1 ;;;; -*- coding: utf-8; fill-column: 78 -*-
2 * minor incompatible change: under weak type checking policy integer
3 types are weakened less aggressively.
4 * minor incompatible change: SAVE-LISP-AND-DIE :TOPLEVEL function is now
5 allowed to return, which causes SBCL to quit with exit status 0. Previously
6 if the function returned with a small integer return value, that value
7 was accidentally reused as the exit status.
8 * new feature: SB-EXT:DEFGLOBAL macro allows defining global non-special
10 * new feature: SB-EXT:ALWAYS-BOUND proclamation inhibits MAKUNBOUND, and
11 allows the compiler to safely elide boundedness checks for special
13 * new feature: SB-EXT:GLOBAL proclamation inhibits SPECIAL proclamations for
14 the symbol, prohibits both lexical and dynamic binding. This is mainly an
15 efficiency measure for threaded platforms, but also valueable in
17 * optimization: the compiler uses a specialized version of FILL when the
18 element type is know in more cases, making eg. (UNSIGNED-BYTE 8) case
20 * optimization: accesses to potentially non-simple arrays where element type
21 is known are 50% faster.
22 * optimization: compiler now generates faster array typechecking code.
23 * optimization: ARRAY-DIMENSION is now faster for multidimensional and
25 * optimization: multidimensional array accesses in the absence of type
26 information regarding array rank are approximately 10% faster due to
27 open coding of ARRAY-RANK.
28 * optimization: result of (FILL (MAKE-ARRAY ...) ...) and (REPLACE
29 (MAKE-ARRAY ...) ...) can be stack allocated if the result of MAKE-ARRAY
31 * optimization: result of call to VECTOR can now be stack allocated.
32 * optimization: MAKE-ARRAY with :INITIAL-CONTENTS is now vastly faster
33 as long as the resulting array is one-dimensional and has a known
34 element type. In particular, :INITIAL-CONTENTS (LIST ...) where the
35 length of the list matches the known length of the vector does not
36 allocate the list as an intermediate step. Ditto for VECTOR and simple
38 * optimization: MAKE-ARRAY can now stack allocate in the presence of
39 :INITIAL-CONTENTS and :INITIAL-ELEMENT as long as the result has a
40 known element type, and is known to be simple and one dimensional.
41 * improvement: SBCL now emits a compiler note where stack allocation was
42 requested but could not be provided (not in all cases, unfortunately)
43 * improvement: better MACHINE-VERSION responses. (thanks to Josh Elsasser)
44 * improvement: pretty-printing loop has been implemented properly. (thanks
46 * documentation: CLOS slot typechecing policy has been documented.
47 * bug fix: better floating point exception handling on x86/OpenBSD.
48 (thanks to Josh Elsasser)
49 * bug fix: exit status from QUIT when called under --script was lost
50 (reported by Hubert Kauker)
51 * bug fix: MAKE-ARRAY for non-zero :INITIAL-ELEMENT always used the
52 same implementation of FILL to initialize the array, even if a more
53 efficient one was available (reported by Stas Boukarev, thanks to
55 * bug fix: potential miscompilation of array stack allocation on x86 and
56 x86-64. (reported by Time Tossavainen)
57 * bug fix: some forms of AND, OR, and COND resulted in expansions that could
58 result in their subforms being treated as top level forms. (reported by
60 * bug fix: On x86/x86-64 alien functions declared to return integers shorter
61 than a machine register could leave garbage in the high bits of the
62 result register (bug 316325).
63 * bug fix: disable address space randomization Linux/x86-64 as well,
64 not just x86-64. (reported by Ken Olum)
65 * bug fix: Attempting to DEREF an (ALIEN (* T)) would produce a WARNING and
66 generate incorrect code.
67 * bug fix: #201; type inference for CONS and ARRAY types could derive
68 wrong results in the presence of eg. RPLACA or ADJUST-ARRAY.
69 * bug fix: special variables with a proclaimed specific subtype of FUNCTION
70 could not be assigned to or bound with PROGV. (reported by Lorenz
72 * bug fix: the value of CL:- in the inspector was the previous expression
73 evaluated rather than the expression being evaluated.
74 * bug fix: constants can no longer be locally declared special.
75 * bug fix: signals delivered to threads started from foreign land (read:
76 directly by pthread_create, not by MAKE-THREAD) are redirected to a Lisp
77 thread by blocking all signals and resignalling.
79 changes in sbcl-1.0.28 relative to 1.0.27:
80 * a number of bugs in cross-compilation have been fixed, with the ultimate
81 result that building under (at least) clisp should be much more reliable.
82 * minor incompatible changes: echo-streams now propagate unread-char to the
83 underlying input stream, and no longer permit unreading more than one
85 * improvement: on x86/x86-64 Lisp call frames now have the same layout as C
86 frames, allowing for instance more reliable backtraces.
87 * improvement: the debugger REPL can now reference lexical variables
88 by name directly for code compiled with (DEBUG 3).
89 * improvement: errors from malformed declarations now have better source
90 paths associated with them. (thanks to Tobias Rittweiler)
91 * optimization: faster local calls on x86/x86-64
92 * bug fix: some error messages for out-of-bound array indexes confused the
93 index and the bound. (thanks to Stas Boukarev)
94 * bug fix: pretty printing malformed DEFPACKAGE forms (thanks to Sidney
96 * bug fix: running regressions tests in shells without OSTYPE set now works.
97 (reported by Harald Hanche-Olsen)
98 * bug fix: more robust static space exhaustion signalling from
99 MAKE-STATIC-VECTOR (thanks to Daniel Lowe)
100 * bug fix: (SETF DOCUMENTATION) for anonymous function now throws the
101 docstring away instead of storing it under names such as (LAMBDA (X)).
102 (reported by Leslie Polzer)
103 * bug fix: timers could go off in the wrong order, be delayed indefinitely
104 (thanks to Ole Arndt for the patch)
105 * bug fix: RESTART-FRAME and RETURN-FROM-FRAME stack corruption
106 * bug fix: the discriminating function for PRINT-OBJECT no longer preserves
107 potentially-invalid effective methods in its cache.
108 * bug fix: SB-INTROSPECT:FIND-DEFINITION-SOURCE now works with funcallable
109 instances as well (thanks to Paul Khuong)
110 * bug fix: using RUN-PROGRAM does not interfere with SB-POSIX:WAIT,
111 SB-POSIX:WAITPID and their C equivalents.
112 * bug fix: RUN-PROGRAM does not crash on Darwin when stressed.
114 changes in sbcl-1.0.27 relative to 1.0.26:
115 * new port: support added for x86-64 OpenBSD. (thanks to Josh Elsasser)
116 * new port: support added for x86-64 Solaris. (thanks to Alex Viskovatoff)
117 * improvement: the system either recovers from stack exhaustion or dies
118 properly as opposed to leaving the user uncertain of whether the handler
119 trampled on some random memory next to the stack or having to rely on
120 --lose-on-corruption (which is still a good idea to use in production
121 because stack exhaustion can happen in signal handlers which will likely
123 * bug fix: fix gc related interrupt handling bug on ppc (regression from
124 1.0.25.37, reported by Harald Hanche-Olsen)
125 * bug fix: work around signal delivery bug in darwin (regression from
126 1.0.25.44, reported by Sidney Markowitz)
127 * bug fix: fix ERROR leaking memory (reported by David Thompson)
129 changes in sbcl-1.0.26 relative to 1.0.25:
130 * incompatible change: an interruption (be it a function passed to
131 INTERRUPT-THREAD or a timer function) runs in an environment where
132 interrupts can be enabled. The interruption can use
133 WITH-INTERRUPTS or WITHOUT-INTERRUPTS as it sees fit. Use
134 WITHOUT-INTERRUPTS to avoid nesting of interruptions and
135 potentially running out of stack. Keep in mind that in the absence
136 of WITHOUT-INTERRUPTS some potentially blocking operation such as
137 acquiring a lock can enable interrupts.
138 * incompatible change: GC-OFF and GC-ON are removed, as they were
139 always unsafe. Use WITHOUT-GCING instead.
140 * new feature: runtime option --disable-ldb
141 * new feature: runtime option --lose-on-corruption to die at the
142 slightest hint of possibly non-recoverable errors: running out of
143 memory, stack, alien stack, binding stack, encountering a memory
144 fault, etc. In the absence of --lose-on-corruption a warning is
146 * enhancement: detect binding stack exhaustion
147 * enhancement: detect alien stack exhaustion on x86/x86-64
148 * improvement: generally more stable and reliable interrupt handling
149 * improvement: there is a per thread interruption queue,
150 interruptions are executed in order of arrival
151 * improvement: a repeating timer reschedules itself when the it has
152 finished, but expiration times are spaced equally. If an
153 expiration time is in the past it will trigger after a short grace
154 period that may give a chance to other things to run.
155 * optimization: slightly faster gc on multithreaded builds
156 * optimization: faster WITHOUT-GCING
157 * bug fix: when JOIN-THREAD signals an error, do it when not holding
158 important locks so that the debugger/handler doesn't produce
159 recursive errors or deadlock.
160 * bug fix: real-time signals are not used anymore, so no more
161 hanging when the system wide real-time signal queue gets full.
162 * bug fix: INTERRUPT-THREAD on a dying thread could produce memory
164 * bug fix: finalizers, gc hooks never run in a WITHOUT-INTERRUPTS
165 * bug fix: fix random memory faults related to interrupts on alpha
166 * bug fix: fix random memory faults related to interrupts on sparc
167 * bug fix: fix deadlocks related to starting threads
168 * bug fix: fix deadlines on locks on futex platforms
169 * bug fix: restore errno in signal handlers
170 * bug fix: fix deadlocks related to hash tables
171 * bug fix: fix deadlocks in pcl
173 changes in sbcl-1.0.25 relative to 1.0.24:
174 * incompatible change: SB-INTROSPECT:FUNCTION-ARGLIST is deprecated, to be
175 removed later. Please use SB-INTROSPECT:FUNCTION-LAMBDA-LIST instead.
176 * new feature: SB-INTROSPECT:DEFTYPE-LAMBDA-LIST allows retrieval of
177 DEFTYPE lambda lists. (thanks to Tobias Rittweiler)
178 * enhancement: MUTEX-VALUE is to be superseded by MUTEX-OWNER that has a
179 better name and does not return values so stale on multiprocessor systems.
180 Also, HOLDING-MUTEX-P was added for about the only sane usage of
182 * improvement: unithread builds keep track of MUTEX-VALUE.
183 * improvement: reading from a TWO-WAY-STREAM does not touch the output
184 stream anymore making it thread safe to have a concurrent reader and
185 a writer, for instance, in a pipe.
186 * improvement: GET-SETF-EXPANDER avoids adding bindings for constant
187 arguments, making compiler-macros for SETF-functions able to inspect
188 their constant arguments.
189 * improvement: COMPILE-FILE reports times with millisecond accuracy
190 (thanks to Luis Oliveira)
191 * optimization: CHAR-CODE type derivation has been improved, making
192 TYPEP elimination on subtypes of CHARACTER work better. (reported
193 by Tobias Rittweiler, patch by Paul Khuong)
194 * bug fix: setting alien structure fields of type struct by value now
195 computes the right offset for the memory copy.
196 * bug fix: compilation problem involving inlined calls to aliens with
197 result type VOID. (reported by Ken Olum)
198 * bug fix: #235a; sequential inline expasion in different policies no
199 longer reuses the functional from the previous expansion site.
200 * bug fix: DEFTYPE no longer breaks on bodies consisting of a single
201 unquoted symbol. Regression from 1.0.22.8. (reported by Ariel Badichi)
202 * bug fix: named ENUMs in multiply-referenced alien record types no longer
203 cause an implied type redefinition. Regression from 1.0.21.29.
204 * bug fix: sign-extension of small signed return values now works with C
205 code compiled with gcc 4.3 or newer. (reported by Liam Healy)
206 * improvements to the Windows port:
207 ** SB-BSD-SOCKETS now works from saved cores as well. (reported by Stephen
208 Westfold, thanks to Rudi Schlatte)
210 changes in sbcl-1.0.24 relative to 1.0.23:
211 * new feature: ARRAY-STORAGE-VECTOR provides access to the underlying data
212 vector of a multidimensional SIMPLE-ARRAY.
213 * new feature: the system now signals a continuable error if standard
214 readtable modification is attempted. (thanks to Tobias Rittweiler)
215 * new feature: DIRECTORY has been extended with a non-standard keyword
216 argument :RESOLVE-SYMLINKS. (thanks to TC-Rucho)
217 * enhancement: SB-BSD-SOCKETS:MAKE-SOCKET-STREAM keyword argument handling
218 has been robustified and documented better. (thanks to Robert Goldman)
219 * optimization: faster generic arithmetic dispatch on x86 and x86-64.
220 * optimization: FORMAT ~D and ~F are now approximately 5% faster.
221 * tradeoff: constant FORMAT control strings are now compiled unless
222 SPACE > SPEED (previously only when SPEED > SPACE.)
223 * bug fix: Red Hat Enterprise 3 mmap randomization workaround. (thanks
225 * bug fix: DEFCLASS and ENSURE-CLASS-USING-CLASS are now expected to
227 * bug fix: lexical type declarations are now correctly reported by
228 SB-CLTL2. (reported by Larry D'Anna)
229 * bug fix: STRING-TO-OCTETS did not handle :START properly when
230 using UTF-8 as external format. (thanks to Luis Oliveira)
231 * bug fix: errors from invalid fill-pointer values to (SETF FILL-POINTER)
232 are signalled correctly. (thanks to Stas Boukarev)
233 * bug fix: SET-MACRO-CHARACTER accepts NIL as the readtable
234 designator. (thanks to Tobias Rittweiler)
235 * bug fix: SET-DISPATCH-MACRO-CHARACTER accepts NIL as the readtable
236 designator, and returns T instead of the function. (thanks to
238 * bug fix: direct superclasses of STANDARD-CLASS and
239 FUNCALLABLE-STANDARD-CLASS now default to STANDARD-OBJECT and
240 FUNCALLABLE-STANDARD-OBJECT as required by AMOP.
241 * bug fix: compiling a call to SLOT-VALUE with a constant slot-name
242 when no class with the named slot yet exists no longer causes a
243 compile-time style-warning.
244 * bug fix: :ALLOCATION :CLASS slots are type-checked properly
245 in safe code. (reported by Didier Verna)
246 * bug fix: #430; nested structure constructors can stack allocate.
247 * bug fix: on some 64-bit platforms dynamic space size was truncated
248 to #xffff0000 bytes. (reported by Benjamin Lambert)
249 * bug fix: setting *READ-SUPPRESS* to T no longer renders the default
250 REPL unusable. (reported by Daniel Herring)
251 * bug fix: return values of READ-SEQUENCE did not take :START into
252 account on file streams, regressions since 1.0.12.22. (reported by
253 Thomas Russ, patch by Paul Khuong)
254 * bug fix: using SET or (SETF SYMBOL-VALUE) to change the value of a
255 method specializer used to confuse permuation vector optimization.
256 * bug fix: system inserted bogus implicit type declarations for local
257 special variables in DEFMETHOD bodies.
258 * bug fix: #354; duplicated frames in backtraces due to
259 non-tail-call-optimized XEPs to functions with return type NIL
261 * bug fix: #357; MAKE-INSTANCE/SHARED-INITIALIZE now
262 initializes structure object slots according to DEFSTRUCT initforms,
263 and DEFSTRUCT forms :INCLUDEind structure classes defined using
264 DEFCLASS :METACLASS STRUCTURE-CLASS now inherit their initforms.
265 (reported by Bruno Haible and Stephen Wilson)
266 * bug fix: #395; fill-pointer output streams used now support
267 element-type BASE-CHAR as well.
268 * bug fix: compiler error when attempting to derive return value of
269 ARRAY-ELEMENT-TYPE when the array type was a union of intersection
271 * bug fix: address-spaces overlapped on OpenBSD. (thanks to Josh
273 * bug fix: Mac OS X binaries should now be portable between Leopard
276 changes in sbcl-1.0.23 relative to 1.0.22:
277 * enhancement: when disassembling method functions, disassembly
278 for the associated fast function is also produced.
279 * enhancement: system stores DEFTYPE lambda-lists, so DESCRIBE can
281 * optimization: printing with *PRINT-PRETTY* true is now more
282 efficient as long as the object being printed doesn't require
283 special handling by the pretty printer.
284 * bug fix: slot symbol-macros from WITH-SLOTS inside DEFMETHOD bodies
285 now interact correctly with type declarations.
286 * partial bug fix: PCL detects infinite recursion during wrapper
287 validation. (thanks to Attila Lendvai)
288 * bug fix: #426; nested function calls are inlined properly.
289 Previously if FOO was an inline function, in calls of the form
290 (FOO (FOO ...)) the outer call was not inlined.
291 * bug fix: long long arguments passed in registers to alien
292 functions on PowerPC Linux and NetBSD are now handled correctly
293 when preceded by SYSTEM-AREA-POINTERs. (reported by Josh Elsasser)
295 changes in sbcl-1.0.22 relative to 1.0.21:
296 * minor incompatible change: LOAD-SHARED-OBJECT no longer by default looks
297 for the shared object in the current directory, but passes the native
298 namestring of the designated pathname to the operation system's shared
299 object loading function as-it.
300 * minor incompatible change: --disable-debugger toplevel option now takes
301 effect before processing of initialization files and --eval or --load
303 * new feature: new commandline argument: --script, which supports
304 shebang lines. See documentation for details. (based on work by
306 * new feature: SAVE-LISP-AND-DIE can save current values of
307 --dynamic-space-size and --control-stack-size in the executable core,
308 causing it to skip normal runtime option processing. See documentation
309 for details. (thanks to Zach Beane)
310 * enhancement: inoccous calls to EVAL or generic functions dispatching
311 on subclasses of eg. STREAM no longer cause compiler notes to appear.
312 * enhancement: the system no longer resignals errors from --load and
313 --eval toplevel arguments as SIMPLE-ERRORS, which caused restarts
314 associated with the original error to be lost. (thanks to Ariel
316 * enhancement: :DONT-SAVE keyword argument has been added to
317 LOAD-SHARED-OBJECT for controlling interaction with
318 SAVE-LISP-AND-DIE, and UNLOAD-SHARED-OBJECT can be used to undo
319 the effects of an earlier LOAD-SHARED-OBJECT call.
320 * bug fix: ADJUST-ARRAY on multidimensional arrays used bogusly give
321 them a fill pointer unless :DISPLACED-TO or :INITIAL-CONTENTS were
322 provided. (reported by Cedric St-Jean)
323 * bug fix: circularity handling in the reader did not treat raw
324 structure slots correctly. (reported by Cedric St-Jean)
325 * bug fix: SERVE-EVENT occasionally signaled an error about bogus
326 file descriptors when there were none.
327 * bug fix: DEFINE-COMPILER-MACRO support of destructuring lambda-lists
328 was broken. (reporteed by Willem Broekema)
329 * bug fix: SB-EXT:NATIVE-NAMESTRING on Windows did not work on
330 pathnames without a directory.
331 * bug fix: COMPILE-FILE did not warn about undefined variable
332 references at toplevel, and LOAD of the resulting fasl did
334 * bug fix: functions with non-required arguments used to end up with
335 (SB-C::&OPTIONAL-DISPATCH ...) as their name.
336 * bug fix: redefining a function with non-required arguments didn't
337 update the system's knowledge about its call signature properly.
338 * bug fix: fixed #431; incompatible alien record type redefinitions
339 are detected and handled. (reported by Neil Haven)
340 * bug fix: using RESTRICT-COMPILER-POLICY with DEBUG 3 could cause
341 PROGV miscompilation. (reported by Matthias Benkard, patch by Juho
343 * bug fix: on x86 and x86-64 backtraces were sometimes truncated
344 after alien stack frames.
345 * bug fix: strings passed to mk{s,d}temp lacked null termination
347 changes in sbcl-1.0.21 relative to 1.0.20:
348 * new feature: the compiler is able to track the effective type of a
349 generic function across method addition and removal.
350 * new feature: SB-EXT:ATOMIC-INCF allows atomic incrementation of
351 appropriately typed structure slots without locking.
352 * new feature: SB-EXT:CALL-WITH-TIMING provides access to timing
353 information like those gathered by TIME using a programming-friendly
355 * new feature: TIME reports time taken even if the form performs a
356 non-local transfer of control.
357 * enhancement: reduced conservativism on GENCGC platforms: on
358 average 45% less pages pinned (measured from SBCL self build).
359 * bug fix: SB-EXT:COMPARE-AND-SWAP on SYMBOL-VALUE can no longer
360 mutate constant symbols or violate declaimed type of the symbol.
361 * bug fix: SB-THREAD:SIGNAL-SEMAPHORE could fail to wakeup threads
362 sleeping on the semaphore in heavily contested semaphores.
363 * bug fix: semaphores and condition variables were not interrupt
365 * bug fix: SB-THREAD:CONDITION-WAIT doesn't allow waits on mutexes
366 owned by other threads anymore.
367 * bug fix: FIND on lists called KEY outside the specified
368 subsequence. (reported by budden)
369 * bug fix: LOG doesn't use single-float intermediate results when
370 given mixed integer and double-float arguments, leading to better
371 precision. (reported by Bob Felts)
372 * bug fix: LOG with base zero returned values of inconsistent type.
374 changes in sbcl-1.0.20 relative to 1.0.19:
375 * minor incompatible change: OPTIMIZE qualities
376 SB-C::STACK-ALLOCATE-DYNAMIC-EXTENT, SB-C::STACK-ALLOCATE-VECTOR,
377 and SB-C::STACK-ALLOCATE-VALUE-CELLS no longer exist. See documentation
378 and SB-EXT:*STACK-ALLOCATE-DYNAMIC-EXTENT* for details.
379 * documentation: some slot access efficiency guidelines have been
380 added to the user manual.
381 * optimization: ASSOC-IF, ASSOC-IF-NOT, MEMBER-IF, MEMBER-IF-NOT,
382 RASSOC, RASSOC-IF, and RASSOC-IF-NOT are now equally efficient
383 as ASSOC and MEMEBER.
384 * optimization: calls to ASSOC, MEMBER, and RASSOC can be transformed
385 to more efficient EQ-comparison versions more often.
386 * optimization: enhanced derivation of DOLIST iteration variable type
388 * optimization: constant folding of simple (LIST ...) forms as DOLIST
390 * optimization: runtime lookup of function definitions can be
391 elided in more cases, eg: (let ((x 'foo)) (funcall foo)).
392 * optimization: compiler is able to derive the return type of
393 (AREF (THE STRING X) Y) as being CHARACTER.
394 * optimization: CLRHASH on empty hash-tables no longer does pointless
395 work. (thanks to Alec Berryman)
396 * optimization: REPLACE deftransforms don't punt when :START1 or
398 * bug fix: compiling DESTRUCTURING-BIND on constant list arguments
399 no longer emits a code deletion note. (reported by Andrew
401 * bug fix: bogus odd-number-of-keywords STYLE-WARNINGs from calls to
402 functions with an odd number of &OPTIONAL arguments, a &REST
403 argument, and one or more &KEY arguments at the call site.
404 * bug fix: STYLE-WARNINGs for functions with both &OPTIONAL and &KEY
405 arguments appeared at call sites as well.
406 * bug fix: fixed #425; CLOSE drops input buffers from streams, so
407 READ-LINE &co can no longer read from them afterwards. (reported
409 * bug fix: fixed #427: unused local aliens no longer cause compiler
410 breakage. (reported by Stelian Ionescu, Andy Hefner and Stanislaw
412 * bug fix: non-local exit from a WITH-ALIEN form no longer causes
413 alien-stack leakage. (reported by Andy Hefner)
414 * bug fix: PROGV signals an error when an attempt to violate declared
415 type of a variable or bind a constant is made.
416 * bug fix: SET signals an error when an attempt to violate declared
417 type of a variable is made.
418 * bug fix: restart computation during the execution of a restart
419 test function no longer causes infinite recursion. (reported by
421 * bug fix: calling SB-COVER:REPORT with a non-directory pathname now
422 signals an error. (thanks to Pierre Mai)
423 * bug fix: EXPORT left symbol unexported in conflict situations.
424 (thanks to Michael Weber)
425 * bug fix: correctly handle name conflicts involving CL:NIL.
426 (thanks to Michael Weber)
427 * bug fix: RESOLVE-CONFLICT restart for name conflicts handles
428 conflicts arising from USEing package with conflicting symbols
429 correctly. (thanks to Michael Weber)
431 changes in sbcl-1.0.19 relative to 1.0.18:
432 * new feature: user-customizable variable SB-EXT:*MUFFLED-WARNINGS*;
433 warnings that go otherwise unhandled will be muffled if they are
434 of the type that's the value of this variable.
435 * optimization: stack allocation is slightly more efficient on x86
437 * bug fix: DEFSTRUCT forms with user-specified :CONSTRUCTOR options,
438 where a raw slot always is initialized using the initform whose
439 type is not know sufficiently well a compile-time are now compiled
440 correctly. (reported by John Morrison)
441 * bug fix: compiler no longer makes erronous assumptions in the
442 presence of non-foldable SATISFIES types.
443 * bug fix: stack analysis missed cleanups of dynamic-extent
444 arguments in non-let-converted calls to local functions.
445 * improvements to the Windows port:
446 ** adjusted address spaces for building on both Win32 and
447 Win64. (thanks for John Connors)
448 * fixed some bugs revealed by Paul Dietz' test suite:
449 ** interval arithmetic during type derivation used inexact integer
450 to single-float coercions.
451 ** arithmetic operations involving large integers and single
452 floats give the same results in compiled and interpreted code.
453 ** deriving the result type of COERCE no longer signals an error
454 if the derived type of the second argument is a MEMBER type
455 containing invalid type specifiers.
456 ** ADJOIN with constant NIL as second argument works correctly.
457 ** FUNCTION-KEYWORDS was called SB-PCL::FUNCTION-KEYWORD-PARAMETERS.
459 changes in sbcl-1.0.18 relative to 1.0.17:
460 * minor incompatible change: SB-SPROF:WITH-PROFILING now by default
461 profiles only the current thread.
462 * minor incompatible change: changes to SYMBOL-VALUE of constants
463 defined with DEFCONSTANT now signal an error.
464 * enhancement: SB-SPROF now has support for wallclock profiling,
465 and is also able to profile specific threads. REPORT output
466 has also additional sorting options.
467 * enhancement: better pretty-printing of DEFPACKAGE forms. (Thanks
469 * optimization: structure allocation has been improved
470 ** constructors created by non-toplevel DEFSTRUCTs are ~40% faster.
471 ** out of line constructors are ~10% faster.
472 ** inline constructors are ~15% faster.
473 ** inline constructors are capable of dynamic extent allocation
474 (generally on x86 and x86-64, in some cases on other platforms
476 * optimization: simple uses of HANDLER-CASE and HANDLER-BIND no
478 * optimization: file compiler is now able to coalesce non-circular
479 lists, non-base strings, and bit-vectors. Additionally, constants
480 are never referenced through SYMBOL-VALUE at runtime.
481 * optimization: code defining methods on PRINT-OBJECT (and other
482 generic functions in the COMMON-LISP package) now loads faster.
483 * bug fix: EAI_NODATA is deprecated since RFC 3493. Stop using it
485 * bug fix: if COMPILE-FILE aborts due to an unwind, the partial
486 fasl is now deleted. (reported by Attila Lendvai)
487 * bug fix: READ-LINE always returned NIL for the last line in files.
488 (reported by Yoshinori Tahara)
489 * bug fix: more accurate disassembly annotations of foreign function
490 calls. (thanks to Andy Hefner)
491 * bug fix: trimming non-simple strings and non-string string
492 designators when the there is nothing to trim works properly.
493 (thanks to James Knight)
494 * new feature: SB-POSIX bindings for mlockall, munlockall, and setsid.
495 (thanks to Travis Cross)
496 * fixed some bugs revealed by Paul Dietz' test suite:
497 ** NIL is a valid function name (regression at 1.0.13.38)
498 ** FILL on lists was missing its return value (regression at 1.0.12.27)
499 ** STRING-TRIM, STRING-LEFT-TRIM, and STRING-RIGHT-TRIM did not respect
500 fill pointers properly (regression at 1.0.12.23)
501 ** STRING-TRIM, STRING-LEFT-TRIM, and STRING-RIGHT-TRIM did not respect
502 displacement indices properly (regression at 1.0.12.23)
504 changes in sbcl-1.0.17 relative to 1.0.16:
505 * temporary regression: user code can no longer allocate closure
506 variable storage on stack, due to bug 419 without explicitly
507 requesting it. Please consult sbcl-devel for advice if you need to
508 use this feature in the meanwhile.
509 * new feature: runtime argument --control-stack-size can be used to
510 adjust thread default control stack size.
511 * enhancement: improved TIME output
512 ** all times are reported using the measured accuracy (milliseconds
513 for real and GC times, microseconds for everything else.)
514 ** processor cycle counts on x86 and x86-64.
515 ** interpreted forms are counted for both evaluator modes.
516 ** number of lambdas converted by the compiler is reported.
517 ** CPU percentage report (computed from real and total run time.)
518 ** more comprehensive run time reporting, using a condenced format
519 ** interperted form, lambda, and page fault counts are omitted
521 * optimization: ADJOIN and PUSHNEW are upto ~70% faster in normal
523 * optimization: APPEND is upto ~10% faster in normal SPEED policies.
524 * optimization: two argument forms of LAST are upto ~10% faster
525 in normal SPEED policies.
526 * optimization: NCONC no longer needs to heap cons its &REST list
527 in normal SPEED policies.
528 * bug fix: SB-FLUID build feature no longer breaks the build. (thanks
530 * bug fix: UNION and NUNION work with :TEST-NOT once more,
531 regression since 1.0.9.1. (thanks to Eric Marsden)
532 * bug fix: result of MAKE-ARRAY can be stack allocated - regression
533 since 1.0.15.36. (thanks to Paul Khuong)
534 * bug fix: LAST when always returned the whole list when given a bignum
535 as the second argument.
536 * bug fix: dynamic extent allocation of nested lists and vectors
537 could leak to otherwise accessible parts.
538 * bug fix: invalid optimization of heap-allocated alien variable
540 * bug fix: fasl header checking is less vulnerable to different
541 platform word lengths.
542 * bug fix: more correct assembler syntax for GNU binutils
543 2.18.50.0.4 support. (thanks to Marijn Schouten)
544 * bug fix: fix ECASE warnings from CMUCL-as-xc-host. (reported by
546 * bug fix: the fopcompiler can handle LOCALLY forms (with no
547 declarations) successfully. (reported by Attila Lendvai)
549 changes in sbcl-1.0.16 relative to 1.0.15:
550 * minor incompatible change: revert the changes to sb-posix's error
551 signaling added in 1.0.14.
552 * minor incompatible change: change PROBE-FILE back to returning
553 NIL whenever we can't get a truename, as was the case before 1.0.14.
554 * minor incompatible change: SB-BSD-SOCKETS:NAME-SERVICE-ERROR now
555 inherits from ERROR instead of just CONDITION.
556 * new feature: SB-INTROSPECT can provide source locations for instances
557 as well. (thanks to Tobias Rittweiler)
558 * optimization: binding special variables now generates smaller code
559 on threaded platforms.
560 * optimization: MEMBER and ASSOC are over 50% faster for :TEST #'EQ
561 and cases where no :TEST is given but the compiler can infer that
562 the element to search is of type (OR FIXNUM (NOT NUMBER)).
563 * optimization: better LOGNOT on fixnums.
564 * optimization: modular arithmetic for a particular requested width
565 is implemented using a tagged representation unless a better
566 representation is available.
567 * fixed bug 423: TRULY-THE and *CHECK-CONSISTENCY* interaction.
568 * bug fix: SB-BSD-SOCKETS:MAKE-INET-ADDRESS checks the input string
569 for wellformedness and returns a specialized vector. (reported by
570 Francois-Rene Rideau)
571 * bug fix: FIND-CLASS was not thread-safe. (reported by Attila Lendvai)
572 * bug fix: ~R was broken for vigtillions. (thanks to Luis Oliveira)
573 * bug fix: attempt to obtain *SCHEDULER-LOCK* recursively when
574 unscheduling timer at the same time as another timer fires.
575 * bug fix: don't reschedule timers for dead threads.
576 * bug fix: periodic polling was broken. (thanks to Espen S Johnsen)
577 * bug fix: copying output from RUN-PROGRAM to a stream signalled
578 bogus errors if select() was interrupted.
579 * enhancement: add support for fcntl's struct flock to SB-POSIX.
581 changes in sbcl-1.0.15 relative to sbcl-1.0.14:
582 * enhancement: cleaner backtraces for interactive interrupts, as
583 well as other cases where the interesting frames used to be
584 obscured by interrupt handling frames.
585 * enhancement: untracing a whole package using (UNTRACE "FOO") is
586 now supported, and tracing a whole package using (TRACE "FOO") now
587 traces SETF-functions as well.
588 * enhancement: implement SB-POSIX:MKTEMP and SB-POSIX:MKDTEMP.
589 * SB-DEBUG:PRINT-FRAME-CALL now prints the entry-point kind even
590 when SB-DEBUG:*SHOW-ENTRY-POINT-DETAILS* is NIL.
591 * unreadably printed representation of hash-tables now includes
593 * bug fix: partially fixed #188: type propagation from assignments
594 is now more efficient.
595 * bug fix: fixed #407: (COERCE X 'SINGLE-FLOAT) and (COERCE X
596 'DOUBLE-FLOAT) are not flushable.
597 * bug fix: on x86 and x86-64 pointer based EQ-hashing now uses the
598 full address of the object, and none of the tag bits.
599 * bug fix: readably printing hash-tables now respects other printer
600 control variables. (reported by Cedric St-Jean)
601 * bug fix: compiler gave a bogus STYLE-WARNING for the :SYNCHRONIZED
602 keyword with MAKE-HASH-TABLE.
603 * bug fix: export SB-POSIX:MKSTEMP.
604 * bug fix: SORT was not interrupt safe.
605 * bug fix: XREF accounts for the last node of each basic-block as
607 * bug fix: MAKE-INSTANCE optimizations interacted badly with
608 non-keyword :DEFAULT-INITARGS in the presence of :BEFORE/:AFTER
609 methods on SHARED-INITIALIZE. (thanks to Matt Marjanovic)
610 * bug fix: the CTOR optimization for MAKE-INSTANCE should no longer
611 create obsolete instances in the case of redefinition or
612 obsoletion of a superclass. (thanks to Andy Hefner)
613 * bug fix: Support for the Alpha architecture has been revived; it had
614 suffered somewhat from lack of maintenance since sbcl-1.0.
615 * improvements to the Windows port:
616 ** The system detects the codepage to use at startup, instead of
617 using the value from the saved core. (thanks to Kei Suzuki)
619 changes in sbcl-1.0.14 relative to sbcl-1.0.13:
620 * new feature: SB-EXT:*EXIT-HOOKS* are called when the process exits
621 (see documentation for details.)
622 * revived support for OpenBSD (contributed by Josh Elsasser)
623 * partially fixed bug #108: ROOM no longer suffers from occasional
624 (AVER (SAP= CURRENT END)) failures.
625 * fixed bug #402: proclaimed non-standard declarations in DEFMETHOD
626 bodies no longer cause a WARNING to be signalled. (reported by
628 * bug fix: (TRUNCATE X 0) when X is a bignum now correctly signals
629 DIVISION-BY-ZERO. Similarly for MOD and REM (which suffered due to
630 the bug in TRUNCATE.) (reported by Michael Weber)
631 * bug fix: SB-SPROF:REPORT no longer signals an error if there are
632 no samples. (reported by Andy Hefner)
633 * bug fix: functions compiled using (COMPILE NIL '(LAMBDA ...))
634 no longer appear as (NIL ...) frames in backtraces.
635 * bug fix: RESOLVE-CONFLICT (and the other name conflict machinery)
636 is now actually exported from SB-EXT as documented. (reported by
638 * bug fix: sb-aclrepl now correctly understands how to inspect
639 single-floats on 64-bit platforms where single-floats are not boxed.
640 * bug fix: SB-MOP:CLASS-SLOTS now signals an error if the class has not
641 yet been finalized. (reported by Levente Meszaros)
642 * bug fix: CLOSE :ABORT T behaves more correctly on Windows.
643 * DESCRIBE and (DOCUMENTATION ... 'OPTIMIZE) describe meaning of
644 SBCL-specific optimize qualities.
646 changes in sbcl-1.0.13 relative to sbcl-1.0.12:
647 * minor incompatible change: RUN-PROGRAM now uses execvp(3) to find
648 an executable in the search path, and does so in the child
649 process's PATH. The function FIND-EXECUTABLE-IN-SEARCH-PATH has
650 been removed; it can be found in the CVS history, for somebody who
651 needs that search behavior (see the manual).
652 * minor incompatible change: compiler policy re. weakening type
653 checks has changed: now type checks are weakened only if SAFETY < 2
655 * SB-EXT:NATIVE-NAMESTRING takes a new keyword AS-FILE, forcing
656 unparsing of directory pathnames as files. Analogously,
657 SB-EXT:PARSE-NATIVE-NAMESTRING takes an AS-DIRECTORY, forcing a
658 filename to parse into a directory pathname.
659 * enhancement: implicit generic function creation now signals a
660 specific style-warning IMPLICIT-GENERIC-FUNCTION-WARNING, which
661 users can bind handlers for and muffle around calls to LOAD.
662 * enhancement: RUN-PROGRAM allows unicode arguments and environments
663 to be used (using the default stream external format), and allows
664 non-simple strings to be used. (thanks to Harald Hanche-Olsen)
665 * optimization: COPY-SEQ, FILL, and SUBSEQ are 30-80% faster for
666 strings and vectors whose element-type or simplicity is not fully
667 known at compile-time.
668 * optimization: STRING-TRIM and related functions no longer allocate
669 a new string when no trimming needs to be performed. These functions
670 are also faster than before when the input string has been declared
672 * optimization: READ-SEQUENCE on simple-strings is up to 80% faster.
673 * optimization: READ-LINE is significantly faster for files containing
675 * optimization: non-open coded uses of character comparison operators
676 (e.g. char=) no longer cons when called with more than one parameter
677 on platforms supporting dynamic-extent allocation.
678 * bug fix: READ-SEQUENCE on composite stream wrapping a Gray stream
679 with STREAM-ELEMENT-TYPE (UNSIGNED-BYTE 8) signalled an error.
680 * bug fix: COPY-SEQ on lists did not signal a type-error on improper
682 * bug fix: some sequence functions elided bounds checking when
684 * bug fix: too liberal weakening of union-type checks when SPEED >
686 * bug fix: more bogus fixnum declarations in ROOM implementation
689 changes in sbcl-1.0.12 relative to sbcl-1.0.11:
690 * new feature: MAKE-HASH-TABLE now experimentally accepts a
691 :SYNCHRONIZED argument, which makes the hash-table safe for
692 concurrent accesses (but not iteration.) See also:
693 SB-EXT:WITH-LOCKED-HASH-TABLE, and
694 SB-EXT:HASH-TABLE-SYNCHRONIZED-P.
695 * optimization: CONCATENATE on strings is an order of magnitude faster
696 in code compiled with (> SPEED SPACE).
697 * optimization: SUBSEQ is ~50% faster on lists.
698 * bug fix: bug 417 fixed -- source location reporting is now more robust.
699 * bug fix: SUBSEQ on a list will now correctly signal an error if if
700 END is smaller then START.
701 * bug fix: SB-PROFILE will no longer report extra consing for nested
702 calls to profiled functions.
703 * bug fix: ROOM implementation had bogus fixnum declarations which
704 could cause type-errors when calling ROOM on large images.
705 * bug fix: if file compilation is aborted, the partial fasl is now
706 deleted, and COMPILE-FILE returns NIL as the primary value.
707 * bug fix: number of thread safety issues relating to SBCL's internal
708 hash-table usage have been fixed.
709 * bug fix: SB-SYS:WITH-PINNED-OBJECTS could cause garbage values to
710 be returned from its body when the values were being returned
711 using unknown-values return convection and the W-P-O was wrapped
712 inside an UNWIND-PROTECT.
713 * bug fix: sb-posix should now compile again under Windows, enabling
716 changes in sbcl-1.0.11 relative to sbcl-1.0.10:
717 * incompatible change: hash-table accessor functions are no longer
718 automatically protected by locks. Concurrent accesses on the same hash-table
719 from multiple threads can give inconsistent results or even corrupt the
720 hash-table completely. Multi-threaded applications should do their own
721 locking at the correct granularity. In the current implementation it is
722 still safe to have multiple readers access the same table, but it's not
723 guaranteed that this property will be maintained in future releases.
724 * minor incompatible change: (SETF MUTEX-VALUE) is not longer supported,
725 and will signal an error at runtime.
726 * enhancement: SB-THREAD package now exports a semaphore interface.
727 * enhancement: CONS can now stack-allocate on x86 and
728 x86-64. (Earlier LIST and LIST* supported stack-allocation, but
730 * enhancement: nested lists can now be stack allocated on
731 platforms providing stack allocation support.
732 * enhancement: dynamic-extent support has been extended to support
733 cases where there are multiple possible sources for the stack
735 * optimization: RELEASE-MUTEX no longer needs to perform a syscall
736 if the mutex is uncontested on Linux.
737 * bug fix: symbol-macro expansion now uses the *MACROEXPAND-HOOK*
738 as specified by the CLHS. (thanks to Tobias Rittweiler)
739 * bug fix: NaN comparison now works on x86-64.
740 * bug fix: CLOSE :ABORT T on a stream with pending output now
742 * bug fix: instances of non-standard metaclasses using standard
743 instance structure protocol sometimes missed the slot type checks
745 * bug fix: known functions can be inlined in high-debug code.
747 changes in sbcl-1.0.10 relative to sbcl-1.0.9:
748 * minor incompatible change: the MSI installer on Windows no longer
749 associates .lisp and .fasl files with the installed SBCL.
750 * minor incompatible change: :UNIX is no longer present in *FEATURES*
751 on Windows. (thanks to Luis Oliviera)
752 * new feature: SB-CLTL2 contrib module now implements
753 FUNCTION-INFORMATION. (thanks to Larry D'Anna)
754 * optimization: scavenging weak pointers is now more efficient,
755 requiring O(1) instead of O(N) per weak pointer to identify
756 scanvenged vs. unscavenged pointers. (thanks to Paul Khuong)
757 * optimization: typechecking (SETF SLOT-VALUE) is now ~5x faster in
758 method bodies using constant slot-names when the first argument is
759 a specializer parameter for the method.
760 * optimization: SLOT-VALUE &co are now ~50% faster for variable slot
761 names, when the class of the instance is a direct instance
762 STANDARD-CLASS or FUNCALLABLE-STANDARD-CLASS (making them only 3x
763 as slow as the constant slot-name case.)
764 * optimization: member type construction is now O(N) instead
766 * optimization: UNION and NUNION are now O(N+M) for large
767 inputs as long as the :TEST function is one of EQ, EQL, EQUAL, or
769 * enhancement: DEFINE-MODIFY-MACRO lambda-list information is
770 now more readable in environments like Slime which display it.
771 (thanks to Tobias C. Rittweiler)
772 * bug fix: SB-EXT:COMPARE-AND-SWAP was non-atomic unless the compiler
773 was able to infer the correct argument type for the object on which
774 the CAS operation was being performed.
775 * bug fix: copy propagation interfered with parallel assignment
776 semantics in local calls. (reported by Paul Khuong)
777 * bug fix: the signed modular fixnum shift compiled to wrong code on
778 x86 and x86-64. (spotted by a slight modification to some of
781 changes in sbcl-1.0.9 relative to sbcl-1.0.8:
782 * minor incompatible change: SB-SYS:OUTPUT-RAW-BYTES is deprecated.
783 * enhancement: SB-EXT:FINALIZE accepts a :DONT-SAVE keyword argument,
784 indicating the finalizer should be cancelled when SAVE-LISP-AND-DIE
786 * enhancement: default PRINT-OBJECT methods for classes, slot-definitions,
787 and generic function now print the object identity for anonymous
788 instances. (thanks to Attila Lendvai)
789 * enhancement: as an extension to MOP, SBCL now supports SETF of
790 STANDARD-INSTANCE-ACCESS and FUNCALLABLE-STANDARD-INSTANCE-ACCESS.
791 (thanks to Attila Lendvai)
792 * enhancement: sb-cover has an alternate annotation mode which puts
793 the form annotations at the car of the form, instead of annotating
794 it completely. The benefit of this mode is that it shows better
795 which forms were instrumented by the compiler.
796 * bug fix: new compiler transforms for MEMBER and ASSOC were affected
797 by printer control variables. (reported by Dan Corkill)
798 * bug fix: system leaked memory when delayed output was performed by
799 the OS in smaller chunks then expected. (thanks to David Smith)
800 * bug fix: system leaked memory when file streams were not closed
802 * bug fix: large objects written to slow streams that were modified
803 after the write could end up with the modified state written to
804 the underlying file descriptor.
805 * bug fix: multiple threads operating in parallel on the same stream
806 could cause buffer-overflows.
807 * bug fix: source location information is stored correctly
808 (broken since 1.0.6). This bug would generally show up as the
809 Slime debugger highlighting the wrong form when the sldb-show-source
811 * bug fix: Branch forms are again annotated as branches in the sb-cover
813 * bug fix: GCD on MOST-NEGATIVE-FIXNUM no longer causes an infinite loop
814 on x86-64. (reported by Gregory Vanuxem)
815 * bug fix: EQUALP could return wrong results for structures with raw
816 slots (slots with a :TYPE of SINGLE-FLOAT, DOUBLE-FLOAT, or a machine
817 word). (reported by Vjacheslav Fyodorov)
818 * bug fix: sb-sprof sampling didn't work on non-x86oid platforms. (patch
821 changes in sbcl-1.0.8 relative to sbcl-1.0.7:
822 * enhancement: experimental macro SB-EXT:COMPARE-AND-SWAP provides
823 atomic compare-and-swap operations on threaded platforms.
824 * enhancement: experimental function SB-EXT:RESTRICT-COMPILER-POLICY
825 allows assigning a global minimum value to optimization qualities
826 (overriding proclamations and declarations).
827 * enhancement: closed over variables can be stack-allocated on x86
829 * performance bug fix: GETHASH and (SETF GETHASH) are once again
831 * optimization: slot definition lookup is now O(1). This speeds up
832 eg. SLOT-VALUE and (SETF SLOT-VALUE) with variable slot names.
833 * optimization: STRING-TO-OCTETS is now up to 60% faster for UTF-8.
834 * optimization: ASSOC and MEMBER can now be open-coded for all
835 combinations of keyword arguments when second argument is constant
836 and SPEED >= SPACE. In other cases a specialized version is
838 * bug fix: using obsoleted structure instances with TYPEP and
839 generic functions now signals a sensible error.
840 * bug fix: threads waiting on GET-FOREGROUND can be interrupted.
841 (reported by Kristoffer Kvello)
842 * bug fix: backtrace construction is now more careful when making
843 lisp-objects from pointers on the stack, to avoid creating bogus
844 objects that can be seen by the GC.
845 * bug fix: defaulting of values in contexts expecting more than 7
846 variables now works on x86-64. (reported by Christopher Laux)
847 * bug fix: modifications to packages (INTERN, EXPORT, etc) are now
849 * bug fix: (SETF SYMBOL-PLIST) no longer allows assigning a non-list
850 as the property-list of a symbol.
851 * bug fix: DEFMETHOD forms with CALL-NEXT-METHOD in the method body,
852 in EVAL-WHEN forms with both :COMPILE-TOPLEVEL and :LOAD-TOPLEVEL
853 situations requested, are once again file-compileable. (reported
856 changes in sbcl-1.0.7 relative to sbcl-1.0.6:
857 * MOP improvement: support for user-defined subclasses of
858 SB-MOP:SPECIALIZER has been enhanced. The experimental interface
859 function SB-PCL:MAKE-METHOD-SPECIALIZERS-FORM, called as part of
860 the expansion of DEFMETHOD, is responsible for generating a form
861 which creates a list of specializers when evaluated. Additional
862 functions SB-PCL:[UN]PARSE-SPECIALIZER-USING-CLASS provide
863 debugging and introspective support.
864 * minor incompatible change: the (unsupported) spinlock interface
865 has changed: free spinlock now has the value NIL, and a held spinlock
866 has the owning thread as its value.
867 * enhancement: WITHOUT-INTERRUPTS now binds ALLOW-WITH-INTERRUPTS and
868 WITH-LOCAL-INTERRUPTS as local macros. Refer to documentation string
870 * enhancement: name of a socket-stream is now "a socket" instead of
872 * enhancement: SB-POSIX now supports lockf(). (Thanks to Zach Beane.)
873 * enhancement: SB-POSIX now supports getcwd(). (Thanks to Tassilo Horn.)
874 * enhancement: SB-INTROSPECT:FUNCTION-ARGLIST shows nicer argument lists
875 for generic functions. (Thanks to Tobias C. Rittweiler)
876 * optimization: bignum printing speed has been improved by 20-40%
877 (depending on the bignum size.)
878 * bug fix: WITH-MUTEX and WITH-RECURSIVE-LOCK are now interrupt safe
880 * bug fix: the cache used by the CLOS to store precomputed effective
881 methods, slot offsets, and constant return values is now thread and
883 * bug fix: generic function dispatch function updating is now thread
884 and interrupt safe (in the sense that the known issues have been
886 * bug fix: ADD/REMOVE-METHOD is now thread and interrupt safe.
887 * bug fix: interning EQL-specializers is now thread and interrupt safe.
888 * bug fix: asdf systems with dependencies to the SB-POSIX or
889 SB-BSD-SOCKETS contribs can be loaded with :FORCE T.
890 * bug fix: interrupt safety of applicable method computation has been
893 changes in sbcl-1.0.6 relative to sbcl-1.0.5:
894 * new contrib: sb-cover, an experimental code coverage tool, is included
896 * optimization: STRING-TO-OCTETS for unibyte encodings and UTF-8 is
897 significantly faster.
898 * optimization: calls of the form (AREF FOO (+ INDEX <constant>)) now
899 produce more efficient code on x86-64 if the compiler can determine
900 that (+ INDEX <constant>) does not require a bounds check and FOO
901 has an element type at least 8 bits wide.
902 * enhancement: a new, experimental synchronous timeout facility is
903 provided. Refer to SB-SYS:WITH-DEADLINE for details.
904 * enhancement: when a symbol name conflict error arises, the
905 conflicting symbols are always printed with a package prefix.
906 (thanks to Kevin Reid)
907 * enhancement: stepping is now once again supported on the SPARC and
908 MIPS platforms. (It is also now more likely to work on CheneyGC
910 * enhancement: sb-sprof can now also track and report accurate call
912 * bug fixes: the treatment of non-standard subclasses of
913 SB-MOP:SPECIALIZER is more correct.
914 * incompatible change: PURIFY no longer copies the data from the
915 dynamic space into the static and read-only spaces on platforms
916 that use the generational garbage collector
917 * bug fix: GETHASH, (SETF GETHASH), CLRHASH and REMHASH are now
919 * bug fix: GC race condition occasionally resulting in crashes with
920 the error message "SIG_STOP_FOR_GC blocked at a bad place" has been
922 * bug fix: &ENVIRONMENT variables in macro lambda lists can now be
924 * bug fix: DEFSETF lambda lists without &ENVIRONMENT no longer cause
925 a STYLE-WARNING to be signalled (regression from 1.0.4.)
926 * bug fix: an asynchronous interrupt could previously leave the
927 system running with GC inhibited.
928 * bug fix: a DECLARE form evaluated at top-level now causes an error
929 rather than silently (or verbosely) returning NIL.
930 * bug fix: trying to dissassemble functions compiled at high DEBUG could
931 result in a error being signalled due to source form lookup errors.
932 (reported by Peter Graves)
934 changes in sbcl-1.0.5 relative to sbcl-1.0.4:
935 * incompatible change: removed writer methods for host-ent-name,
936 host-ent-addresses -- changing the values did not update the DNS
938 * minor incompatible change: changed experimental JOIN-THREAD interface
939 * documentation: the manual now lists reader and writer methods
940 in class slot documentation sections. (thanks to Richard M Kreuter)
941 * documentation: unwinding from asyncronous events has been
942 documented as unsafe.
943 * documentation: SB-SYS:WITHOUT-GCING has been documented as unsafe
944 in multithreaded application code.
945 * optimization: GET-INTERNAL-REAL-TIME has been optimized on POSIX
946 platforms. (thanks to James Anderson for the optimization hint)
947 * optimization: REPLACE, SUBSEQ, and COPY-SEQ are now optimized in
949 * optimization: STRING-{EQUAL,LESSP,GREATER-P} and their NOT-
950 variants no longer cons.
951 * optimization: Direct calls to CHAR-{EQUAL,LESSP,GREATERP} and
952 their NOT- variants no longer cons.
953 * optimization: EQUAL hash tables no longer use SXHASH for objects
954 of all data types, but instead use an EQL hash for types for which
955 EQUAL is the same as EQL.
956 * optimization: the non-inlined generic versions of AREF and (SETF AREF)
957 are significantly faster.
958 * optimization: new STRING-OUTPUT-STREAM implementation is much
959 faster for large outputs and conses 30% less on average.
960 * enhancement: XREF information is now collected for references made
961 to global variables using SYMBOL-VALUE and a constant argument.
962 * enhancement: SIGINT now causes a specific condition
963 SB-SYS:INTERACTIVE-INTERRUPT to be signalled.
964 * bug fix: FILE-POSITION used to signal an error in some cases where
965 ANSI requires it to return NIL.
966 * bug fix: ADJUST-ARRAY is now interrupt-safe.
967 * bug fix: adding and removing fd-handlers is now interrupt-safe.
968 * bug fix: inlined calls to C now ensure 16byte stack alignment on
970 * bug fix: bad type declaration in the CLOS implementation has
971 been fixed. (reported by James Anderson)
972 * bug fix: incorrect ROOM reporting on x86-64 has been fixed.
973 (thanks to Lutz Euler)
974 * bug fix: DEFSETF now allows &ENVIRONMENT and disallows &AUX as
975 required by the CLHS. (reported by Samium Gromoff)
976 * bug fix: dead unbound variable references now signal an error.
977 (reported by Marco Monteiro)
978 * bug fix: / with an unused value was being deleted in safe code.
979 (thanks to Marco Monteiro and Kevin Reid)
980 * bug fix: number of characters that can be written onto a single
981 line in a file is unlimited.
982 * bug fix: some GC deadlocks caused by asynchronous interrupts have
983 been fixed by inhibiting interrupts for when GC is disabled.
984 * bug fix: some interrupt safety issues with GETHASH, (SETF
985 GETHASH), CLRHASH and REMHASH have been fixed.
986 * bug fix: binding *BREAK-ON-SIGNALS* to a value that is not a type
987 specifier no longer causes infinite recursion.
988 * bug fix: SB-EXT:MUFFLE-CONDITIONS declarations no longer trigger a
989 bogus warning in DEFMETHOD bodies. (reported by Kevin Reid)
990 * bug fix: an &environment argument with the correct variable information
991 is passed to macros that are expanded during byte compilation.
992 (reported by Samium Gromoff)
993 * bug fix: a logic error was causing the hash function for bignums to
994 have excessive amounts of collisions. (reported by Faré Rideau)
995 * bug fix: modifying the contents of an array could change the return
996 value of SXHASH on that array, which is only allowed for strings
997 and bit vectors. (bug introduced in 0.9.16)
998 * improvement: the x86-64/darwin port now passes all tests (except
999 for the debugger tests) but should still be considered
1000 experimental until this is fixed.
1001 * improvement: a style-warning is signaled for CASE (etc) clauses with
1002 duplicate keys. (thanks to Kevin Reid)
1003 * improvement: macos/x86 and macos/x86-64 now use mach exception
1004 handlers for dealing with illegal instructions (for trapping and
1005 error handling) and memory protection violations (for GC).
1007 changes in sbcl-1.0.4 relative to sbcl-1.0.3:
1008 * new platform: experimental support for x86-64/darwin (MacOS).
1009 * incompatible change: the thread-safe (on most platforms) getaddrinfo
1010 and getnameinfo sockets functions are used instead of gethostbyaddr
1011 and gethostbyname, on platforms where the newer functions are available.
1012 As a result, the ALIASES field of HOST-ENT will always be NIL on these
1014 * change: runtimes with embedded cores (i.e. saved with :EXECUTABLE T)
1015 don't print the startup banner, but behave as if --noinform was passed
1016 as a command line argument. (thanks to Kevin Reid)
1017 * new experimental feature: added JOIN-THREAD (by NIIMI Satoshi)
1018 * optimization: code using alien values with undeclared types is much faster.
1019 * optimization: the compiler is now able to open code SEARCH in more cases.
1020 * optimization: more compact typechecks on x86-64 (thanks to Lutz Euler)
1021 * bug fix: using standardized COMMON-LISP special variables as loop
1022 variables no longer signals bogus package lock violations. (reported
1024 * bug fix: declaring local loop variables to be of a range-limited type
1025 such as (SINGLE-FLOAT 1.0 2.0) no longer causes a compile-time error.
1026 (reported by Andras Simon)
1027 * bug fix: >= and <= gave wrong results when used with NaNs. (Some NaN
1028 bugs remain on x86-64.)
1029 * bug fix: the #= and ## reader macros now interact reasonably with
1030 funcallable instances.
1031 * bug fix: type-checks for function arguments were compiled using the
1032 compiler policy settings of the wrong lexical environment (in 1.0.2
1034 * bug fix: SHADOW accepts characters as string designators, as required
1035 by the spec (thanks to Eric Marsden)
1036 * bug fix: fixed GC safety issues when foreign functions are called with
1037 non-base strings as arguments
1038 * bug fix: more consistent error output for fatal-compiler-warnings, like
1040 * bug fix: functions with proper names were showing up as NIL in
1041 backtraces (reported by Edi Weitz, regression in 1.0.2)
1043 changes in sbcl-1.0.3 relative to sbcl-1.0.2:
1044 * new platform: NetBSD/PPC. (thanks to Aymeric Vincent)
1045 * optimization: calls of the form (AREF FOO (+ INDEX <constant>)) now
1046 produce more efficient code on the x86 if the compiler can determine
1047 that (+ INDEX <constant>) does not require a bounds check and FOO
1048 has an element type at least 8 bits wide.
1049 * bug fix: references to undefined variables are handled the same way
1050 in toplevel forms as in the normal compiler
1051 * bug fix: the build scripts again work with non-bash /bin/sh (thanks
1053 * bug fix: use "gtar" as the asdf-install *TAR-PROGRAM* on NetBSD
1054 (thanks to Jon Buller)
1055 * improvement: faster compilation times for complex functions
1056 * improvement: added readlink support to SB-POSIX (thanks to Richard
1059 changes in sbcl-1.0.2 relative to sbcl-1.0.1:
1060 * improvement: experimental support for mach exception handling on
1061 x86/macos. requires building with :MACH-EXCEPTION-HANDLER feature
1063 * improvement: support for GBK external format.
1064 (thanks to Chun Tian (binghe))
1065 * improvement: the debugger now displays variables that have been closed
1066 over, in code compiled with (DEBUG 2) or higher
1067 * improvement: support for executable cores on NetBSD (thanks to
1069 * new feature: added a RESTART-FRAME debugger command
1070 * new feature: new generic function SB-GRAY:STREAM-FILE-POSITION can
1071 be used to provide an implementation for FILE-POSITION on Gray streams
1072 (thanks to Eric Marsden)
1073 * new feature: add syslog, openlog and closelog support to SB-POSIX
1074 (thanks to Richard Kreuter)
1075 * optimization: the function call overhead in code compiled with
1076 a high DEBUG optimization setting is significantly
1077 * bug fix: an error is signaled for attempts to use READ-SEQUENCE
1078 for a (SIGNED-BYTE 8) stream and (UNSIGNED-BYTE 8) vector, or vice versa.
1079 (thanks to Tony Martinez)
1080 * bug fix: the initforms for DEFMETHOD &AUX parameters are only
1081 evaluated once (reported by Kevin Reid)
1082 * bug fix: the :SHOW-PROGRESS keyword parameter to SB-SPROF:WITH-PROFILING
1083 works again (thanks to Kilian Sprotte)
1084 * bug fix: an error is signaled for tagbodies with duplicate tags
1085 (thanks to Stephen Wilson)
1086 * bug fix: NIL can be used as a tagbody tag (thanks to Stephen Wilson)
1087 * bug fix: Win32 port can now handle foreign code unwinding Lisp
1088 stack frames from alien callbacks.
1089 * bug fix: ATANH returned incorrect results on win32 (thanks to Pierre Mai)
1090 * bug fix: SBCL works on Linux/ppc systems with a kernel configured to use
1091 65k pages (thanks to David Woodhouse)
1092 * bug fix: fix SB-POSIX dirent and socket on NetBSD (thanks to Richard Kreuter)
1094 changes in sbcl-1.0.1 relative to sbcl-1.0:
1095 * new platform: FreeBSD/x86-64, including support for threading.
1096 * new feature: the compiler stores cross-referencing information
1097 about function calls (who-calls), macroexpansion (who-macroexpands)
1098 and special variables (who-binds, who-sets, who-references) for code
1099 compiled with (< SPACE 3). This information is available through the
1100 sb-introspect contrib.
1101 * new feature: users may subclass SEQUENCE, and have instances of
1102 these classes interoperate with standard Common Lisp functions if
1103 a number of methods are defined. (This feature is experimental
1104 and the interface subject to change based on feedback from SBCL
1105 users and the general community)
1106 * improvement: sb-sprof traces call stacks to an arbitrary depth on
1107 x86/x86-64, rather than the previous fixed depth of 8
1108 * improvement: another pthread back-end of mutex "pthread-futex".
1109 * bug fix: non-ascii command-line arguments are processed correctly
1110 (thanks to Yaroslav Kavenchuk)
1111 * bug fix: non-required arguments were not passed correctly when a method
1112 defined using DEFMETHOD was called from a mop-generated method using
1113 CALL-NEXT-METHOD (reported by Pascal Costanza)
1114 * bug fix: recursion is now permitted in accessors through
1115 SLOT-UNBOUND. (reported by Pascal Costanza)
1116 * bug fix: an error was signaled at startup if the HOME environment
1117 variable was defined, but had an empty value (reported by Peter Van Eynde)
1118 * bug fix: non ordinary lambda-list keyword in ordinary lambda lists
1119 signal a PROGRAM-ERROR, not a BUG.
1120 * bug fix: SB-POSIX:READDIR works when built with large file support.
1121 * bug fix: ENOUGH-NAMESTRING works relative to #p"/" as well. (thanks
1123 * enhancement: DESTRUCTURING-BIND lambda-list uses &BODY instead of &REST
1124 for better automatic indentation support. (thanks to Matt Pillsbury)
1125 * optimization: loading generic functions no longer takes O(n^2) time,
1126 proportional to the amount of methods in the generic function
1127 (reported by Todd Sabin and Jeremy Brown)
1128 * optimization: the FIND and POSITION family of sequence functions
1129 are significantly faster on arrays whose element types have been
1131 * improvements to the Windows port:
1132 ** Intermittent heap corruption problems have been fixed. (thanks
1133 to Alastair Bridgewater)
1134 ** TRACE :ENCAPSULATE NIL (and function end breakpoints)
1136 ** Lisp is able to unwind foreign exception frames from alien
1137 callbacks. (thanks to Alastair Bridgewater)
1139 changes in sbcl-1.0 relative to sbcl-0.9.18:
1140 * improvement: experimental support for threading on FreeBSD/x86.
1141 (thanks to NIIMI Satoshi)
1142 * improvement: runtime option --dynamic-space-size can be used
1143 to set the size of the dynamic space reserved on startup.
1144 * improvement: floating point modes in effect are now saved in
1145 core, and restored on startup.
1146 * improvement: GET-INTERNAL-REAL-TIME now reports the time since
1147 startup, not time since first call to GET-INTERNAL-REAL-TIME.
1148 * improvement: SAVE-LISP-AND-DIE explicitly checks that multiple
1149 threads are not running after *SAVE-HOOKS* have run.
1150 * improvement: writes to CLOS instance slots are type-checked in code
1151 compiled with (SAFETY 3)
1152 * improvement: floating-point exception handling on FreeBSD (thanks to
1154 * improvement: SB-POSIX supports time(2), utime(2) and utimes(2)
1155 (thanks to Zach Beane)
1156 * improvement: support for files larger than 2GB for CL streams and SB-POSIX
1158 * improvement: added support for the Shift-JIS external format.
1159 (contributed by NIIMI Satoshi)
1160 * improvement: callbacks are supported on Linux/PPC. (thanks to
1162 * bug fix: compiler bug triggered by a (non-standard) VALUES
1163 declaration in a LET* was fixed. (reported by Kaersten Poeck)
1164 * bug fix: file compiler no longer confuses validated and already
1165 dumped structurres. (reported by Kaersten Poeck)
1166 * bug fix: ADJUST-ARRAY :FILL-POINTER T on an array without a
1167 fill-pointer signals a type-error as required. (thanks to
1169 * bug fix: disassemly of funcallable instances works.
1170 * bug fix: single stepping on PPC.
1171 * bug fix: fix thread-safety problems in the type system (generally
1172 manifesting as nonsensical errors like "STRING is a bad type specifier
1173 for sequences" or "The value 1 is not of type FIXNUM")
1174 * bug fix: version components are handled correctly in TRANSLATE-PATHNAME
1175 (reported by Josip Gracin)
1176 * bug fix: an error is signaled for attempts to displace arrays with
1177 incompatible element types (thanks to Mario Mommer)
1178 * bug fix: more correct handling of wide characters in debug info
1179 (bug reported by Attila Lendvai and fixed by Juho Snellman)
1180 * optimization: method calls with &OPTIONAL or &KEY arguments are faster
1181 and don't cause extra consing
1182 * optimization: MAP and MAP-INTO are significantly faster on vectors
1183 whose elements types have been declared.
1184 * Improvements to SB-SPROF:
1185 ** Support for allocation profiling
1186 ** Reduced profiling overhead, especially for long profiling runs
1187 * Improvements to the Windows port:
1188 ** floating point exceptions are now reported correctly.
1189 ** stack exhaustion detection works partially.
1190 ** more accurate GET-INTERNAL-REAL-TIME.
1191 ** better RUN-PROGRAM behaviour in case of errors in the child
1193 ** PROBE-FILE now simplifies pathnames correctly.
1194 ** DIRECTORY now works correctly with :WILD-INFERIORS.
1195 ** DECODE-UNIVERSAL-TIME works on times before 00:00:00 January
1196 1st 1970 (although time timezone data will be unreliable).
1197 ** Hyperbolic functions TANH, ASINH, ACOSH, and ATANH work.
1199 changes in sbcl-0.9.18 (1.0.beta?) relative to sbcl-0.9.17:
1200 * enhancement: SB-POSIX now supports cfsetispeed(3), cfsetospeed(3),
1201 cfgetispeed(3), cfgetospeed(3), and related constants. (thanks to
1203 * bug fix: two potential GC deadlocks affecting threaded builds.
1204 * bug fix: (TYPEP #\A '(NOT (MEMBER #\" #\{ #\:))) now correctly
1205 returns T (reported by Anton Kazennikov)
1206 * bug fix: the STORE-VALUE restart of CHECK-TYPE works correctly
1207 with non-variable places
1208 * bug fix: remove a race condition in the setting of
1209 funcallable-instance functions, this should make threaded CLOS
1210 code more stable against memory faults.
1211 * bug fix: corruption of specials when unbinding is interrupted by an
1212 asynchronous unwind (reported by Hannu Koivisto)
1213 * improvement: the debugger will now also display local variables that
1214 are only used once, for code compiled with a DEBUG optimization quality
1217 changes in sbcl-0.9.17 (0.9.99?) relative to sbcl-0.9.16:
1218 * feature: weak hash tables, see MAKE-HASH-TABLE documentation
1219 * incompatible change: External-format support for FFI calls. The
1220 SB-ALIEN:C-STRING no longer implies an ASCII
1221 external-format. Instead the string is subject to external-format
1222 conversion. Additionally return values of type SB-ALIEN:C-STRING
1223 are by default of type (SIMPLE-ARRAY CHARACTER), not
1224 SIMPLE-BASE-STRING. For an alien type that behaves like the old
1225 SB-ALIEN:C-STRING type, use (SB-ALIEN:C-STRING :EXTERNAL-FORMAT
1226 :ASCII :ELEMENT-TYPE BASE-CHAR). (thanks to Yaroslav Kavenchuk)
1227 * incompatible change: SB-EXT package no longer contains the
1228 following unused symbols: *GC-NOTIFY-AFTER*, *GC-NOTIFY-BEFORE*,
1229 *GC-NOTIFY-STREAM*, *ERROR-PRINT-LENGTH*, *ERROR-PRINT-LEVEL*,
1231 * incompatible change: the single-stepper is no longer available
1232 on Alpha, Mips and Sparc platforms.
1233 * minor incompatible change: the direct superclasses of
1234 SB-MOP:FUNCALLABLE-STANDARD-OBJECT are (FUNCTION STANDARD-OBJECT),
1235 not (STANDARD-OBJECT FUNCTION). This makes the
1236 class-precedence-lists of GENERIC-FUNCTION and
1237 STANDARD-GENERIC-FUNCTION comply with the requirement of ANSI
1239 * minor incompatible change: the default stream external format on
1240 non-unicode SBCL is no longer detected from the locale, but is always
1242 * new feature: Add a version of evaluator that uses an interpreter instead
1243 of the compiler. EVAL still uses the compiler by default, to switch it
1244 to use the interpreter, set the value of the variable
1245 SB-EXT:*EVALUATOR-MODE* to :INTERPRET.
1246 * minor incompatible change: the single-stepper REPL has been merged
1247 with the normal debugger (see the "Stepping" heading of the debugger help
1248 for more details). The debugger command STEP will no longer switch
1249 to the single-stepper REPL.
1250 * bug fix: ENOUGH-NAMESTRING on pathnames with no name and a pattern
1251 for a type now works.
1252 * bug fix: loading of default sysinit file works. (thanks to Leonid
1254 * bug fix: better detection of circularities in the file-compiler.
1255 (reported by Marco Monteiro)
1256 * bug fix: the CL pathname functions now work with files that have
1257 non-ASCII characters in their names (thanks to Yaroslav Kavenchuk)
1258 * bug fix: The :PTY argument for RUN-PROGRAM will now work on
1259 systems with Unix98 pty semantics.
1260 * bug fix: ASDF-INSTALL will now work with bsd tar.
1261 * bug fix: ASDF-INSTALL uses GNU tar on Solaris (thanks to Josip
1263 * bug fix: timers expiring in dead threads no longer cause a
1264 type-error (reported by Paul "Nonny Mouse").
1265 * bug fix: thanks to more lightweight single-stepper instrumentation,
1266 code compiled with (DEBUG 3) will compile and execute significantly faster,
1267 and will have more accurate type-inferencing than before
1268 * bug fix: SLOT-VALUE optimizations are no longer done on method parameters
1269 whose bindings are modified
1270 * improvements to the win32 port (thanks to Yaroslav Kavenchuk):
1271 ** bug fix: arguments to RUN-PROGRAM are escaped correctly
1272 ** replace dummy implementations of CL:MACHINE-INSTANCE and
1273 CL:SOFTWARE-VERSION with proper version
1275 changes in sbcl-0.9.16 relative to sbcl-0.9.15:
1276 * feature: implemented the READER-METHOD-CLASS and
1277 WRITER-METHOD-CLASS portion of the Class Initialization Protocol
1278 as specified by AMOP.
1279 * incompatible change: variable SB-EXT:*USE-IMPLEMENTATION-TYPES*
1281 * optimization: faster LOGCOUNT implementation on x86 and x86-64
1282 (thanks to Lutz Euler)
1283 * optimization: hashing of general arrays and vectors has been
1284 improved. (reported by Any Fingerhut)
1285 * enhancement: SB-INTROSPECT is now able to find definitions of
1286 profiled functions. (thanks to Troels Henriksen)
1287 * enhancement: compiler-macro expansion applies now to FUNCALL forms
1288 as well, allowing compiler-macros for SETF-functions to expand.
1289 * enhancement: step-instrumentation no longer wraps known
1290 single-value functions with multiple-value context, allowing
1291 better type inference.
1292 * fixed bug #337: use of MAKE-METHOD in method combination now works
1293 even in the presence of user-defined method classes. (reported by
1294 Bruno Haible and Pascal Costanza)
1295 * fixed bug #339(c): if there are applicable methods not part of any
1296 long-form method-combination group, call INVALID-METHOD-ERROR.
1297 (reported by Bruno Haible)
1298 * fixed bug #361: the :FUNCTION initarg in the protocol for
1299 initialization of methods can now be used to override
1300 internally-produced optimized functions. (reported by Bruno
1302 * bug fix: extensions of MAKE-METHOD-LAMBDA which wrap the
1303 system-provided lambda expression no longer cause warnings about
1304 unbound #:|pv-table| symbols.
1305 * bug fix: improved the handling of type declarations and the
1306 detection of violations for keyword arguments with non-constant
1308 * bug fix: potentially erronous calls to PATHNAME and
1309 MERGE-PATHNAMES were being flushed in some cases.
1310 (reported by Richard Kreuter)
1311 * bug fix: compiled calls to TYPEP were mishandling obsolete
1312 instances. (reported by James Bielman and Attila Lendvai)
1313 * bug fix: format strings with ~> without matching ~< no longer
1314 trigger an AVER, but signal an understandable error instead.
1315 (reported by Antonio Martinez)
1316 * bug fix: specifying an output-file without a pathname-type for
1317 COMPILE-FILE or COMPILE-FILE-PATHNAME resulted in using the type
1318 of input-file instead of "fasl". (reported by Robert Dodier)
1319 * bug fix: compiler-macro lambda-list parsing of FUNCALL forms.
1320 (reported by James Y Knight).
1321 * bug fix: compiler-macros-function did not consider the environment
1322 argument for shadowing by local functions.
1323 * bug fix: compiler-macros expansion was inhibited by local INLINE
1325 * bug fix: inline expansions of known functions were subject to
1326 step-instrumentation in high debug policies, leading to problems
1327 with type-inference.
1328 * bug fix: compiler failed to differentiate between different CONS
1329 types in some cases.
1330 * bug fix: fixed input, output and error redirection in RUN-PROGRAM
1331 for win32. (thanks to Mike Thomas and Yaroslav Kavenchuk)
1332 * bug fix: #368: incorrect use of expressed vs. upgraded array
1334 * bug fix: #306a: more precise unions of array types.
1335 * thread-safety improvements:
1336 ** CONDITION-WAIT could return early on Linux, if the thread was
1337 interrupted and subsequently continued with SIGCONT.
1338 ** STABLE-SORT and ADJUST-ARRAY were not reentrant.
1340 changes in sbcl-0.9.15 relative to sbcl-0.9.14:
1341 * added support for the ucs-2 external format. (contributed by Ivan
1343 * minor incompatible change: pretty printing of objects of type
1344 (cons symbol) is, in the default pprint-dispatch-table, now
1345 sensitive to whether the symbol satisfies FBOUNDP. (thanks to
1347 * minor incompatible change: SB-MOP:FINALIZE-INHERITANCE is now
1348 called later in a class's lifetime, possibly as late as when the
1349 first instance of the class is created. Previously,
1350 SB-MOP:FINALIZE-INHERITANCE was called by the system as soon as a
1351 class became finalizeable.
1352 * fixed bug: FILE-POSITION sometimes returned inconsistent results
1353 for multibyte external-format streams. (thanks to "vbzoli")
1354 * fixed bug: CHANGE-CLASS would fail to preserve the values of slots
1355 with :ALLOCATION :CLASS inherited from superclasses of the
1357 * fixed bug: anonymous classes can now be created using the :NAME
1358 initarg and MAKE-INSTANCE / REINITIALIZE-INSTANCE, as specified by
1359 AMOP. (reported by Leonid Slobodov on comp.lang.lisp)
1360 * fixed bug: core-files saved with :EXECUTABLE T can again be
1361 executed when SBCL_HOME isn't set. (reported by James Knight)
1362 * fixed bug: toplevel LOCALLY forms with declarations could
1363 occasionally get miscompiled. (reported by Yaroslav Kavenchuk)
1364 * fixed bug: printing from several different threads using different
1365 values of *print-case* could cause invalid output, due to
1366 some internal special variables of the printer not being bound
1367 thread-locally (reported by Max Mikhanosha)
1368 * fixed bug: SPECIALIZER metaobjects (including anonymous classes
1369 and EQL-SPECIALIZERs) can be used as specializers to DEFMETHOD.
1370 (reported by Pascal Costanza)
1371 * fixed bug: FINALIZE-INHERITANCE is called from
1372 REINITIALIZE-INSTANCE on classes when the class has previously
1373 been finalized, as required by AMOP.
1374 * minor code generation optimizations:
1375 ** better register allocation in CLOS dispatching functions
1376 ** overflow detection when coercing signed bytes to fixnums on x86-64
1377 ** is now implemented with one IMUL instruction instead of three shifts
1378 ** more efficient bit-vector access on x86 and x86-64
1379 ** more efficient access to raw structure slots on x86 and x86-64
1380 * fixed some bugs revealed by Paul Dietz' test suite:
1381 ** ENSURE-DIRECTORIES-EXIST.8: ENSURE-DIRECTORIES-EXIST must
1382 return its argument.
1384 changes in sbcl-0.9.14 relative to sbcl-0.9.13:
1385 * feature: thread support on Solaris/x86, and experimental thread support
1387 * feature: SBCL now tries to signal a STORAGE-CONDITION when running out
1389 * feature: SBCL now provides USE-VALUE and STORE-VALUE restarts in the
1390 default method for SLOT-UNBOUND.
1391 * minor incompatible change: prevent the user from specializing the
1392 new-value argument to SB-MOP:SLOT-VALUE-USING-CLASS. It's
1393 somewhat counter to the intent of the protocol, I (CSR) think, and
1394 additionally it just doesn't work in SBCL as currently
1395 implemented, thanks to optimizations (that are always valid for
1396 the other three SLOT-VALUEish functions, but not for the setter).
1397 * minor incompatible change: the :SB-LDB feature is now enabled by
1398 default, and DISABLE-DEBUGGER and ENABLE-DEBUGGER also affect
1399 the low-level debugger.
1400 * enchancement: RUN-PROGRAM is now able to extract the file-descriptor
1401 from SYNONYM-STREAM and TWO-WAY-STREAM provided they can be decomposed
1402 down to an SB-SYS:FD-STREAM, allowing direct communication in
1403 more cases. Temporary files and pipes are still used as a fallback
1405 * thread-safety: DECODE-UNIVERSAL-TIME & GET-DECODED-TIME depended
1406 on the non-reentrant C functions localtime(3) and gmtime(3).
1407 * bug fix: global optimization policy was not visible in LOCALLY and
1409 * bug fix: class objects can be used as specializers in methods.
1410 (reported by Pascal Costanza)
1411 * bug fix: native unparsing of pathnames with :DIRECTORY NIL failed
1412 with a type error. (reported by blitz_ on #lisp)
1413 * bug fix: unparsing logical pathnames with :NAME :WILD :TYPE NIL
1414 failed with a type error. (reported by Pascal Bourguignon)
1415 * bug fix: merging pathnames against defaults with :DIRECTORY
1416 starting with '(:RELATIVE :BACK) should preserve the :BACK.
1417 (reported by James Y Knight)
1418 * bug fix: saving large (>2GB) cores on x86-64 now works
1419 * bug fix: a x86-64 backend bug when compiling (setf aref) with a
1420 constant index and a (simple-array (signed-byte 32)) array
1421 * bug fix: NAME-CHAR on an invalid symbol no longer signals an
1422 error (patch by Robert J. Macomber)
1423 * feature: TIME now displays the amount of run-time spent in GC
1424 * bug fix: The debugger now does a better job of respecting
1425 (PUSH '(*PRINT-CIRCLE* . T) SB-DEBUG:*DEBUG-PRINT-VARIABLE-ALIST*)
1426 when printing SB-DEBUG:*DEBUG-CONDITION*. (This is a debugger-only
1427 workaround for bug 403.)
1428 * bug fix: floating point exception handling now works on FreeBSD
1429 (thanks to NIIMI Satoshi)
1430 * fixed some bugs revealed by Paul Dietz' test suite:
1431 ** MISC.641: LET-conversion were not supposed to work in late
1433 * improvements to the Win32/x86 port:
1434 ** GET-INTERNAL-RUN-TIME implemented, thanks to Frank Buss.
1435 ** improved SB-BSD-SOCKETS support, thanks to Timothy Ritchey.
1437 changes in sbcl-0.9.13 relative to sbcl-0.9.12:
1438 * new feature: source path information is generated for macro-expansion
1439 errors for use in IDE's like Slime (thanks to Helmut Eller)
1440 * bug fix: calls to the compiler no longer modify *RANDOM-STATE*
1441 * bug fix: compiler does not loop forever on an invalid type in
1443 * improvement: compilation of most CLOS applications is significantly
1445 * optimization: added a limited bytecode compiler for simple toplevel
1446 forms, speeding up compilation and FASL loading
1447 * bug fix: the statistical profiler now properly distinguishes anonymous
1450 changes in sbcl-0.9.12 relative to sbcl-0.9.11:
1451 * minor incompatible change: in sbcl-0.9.11 (but not earlier
1452 versions) SB-MOP:FINALIZE-INHERITANCE would recursively descend
1453 into subclasses of the finalized class. Now user calls to
1454 FINALIZE-INHERITANCE finalize just the one class, and calls by the
1455 system return before any subclasses are finalized.
1456 * minor incompatible change: The reader no longer ignores errors
1457 regarding non-existent packages in #+ and #- feature tests.
1458 * new feature: command line options --no-sysinit, --no-userinit to
1459 inhibit loading the corresponding init files
1460 * new feature: add a generic function SOCKET-SEND to SB-BSD-SOCKETS,
1461 for sending data through UDP sockets (thanks to François-René Rideau)
1462 * minor incompatible change: SIGPIPE is ignored and "Broken pipe"
1463 error is signalled instead (thanks to François-René Rideau)
1464 * minor incompatible change: Error signalling behaviour of lexical
1465 operations violating package locks has changed slightly. Refer to
1466 documentation on package locks for details.
1467 * enhancement: EVAL can process IF-expressions without resorting to the
1469 * bug fix: Some operations involving SIN, COS, or TAN weren't being
1470 constant-folded properly due to missing out-of-line functions.
1471 (reported by Mika Pihlajamäki)
1472 * bug fix: LISTEN sometimes returned T even in cases where no data was
1473 immediately available from the stream
1474 * fixed bug: types of the last two arguments to SET-SYNTAX-FROM-CHAR
1475 were reversed. (reported by Levente Mészáros)
1476 * fixed bug: Tests for the (VECTOR T) type gave the wrong answer
1477 when given a vector displaced to an adjustable array. (reported
1479 * bug fix: derivation of float boundaries from numbers outside the
1480 appropriate float range (reported by John Wiseman)
1481 * bug fix: MAKE-LOAD-FORM-SAVING-SLOTS accepts en empty slot name
1483 * bug fix: precomputing cache entries for generic functions with
1484 some subclasses of specializers as yet invalid does not attempt to
1485 fill a cache line with a negative offset. (reported by Levente
1487 * improvements to DOCUMENTATION for TYPE and STRUCTURE doc-types:
1488 allow condition class objects as arguments to DOCUMENTATION and
1489 (SETF DOCUMENTATION); only find and set documentation for
1490 structure names for the STRUCTURE doc-type. (suggested by Gary
1492 * improvements to the Win32/x86 port:
1493 ** (user-homedir-pathname) and default initialization file
1494 locations now know about the user's "Documents and Settings"
1495 directory (thanks to Yaroslav Kavenchuk)
1496 ** run-program is implemented (thanks to Mike Thomas)
1497 ** sockets support (thanks to Timothy Ritchey)
1498 ** better backtrace support (thanks to Alastair Bridgewater)
1499 ** sb-grovel supported
1500 ** asdf-install and sb-posix work somewhat
1501 ** capable of running Slime using SWANK:*COMMUNICATION-STYLE* NIL
1502 * improvements to the Solaris/x86 port:
1503 ** works on Solaris 11/Solaris Express
1504 ** floating-point exception handling support
1505 ** support for the breakpoint-based TRACE :ENCAPSULATE NIL facility
1506 * fixed some bugs revealed by Paul Dietz' test suite:
1507 ** REMOVE-METHOD returns its generic function argument even when
1508 the method is not one of the generic functions' methods.
1509 ** objects declared as MEMBER types can be admissible arguments to
1510 structure accessors.
1511 ** printing characters should simply be printed by the FORMAT ~:C
1513 ** compiler failure when compiling functions with hairy constant
1514 defaults for optional parameters.
1515 ** compiler produces wrong code when MAYBE-INLINE-expanding a
1516 function, which is already optimized.
1518 changes in sbcl-0.9.11 relative to sbcl-0.9.10:
1519 * new platform: experimental support for SBCL x86/Darwin, including
1520 MacOS X 10.4.5 on Intel.
1521 * new feature: Unicode character names are now known to the system
1522 (through CHAR-NAME and NAME-CHAR).
1523 * new feature: the filesystem location of SBCL's core file is
1524 exposed to lisp through the variable SB-EXT:*CORE-PATHNAME*.
1525 * minor incompatible change: the contrib modules SB-POSIX and
1526 SB-BSD-SOCKETS no longer depend on stub C libraries; the intent of
1527 this change is to make it easier to distribute
1528 location-independent binaries.
1529 * bug fix: as implied by AMOP, standardized classes no longer have
1530 slots named by external symbols of public packages. (reported by
1532 * optimization: calling functions via constant symbols -- as in
1533 (FUNCALL 'FOO) -- is now roughly as efficient as calling them
1534 via the function object as in (FUNCALL #'FOO).
1535 * optimization: on x86 and x86-64, the calling convention for the
1536 case of dealing with an unknown number of values has been altered
1537 to be friendlier to the prediction heuristics implemented,
1538 particularly with reference to CALL and RET pairing. (thanks to
1539 Alastair Bridgewater)
1540 * optimization: on x86, the code for access to thread-local symbol
1541 values and binding/unbinding of thread-local symbols is smaller.
1542 (thanks to Alastair Bridgewater)
1543 * enhancement: CONSTANTP is now able to determine constantness of
1544 more complex forms, including calls to constant-foldable standardized
1545 functions and some special forms beyond QUOTE.
1546 * fixed bug: occasional GC crashes on Solaris/x86
1547 * optimization: x86-64 supports stack allocation of results of simple
1548 calls of MAKE-ARRAY, bound to variables, declared DYNAMIC-EXTENT
1549 * enchancement: the PROCESS-INPUT and -OUTPUT streams created by
1550 SB-EXT:RUN-PROGRAM can be used for both character and byte IO
1551 (thanks to James Knight)
1552 * fixed bug: CL:LISTEN always returns NIL at end of file, as required
1553 by the standard (thanks to Stephen Compall)
1555 changes in sbcl-0.9.10 relative to sbcl-0.9.9:
1556 * new feature: new SAVE-LISP-AND-DIE keyword argument :EXECUTABLE can
1557 be used for bundling the runtime and the core file into one
1558 executable binary. This feature is not currently supported on all SBCL
1559 platforms. (thanks to James Bielman and NIIMI Satoshi)
1560 * new feature: a generational or ephemeral garbage collector is now
1561 the default on the PowerPC platform (both Linux and Darwin). The
1562 old Cheney (stop and copy) collector is a build-time option.
1563 (thanks to Cyrus Harmon, after Raymond Toy)
1564 * minor incompatible change: the method by which SBCL finds its
1565 contributed modules has changed; it no longer relies on symbolic
1566 links from an $SBCL_HOME/systems directory, but searches directly
1567 in the subdirectories of $SBCL_HOME.
1568 * enhancement: the dynamic heap size on the Linux/PPC platform is
1569 markedly larger, even using the older Cheney garbage collector.
1570 * fixed bug #399: full call to DATA-VECTOR-REF in accesses to
1571 certain complicated string types. (reported by Gary King)
1572 * fixed bug: STRING-TO-OCTETS and OCTETS-TO-STRING did not convert
1573 character codes. (thanks to Yaroslav Kavenchuk and Ivan Boldyrev)
1574 * fixed bug: DOCUMENTATION on structures no longer signals an
1575 error if no structure type of the right name exists. (reported by
1577 * fixed bug: compiled calls to MAKE-INSTANCE where there is an
1578 applicable non-standard primary or :AROUND method on
1579 INITIALIZE-INSTANCE or SHARED-INITIALIZE and a non-keyword initarg
1580 no longer cause unbound variable errors. (reported by Kevin Reid)
1581 * fixed bug: metacircle resolution in cases where methods have slots
1582 added before the slots from STANDARD-METHOD. (reported by Jean
1584 * fixed bug: the Power PC assembly code for calling into lisp
1585 sometimes computed its return address wrongly (depending on the
1586 15th bit of the address, fixed at link time). This used to
1587 manifest itself in a segmentation violation while building PCL.
1588 (reported by Kevin Rosenberg, Eric Marsden, Lars Brinkhoff and
1589 many others over the years)
1590 * fixed bug: ROOM no longer reports silly numbers for stack usage if
1591 the stack pointer (treated as a signed integer) is negative.
1592 (thanks to Peter van Eynde)
1594 changes in sbcl-0.9.9 relative to sbcl-0.9.8:
1595 * new platform: experimental support for the Windows operating
1596 system has been added. (thanks to Alastair Bridgewater)
1597 * fixed several bugs in and robustified the PPC FFI (including
1598 callbacks). (thanks to Cyrus Harmon and Heiner Schwarte)
1599 * bug fix: allow non-simple string symbol names (reported by Paul
1601 * bug fix: interrupt handling on NetBSD (thanks to Richard M
1603 * bug fix: saving a core corrupted callbacks on x86/x86-64
1604 * bug fix: closed a loophole in metacircularity detection and
1605 grounding in the PCL implementation of CLOS.
1606 * bug fix: TRUENAME on "/" no longer returns a relative pathname.
1607 (reported by tomppa on #lisp)
1608 * bug fix: clear the direction flag on Lisp -> C transitions, as
1609 required by the x86-64 ABI. Fixes mysterious GC crashes on SuSE.
1610 (reported by Andrej Grozin and Hendrik Maryns)
1611 * optimization: major improvements to GC efficiency on GENCGC platforms
1612 * optimization: faster implementation of EQUAL
1613 * optimization: emit more efficient opcodes for some common
1614 immediate->register MOV instructions on x86-64. (thanks to Lutz Euler)
1615 * optimization: several other minor code-generation improvements on x86-64
1616 * fixed segfaults on x86 FreeBSD 7-current. (thanks to NIIMI Satoshi)
1618 changes in sbcl-0.9.8 relative to sbcl-0.9.7:
1619 * minor incompatible change: (SETF CLASS-NAME) and (SETF
1620 GENERIC-FUNCTION-NAME) are generic functions once more (reverting
1621 the change in the 0.9.7 release). (SETF CLASS-NAME) is specified
1622 by ANSI as a generic function, and for consistency (SETF
1623 GENERIC-FUNCTION-NAME) is treated likewise.
1624 * fixed bug #233.b: make constraint propagation notice when a variable
1625 value is changed after it is referenced but before it is used
1626 * fixed bug #296: no more arbitrary behaviour from filenames with
1627 odd characters as --load arguments. (reported by Adam Warner)
1628 * fixed bug #390: :CHARACTER-SET pathname components now work as
1629 expected. (reported by Tim Daly Jr)
1630 * fixed bug #391: complicated :TYPE intersections in slot
1631 definitions no longer cause an error in PCL internals.
1632 * fixed bug #392: plugged a hole in the obsolete-instance protocol.
1633 * bug fix: FILE-STRING-LENGTH is now external-format sensitive,
1634 returning the number of octets which would be written to the
1635 file-stream. (thanks to Robert J. Macomber)
1636 * bug fix: the SB-SIMPLE-STREAMS contrib passes external-format
1637 arguments to the internal stream functions. (thanks to David
1639 * bug fix: stack exhaustion is now handled on the x86/SunOS
1640 platform. (thanks to Daisuke Homma)
1641 * bug fix: more accurate ROOM results on GENCGC platforms
1642 * optimization: improved type inference for arithmetic-for
1643 index variables in LOOP
1644 * optimization: faster floating-point SQRT on x86-64
1645 * fixed some bugs revealed by Paul Dietz' test suite:
1646 ** DOCUMENTATION returns NIL instead of "" for method combinations
1647 that don't have a docstring
1649 changes in sbcl-0.9.7 relative to sbcl-0.9.6:
1650 * minor incompatible change: (SETF CLASS-NAME) and (SETF
1651 GENERIC-FUNCTION-NAME) are no longer generic functions, and
1652 therefore are not customizeable by user code (as seems to be at
1653 least permitted and maybe required by AMOP). As a consolation,
1654 however, the SBCL implementation of these functions now calls
1655 REINITIALIZE-INSTANCE as specified by AMOP.
1656 * bug fix: slot-definition documentation is propagated as per ANSI
1657 7.6.2 to EFFECTIVE-SLOT-DEFINITIONS.
1658 COMPUTE-EFFECTIVE-SLOT-DEFINITION now receives a :DOCUMENTATION
1659 argument, as do eslotd initialization methods. (from Pascal
1660 Costanza's "Closer" project)
1661 * bug fix: REINITIALIZE-INSTANCE on generic functions calls
1662 COMPUTE-DISCRIMINATING-FUNCTION (almost) unconditionally, as
1664 * bug fix: it is now possible to have more than one subclass of
1665 STANDARD-GENERIC-FUNCTION without causing stack overflow.
1666 (reported by Bruno Haible, Pascal Costanza and others)
1667 * bug fix: the dependent update protocol now works for generic
1668 functions. (thanks to Gerd Moellmann; reported by Bruno Haible
1669 and Pascal Costanza)
1670 * bug fix: condition-class instances corresponding to
1671 DEFINE-CONDITION forms are now created eagerly. (reported by
1672 Kalle Olavi Niemitalo on comp.lang.lisp)
1673 * bug fix: floating point printing is more accurate in some
1674 circumstances. (thanks to Simon Alexander)
1675 * bug fix: *COMPILE-FILE-PATHNAME* now contains the user's pathname
1676 merged with *DEFAULT-PATHNAME-DEFAULTS*.
1677 * bug fix: callbacks on OS X now preserve stack-alignment.
1678 (thanks to Cyrus Harmon)
1679 * enhancement: the x86-64 disassembler is much better at
1680 disassembling SSE instructions. (thanks to Lutz Euler)
1681 * enhancement: CHECK-TYPE on a variable now allows type-inference
1682 (thanks to Vincent Arkesteijn)
1683 * optimization: improved performance of EUC-JP external format.
1684 (thanks to NIIMI Satoshi)
1685 * optimization: performance improvements to IO on file streams of
1686 :ELEMENT-TYPE CHARACTER
1687 * optimization: much faster memory allocation on x86-64
1688 * optimization: faster garbage collections (latency and throughput)
1690 * optimization: faster bignum subtraction and division on x86 and x86-64
1691 (thanks to Lutz Euler)
1692 * optimization: more accurate type-inference for loops that use a
1693 floating point index variable or a negative step.
1695 changes in sbcl-0.9.6 relative to sbcl-0.9.5:
1696 * bug fix: add a workaround to SBCL looping infinitely at startup on
1697 Linux kernels with apparently buggy implementations of personality().
1698 (thanks to Svein Ove Aas)
1699 * bug fix: Unicode symbols are correctly printed in LDB backtraces
1700 (thanks to David Lichteblau)
1701 * bug fix: local bindings shadow global symbol macros for
1702 macroexpansion. (reported by Matthew D Swank on comp.lang.lisp)
1703 * optimization: non-open coded uses of numeric comparison operators
1704 (e.g. >) no longer cons when called with more than one parameter
1705 on platforms supporting dynamic-extent allocation.
1706 * enhancement: saving cores with foreign code loaded is now supported
1707 on MIPS/Linux in addition to the previously supported platforms.
1708 * bug fix: FIND-RESTART now tests for activity, not applicability when given
1709 a restart object as identifier. (reported by Helmut Eller for CMUCL)
1710 * bug fix: division by zero in sb-sprof when no samples were collected
1711 * bug fix: a race when a slow to arrive sigprof signal killed sbcl
1712 * bug fix: asdf-install uses CRLF as required by the HTTP spec.
1713 (thanks to Alexander Kjeldaas)
1714 * new feature: ignoring the return values of destructive functions
1715 like NREVERSE signals a compile-time style-warning.
1716 (thanks to Kevin Reid)
1717 * new feature: the alignment of alien structure fields can be
1718 explicitly specified. (thanks to Cyrus Harmon)
1719 * bug fix: run cleanup forms (in all threads) when receiving a SIGTERM
1720 and dump core on SIGQUIT
1722 ** incompatible change: threads do not inherit values of specials
1723 from their parents (see manual)
1724 ** bug fix: threads stacks belonging to dead threads are freed by the
1725 next exiting thread, no need to gc to collect thread stacks anymore
1726 ** minor incompatible change: INTERRUPT-THREAD-ERROR-ERRNO removed
1727 ** WITH-RECURSIVE-LOCK can be nested in a WITH-MUTEX for the same lock
1728 ** bug fix: dynamic variable and thread start related gc lossage
1729 ** bug fix: job control (SIGSTOP/SIGCONT) no longer confuses threads
1731 * fixed some bugs revealed by Paul Dietz' test suite:
1732 ** SUBTYPEP is slightly more accurate on heinously complicated
1733 CONS types where some of the members have uncertain (in the
1734 NIL, NIL sense) type relationships to each other.
1736 ** Cores produced by SAVE-LISP-AND-DIE on GENCGC platforms are
1737 no longer purified unless :PURIFY T is explicitly specified.
1738 ** Non-purified cores are significantly smaller than before
1740 changes in sbcl-0.9.5 relative to sbcl-0.9.4:
1741 * new feature: timers based on Zach Beane's excellent timer package
1742 * added support for the following external formats: koi8-u,
1743 x-mac-cyrillic, cp437, cp850, cp852, cp855, cp857, cp860, cp861,
1744 cp862, cp863, cp864, cp865, cp866, cp869, cp874, iso-8859-2,
1745 iso-8859-3, iso-8859-4, iso-8859-5, iso-8859-6, iso-8859-7,
1746 iso-8859-8, iso-8859-9, iso-8859-10, iso-8859-11, iso-8859-13,
1747 iso-8859-14, cp1250, cp1251, cp1252, cp1253, cp1254,
1748 cp1255,cp1256, cp1257, cp1258 (contributed by Ivan Boldyrev)
1749 * incompatible change: a threaded SBCL will no longer revert to
1750 non-threaded mode on non-NPTL systems, but refuse to start entirely.
1751 * bug fix: interrupts are disabled until startup is complete; no
1752 more sigsegvs when receiving a signal to soon
1753 * optimization: faster 32-bit SB-ROTATE-BYTE:ROTATE-BYTE on non-x86/ppc
1755 * bug fix: add a workaround for the memory randomization features in
1756 Linux kernels >= 2.6.12 that interfere with SBCL's memory maps. This
1757 workaround will only be in effect on systems with the proc filesystem
1759 * bug fix: printing objects of type HASH-TABLE signals a
1760 PRINT-NOT-READABLE error when *READ-EVAL* is NIL. (reported by
1762 * bug fix: GET-INTERNAL-REAL-TIME now works even for processes that
1763 have been running for over 50 days. (reported by Gilbert Baumann)
1764 * bug fix: the logic for getting names of functions gets less
1765 confused when confronded with alternate-metaclass
1766 funcallable-instances. (reported by Cyrus Harmon)
1767 * bug fix: FUNCTIONP and (LAMBDA (X) (TYPEP X 'FUNCTION)) are now
1768 consistent, even on internal alternate-metaclass objects.
1769 * bug fix: flush closure information collected by physical
1770 environment analysis prepass before the main pass. (bug reported
1772 * bug fix: compiler pack phase does not modify a hashtable, which is
1773 iterated. (reported by Bryan O'Connor, fixed by Rob MacLachlan)
1774 * bug fix: backquote does not optimize (LIST* x y) to CONS when x
1775 may be expanded. (reported by Alexander <malishev> on c.l.l)
1776 * bug fix: no more style-warnings when compiling code using LAST
1777 under high-SPEED low-SPACE settings. (reported by David Wallin
1779 * fixed some incompatibilities between SBCL's MOP and the MOP
1781 ** the METAOBJECT class is now implemented;
1782 ** FUNCALLABLE-STANDARD-OBJECT is now a subclass of
1783 STANDARD-OBJECT, as required;
1784 ** the classes STANDARD-CLASS and FUNCALLABLE-STANDARD-CLASS are
1785 now compatible (as required by VALIDATE-SUPERCLASS); there
1786 remains a consistency requirement over the presence or absence
1787 of the FUNCTION class in the superclass list of finalized
1788 classes; see the manual for more details;
1789 ** the :AROUND method for COMPUTE-SLOTS on
1790 FUNCALLABLE-STANDARD-CLASS now honours the primary method's
1791 requested slot ordering.
1793 ** bug fix: parent thread now can be gc'ed even with a live
1795 ** bug fix: binding a special with PROGV to no value is not
1796 the same as that symbol not having been bound (thanks to
1798 ** bug fix: binding specials is thread safe (thanks to
1800 ** bug fix: interrupt handlers are now per-process, RUN-PROGRAM
1801 and SB-SPROF do not die with 'no handler for signal XX in
1802 interrupt_handle_now(..)' anymore
1803 ** bug fix: WITH-TIMEOUT works with multiple running threads
1804 * fixed some bugs revealed by Paul Dietz' test suite:
1805 ** ENSURE-GENERIC-FUNCTION should take a method class object for
1806 the :method-class keyword argument.
1808 changes in sbcl-0.9.4 relative to sbcl-0.9.3:
1809 * new port: the Solaris operating system on x86 processors is now
1810 mostly supported, though some rough edges in the environment
1811 remain. (thanks to Daisuke Homma)
1812 * enhancement: SBCL on MIPS platforms now has a much larger dynamic
1813 space for its heap. (thanks to Thiemo Seufer)
1814 * optimization: SBCL on MIPS platforms now supports dynamic-extent
1815 closures. (thanks to Thiemo Seufer)
1816 * minor incompatible change: eof selects abort in the debugger.
1817 * minor incompatible change: *INVOKE-DEBUGGER-HOOK* is run before
1818 *DEBUGGER-HOOK* => *DEBUGGER-HOOK* is not run when the debugger
1820 * minor incompatible change: SB-KERNEL:INSTANCE-LAMBDA is
1821 deprecated, and will go away in a future revision of SBCL.
1822 * minor incompatible change: GC-ON and GC-OFF are no longer
1823 implemented with a counter, it does not matter how many times gc
1824 is switched on or off
1825 * bug fix: discriminating functions for generic function classes
1826 with non-standard methods for COMPUTE-APPLICABLE-METHODS no longer
1827 make invalid assumptions about method precedence order. (reported
1829 * bug fix: TRUNCATE now correctly signals division by zero on MIPS
1830 platforms. (thanks to Thiemo Seufer)
1831 * bug fix: degree sign (<U00B0>) could not be encoded in KOI8-R.
1832 * bug fix: correct pathname printing with printer escaping is on.
1833 (thanks to Kevin Reid)
1834 * bug fix: complex VOP definitions in "user-space" no longer trigger
1835 package locks. (reported by Zach Beane)
1836 * fixed bug 343: SB-KERNEL:INSTANCE-LAMBDA is no longer necessary
1837 for funcallable-instance functions, and is no different from
1839 * bug fix: PARSE-INTEGER no longer depends on the whitespaceness of
1840 characters in the current readtable. (reported by Nicholas Neuss)
1841 * bug fix: SUBTYPEP on various CONS types returns more a more
1842 accurate acknowledgment of its certainty.
1843 * optimizations: REMOVE-DUPLICATES now runs in linear time on
1844 lists in some cases. This partially fixes bug 384.
1845 * flush all standard streams before prompting in the REPL and the
1847 * bug fix: signal handling and triggering gc do not conflict
1848 directly anymore, in particular a high frequency sb-sprof does
1849 not prevent gc from running
1850 * bug fix: DECODE-UNIVERSAL-TIME now uses a more reasonable
1851 approximation for timezone and DST information between the
1852 universal time epoch and the smallest negative 32-bit time_t.
1853 * bug fix: ENCODE-UNIVERSAL-TIME no longer errors when passed the
1854 year 1899 so long as the encoded time is non-negative.
1855 * bug fix: on the x86 platform, SB-ROTATE-BYTE:ROTATE-BYTE returns
1856 the correct answer for non-constant rotate amounts when performing
1857 an inline 32-bit rotation.
1859 ** bug fix: RELEASE-FOREGROUND doesn't choke on session lock if
1860 there is only one thread in the session
1861 ** bug fix: memory leak for streams created in one thread and
1862 written to in another
1863 ** bug fix: lockup when compiled with gcc4
1864 ** bug fix: race that allows the gc to be triggered when gc is
1866 ** bug fix: one less memory fault in INTERRUPT-THREAD, again
1867 ** bug fix: gc and INTERRUPT-THREAD don't hang when the RT signal
1869 ** bug fix: finished threads are gc'ed properly
1870 * fixed some bugs revealed by Paul Dietz' test suite:
1871 ** CALL-NEXT-METHOD signals an error (in safe code) when the call
1872 has arguments with a different set of applicable methods from
1873 the orignal arguments.
1874 ** The type error thrown by MAP now has a correct expected-type
1876 ** DISASSEMBLE now throws a TYPE-ERROR when its argument does not
1877 name a compiled function.
1878 ** Three MISC tests where a large bignum was improperly coerced to
1879 a float that couldn't represent that bignum during type
1880 derivation were fixed.
1881 ** SUBTYPEP can now handle types involving the negation of a
1882 list-form FUNCTION type.
1883 ** SUBTYPEP also now handles certain COMPLEX type specifiers such
1884 as (COMPLEX (AND RATIO (NOT FIXNUM))).
1885 ** READ-BYTE and WRITE-BYTE no longer take stream designators.
1887 changes in sbcl-0.9.3 relative to sbcl-0.9.2:
1888 * New feature: Experimental support for bivalent streams: streams
1889 opened with :element-type :default now allow character and binary
1890 (unsigned-byte 8) I/O
1891 * Support for the koi8-r external format. (thanks to Ivan Boldyrev)
1892 * Bug fix: OPEN no longer fails when *PRINT-READABLY* is T. (thanks
1894 * bug fix: no more highly sporadic "couldn't check whether ~S is
1895 readable" when reading a stream and an interrupt hits in the middle
1896 of a select system call
1897 * compiler better recognizes complex arrays (reported by Eduardo
1899 * bug fix: out-of-line SB-SYS:FOREIGN-SYMBOL-ADDRESS did not work
1900 for variables on SBCL built with linkage-tables. (reported by Luis
1902 * various error reporting improvements.
1903 * optimizations: LOGNOR on fixnums is improved in the MIPS backend.
1904 (Thanks to Thiemo Seufer)
1905 * bug fix: nested reader invokations work correctly
1906 * bug fix: it is possible to have simultaneous references to foreign
1907 code and foreign data with the same name.
1909 ** added x86-64 support
1910 ** incompatible change: the threading api now works with thread
1911 objects instead of thread ids
1912 ** bug fix: threads are protected from signals and interruption when
1913 starting up or going down
1914 ** bug fix: a race where an exiting thread could lose its stack to gc
1915 ** bug fix: don't halt on infinite error in threads if possible
1916 ** fixed numerous gc deadlocks introduced in the pthread merge
1917 ** bug fix: fixed thread safety issues in read and print
1918 ** bug fix: debugger doesn't hang on session lock if interrupted at
1919 an inappropriate moment
1920 ** bug fix: run-program is now thread safe(r)
1921 ** bug fix: inner with-recursive-lock no longer releases the mutex
1922 * fixed a bug in (DECODE-UNIVERSAL-TIME 0) (reported by Paul Dietz)
1923 * fixed some bugs revealed by Paul Dietz' test suite:
1924 ** TYPE-ERRORs from signalled by COERCE now have DATUM and
1925 EXPECTED-TYPE slots filled.
1926 ** MULTIPLE-VALUE-SETQ always returns the primary value of the
1928 ** MAKE-CONDITION accepts classes as type-designators.
1929 ** COMPILE may never return NIL.
1930 ** ENCODE-UNIVERSAL-TIME now guards against times out of Unix's
1931 range before calling Unix time functions
1933 changes in sbcl-0.9.2 relative to sbcl-0.9.1:
1934 * numerous signal handling fixes to increase stability
1935 * Support for EUC-JP external format. (thanks to NIIMI Satoshi)
1936 * minor incompatible change: we now correctly canonize default
1937 initargs, making them be a list of (INITARG INITFORM INITFUNCTION)
1938 as per the MOP, rather than the historical (INITARG INITFUNCTION
1939 INITFORM). (reported by Bruno Haible)
1940 * new feature: WITH-COMPILATION-UNIT now accepts a non-standard
1941 :SOURCE-PLIST option. See (DOCUMENTATION #'WITH-COMPILATION-UNIT T)
1942 for more information.
1943 * TRUENAME and PROBE-FILE now correctly resolve symlinks even if the
1944 pathname is a directory pathname.
1945 * SB-SPROF now works (more) reliably on non-GENCGC platforms.
1946 * dynamic space size on PPC has been increased to 768Mb. (thanks to
1948 * SB-MOP:ENSURE-CLASS-USING-CLASS now accepts a class as the
1949 :METACLASS argument in addition to a class name. (reported by
1950 Bruno Haible for CMUCL, patch for CMUCL by Gerd Moellmann)
1951 * RESTART-CASE can now be wrapped around CALL-METHOD forms.
1952 (reported by Bruno Haible; patch from Gerd Moellmann)
1953 * bug fix: sbcl runtime can now be compiled with gcc4 (thanks to
1955 * bug fix: more cleanups to the floating point exception handling on
1956 x86-64 (thanks to James Knight)
1957 * bug fix: the compiler does not try to constant fold calls of
1958 COERCE and BIT-* functions when they return freshly constructed
1959 objects. (reported by Nikodemus Siivola and Paul Dietz)
1960 * optimization: improved the MIPS versions of generic (in the
1961 generic sense) arithmetic routines. (thanks to Thiemo Seufer)
1962 * optimization: direct conversion of (unsigned-byte 32) to floats on
1963 the PowerPC platform.
1964 * optimization: structure instances with raw slots now use less
1965 memory, and probably show better memory locality. (thanks to
1967 * optimization: DYNAMIC-EXTENT declarations for lists and closures
1968 are treated as requests for stack allocation on the x86-64,
1969 Alpha-32, PPC and SPARC platforms.
1970 * contrib improvement: it's harder to cause SOCKET-CLOSE to close()
1971 the wrong file descriptor; implementation of SOCKET-OPEN-P.
1972 (thanks to Tony Martinez)
1974 ** gcing a dead thread can no longer lead to lockups
1975 ** threads block signals until they are set up properly
1976 ** errno is no longer shared by threads
1977 ** interrupt-thread restores the eflags register on x86
1978 ** fixed some lockups due to gc/thread interaction
1979 * fixed some bugs revealed by Paul Dietz' test suite:
1980 ** invalid dotted lists no longer raise a read error when
1981 *READ-SUPPRESS* is T
1982 ** signal an error if a symbol that names a declaration is used
1983 as the name of a type, or vice versa
1984 ** allow using the (declare (typespec var*)) abbreviation for
1985 (declare (type typespec var*)) with all type specifiers
1986 ** circularity detection works properly with PPRINT-LOGICAL-BLOCK
1987 ** always return NIL from PPRINT-POP when OBJECT is NIL
1988 ** don't signal errors when pretty-printing malformed LABELS,
1989 FLET or MACROLET forms
1990 ** declarations in a DOLIST body are also in scope for the
1992 ** COMPILE-FILE accepts all pathname designators as INPUT-FILE
1994 ** the ELEMENT-TYPE keyword argument to WITH-OUTPUT-STREAM is
1997 changes in sbcl-0.9.1 relative to sbcl-0.9.0:
1998 * fixed cross-compiler leakages that prevented building a 32-bit
1999 target with a 64-bit host compiler.
2000 * fixed a bug in CLOSE :ABORT T: no longer attempts to remove files
2001 opened with :IF-EXISTS :OVERWRITE.
2002 * fixed bug 281: error for an invalid qualifier in a short-form method
2003 combination method is not signalled until the faulty method is called.
2004 * bug fix: iteration variable type inferrer failed to deal with open
2005 intervals. (reported by Alan Shields)
2006 * bug fix: dynamically loading (via LOAD-SHARED-OBJECT or similar)
2007 "frameworks" on Mac OS X no longer causes an EXC_BAD_ACCESS if two
2008 or more runtime options were provided to the sbcl binary.
2009 * compiled code is not steppable if COMPILATION-SPEED >= DEBUG.
2010 * contrib improvement: implement SB-POSIX:MKSTEMP (Yannick Gingras)
2011 * optimization: there's now a fast-path for fixnum arguments in the
2012 generic subtraction routines on x86/x86-64. (Thanks to Lutz Euler)
2013 * optimization: the code generated on x86-64 is more compact thanks
2014 to not outputting unneccessary prefix bytes. (Thanks to Lutz Euler)
2015 * bug fix: floating-point exception handling now partly works on
2016 x86-64. (Thanks to James Knight)
2017 * improvement to the MIPS backend from Thiemo Seufer: C-style
2018 64-bit long long arguments and return values to alien functions
2020 * fixed some bugs revealed by Paul Dietz' test suite:
2021 ** the type-error signalled from WARN has a filled-in DATUM slot.
2022 ** the type-error required when a stream is not associated with
2023 a file has the stream as its datum.
2024 ** type-errors on single-floats on x86-64 no longer have
2025 :INVALID-OBJECT as the datum
2026 ** the type-errors signalled for invalid function names now have
2027 a correct expected type
2028 ** (SETF (DOCUMENTATION ... 'STRUCTURE)) no longer signals an error
2029 for structures defined with a :TYPE. Documentation strings for
2030 typed structures are no longer immediately discarded
2031 ** FILE-STRING-LENGTH and STREAM-EXTERNAL-FORMAT now work on
2032 broadcast streams and synonym streams. FILE-LENGTH now also works
2033 on broadcast streams.
2035 changes in sbcl-0.9.0 relative to sbcl-0.8.21:
2036 * incompatible change: the --noprogrammer option, deprecated since
2037 version 0.7.5, has been removed. Please use the equivalent
2038 --disable-debugger option instead.
2039 * incompatible change: finalizers and *AFTER-GC-HOOKS* are now run with
2041 * incompatible change: support for *BEFORE-GC-HOOKS* (that have been
2042 inoperational for a while now) has been completely removed.
2043 * null lexical environments are now printed as #<NULL-LEXENV>,
2044 significantly reducing the amount of clutter in typical backtraces.
2045 * documentation on weak pointers, finalization, and after GC hooks
2046 has been added to the manual.
2047 * optimization: REPLACE on declared (UNSIGNED-BYTE 8) vectors, as well
2048 as other specialized array types, is much faster. SUBSEQ and
2049 COPY-SEQ on such arrays have also been sped up.
2050 * optimization: EQL is now more efficient when at least other argument
2051 is known to be of type (OR FIXNUM (NOT NUMBER)).
2052 * fixed bug: the runtime is now less vulnerable to changes in the
2053 size of the SBCL object on OS X, and virtual memory is reserved for
2054 all spaces that need to be at a fixed address.
2055 * fixed bug: finalizers are now thread-safe. (thanks to Gabor Mellis)
2056 * fixed bug: finalizers and after GC hooks that cause consing are now
2058 * fixed bug: compiler error messages and summaries are now printed to
2059 *ERROR-OUTPUT*, not *STANDARD-OUTPUT*.
2060 * fixed inference of the upper bound of an iteration variable.
2061 (reported by Rajat Datta).
2062 * fixed bug 211e: calling local functions with duplicated constant
2063 keyword argument no longer causes a bogus style warning about an
2065 * fixed bug 305: INLINE/NOTINLINE declaration no longer causes local
2066 ftype declaration to be disregarded. (reported by Dave Roberts)
2067 * fixed bug 373: caused by erronous compilation of references to alien
2068 variables in the runtime on ppc/darwin.
2069 * fixed bug 376: CONJUGATE type deriver.
2070 * fixed infinite looping of ALIEN-FUNCALL, compiled with high DEBUG.
2071 (reported by Baughn on #lisp)
2072 * a cross-compiler bug on non-x86 platforms has been identified and
2073 fixed. (thanks to Bruno Haible)
2074 * improvements to the MIPS runtime code for increased stability.
2075 (thanks to Thiemo Seufer)
2076 * increased the maximimum compact environment size to allow
2077 purify on images with large amounts of functions. (thanks to Cyrus Harmon)
2078 * partial workaround for bug 135: don't name the function we're
2079 calling for hairy cases of EVAL, so as not to accumulate one environment
2080 entry per call to EVAL. (reported by Kevin Reid)
2081 * improvements to the x86-64 disassembler. (thanks to Lutz Euler)
2082 * optimization: type testing for non-vector arrays should be faster.
2083 * fixed TRACE :ENCAPSULATE NIL, added support for :ENCAPSULATE NIL
2085 * bug fix: setting 31st element of a bit vector to zero did not work
2087 * bug fix: redefining a class definition which failed due to a
2088 previous accessor / function clash now works (but see BUGS entry
2089 #380 for more problems in this area). (thanks to Zach Beane)
2090 * the long form of DEFINE-METHOD-COMBINATION disables method group
2091 checking when given a single method group with pattern *. (thanks
2093 * on x86 compiler supports stack allocation of results of simple
2094 calls of MAKE-ARRAY, bound to variables, declared DYNAMIC-EXTENT.
2095 * support for building and running on Mac OS X 10.4 "Tiger" has been added
2096 ** Binaries built on Tiger will not run on 10.2 "Jaguar" currently
2097 * fixed some bugs related to Unicode integration:
2098 ** the restarts for recovering from input and output encoding
2099 errors only appear when there is in fact such an error to
2101 * fixed some bugs revealed by Paul Dietz' test suite:
2102 ** MISC.549 and similar: late transformation of unsafe type
2103 assertions into derived types caused unexpected code
2105 ** SCALE-FLOAT type deriver is less wrong.
2106 ** type derivers for EXP, LOG and similar functions compute result
2107 types for complex arguments better.
2108 ** (MISC.563) CONJUGATE type deriver works for very restricted
2110 ** out-of-line type testers for character strings are available.
2111 ** EQUAL compiler transform understands specialness of objects
2113 ** accessing double-floats stored on the stack now works on x86-64.
2114 ** debugger internals could sometimes create invalid lispobjs,
2115 resulting in GC crashes.
2116 ** MISC.548: type check weakening can convert required type into
2118 ** initialization forms for bindings are not in scope of free special
2121 changes in sbcl-0.8.21 (0.9alpha.1?) relative to sbcl-0.8.20:
2122 * incompatible change: thread support for non-NPTL systems has
2123 been removed - locking is buggy and unreliable. A threaded
2124 SBCL build will now warn at startup and refuse to create
2125 new threads, unless futex support is detected in the kernel
2126 * incompatible change: the top level REPL now has only an ABORT
2127 restart associated with it, not TOPLEVEL and ABORT as it used to.
2128 TOP and TOPLEVEL are now available as debugger commands for
2129 returning to the top level.
2130 * incompatible change: forms evaluated in the REPL now use the
2131 global optimization policy.
2132 * incompatible change: user- and system-initialization files are
2133 no longer processed with LOAD, but by READ and EVAL; hence the
2134 global optimization policy, startup package, readtable, etc,
2136 * The .fasl file format number has been incremented because of
2137 various incompatible changes.
2138 * internal entry point details and argument counts no longer appear
2139 in backtraces unless explicitly requested by setting
2140 SB-DEBUG:*SHOW-ENTRY-POINT-DETAILS*.
2141 * built-in and standard functions no longer have names like "top
2142 level local call to FOO".
2143 * fixed bug 32: functions defined in non-null lexical environments
2144 now have more legible printed representation
2145 * fixed bug 33: functions defined in non-null lexical environemnts
2146 are now more amenable to inspection by INSPECT.
2147 * workaround for bug 354: XEPs no longer appear in backtraces unless
2148 explicitly requested.
2149 * fixed bug: receiving the signal which results from attempting to
2150 write to mprotect()ed memory (SIGSEGV on Linux and some *BSDs,
2151 SIGBUS on other *BSDs) on architectures where the C stack is also
2152 the Lisp stack (x86 and x86-64) from foreign code no longer leads
2153 to debugger confusion or wild instability. (reported by Cheuksan
2155 * fixed bug: COUNT and EQUAL no longer issue compiler efficiency
2156 notes when operating on objects known to be SIMPLE-BIT-VECTORs.
2157 (reported by Lutz Euler)
2158 * fixed bug: (TYPEP X '(MEMBER ...)) no longer returns a list in
2159 compiled code. (reported by Paul Dietz)
2160 * fixed bug 276b: mutating with MULTIPLE-VALUE-SETQ a binding of a
2161 specialized parameter to a method to something that is not TYPEP
2162 the specializer is now possible.
2163 * fixed bug: the MAKE-INSTANCE optimization is now correct in the
2164 face of package deletion.
2165 * fixed bug: LOAD should bind *LOAD-PATHNAME* to the merged
2166 pathname. (reported by David Tolpin on c.l.l)
2167 * contrib improvement: the SB-SIMPLE-STREAMS contrib now defines
2168 STRING-SIMPLE-STREAM and FILE-SIMPLE-STREAM as subclasses of
2169 STRING-STREAM and FILE-STREAM, respectively.
2170 * contrib improvement: SB-INTROSPECT handles more of SLIME's needs
2171 than previously; in addition, its test suite is now run on build.
2172 (thanks to Luke Gorrie)
2173 * a more robust x86-64 disassembler. (thanks to Lutz Euler)
2174 * optimization: added a immediate representation for single-floats
2176 * fixed some bugs revealed by Paul Dietz' test suite:
2177 ** MISC.564: defined out-of-line version of %ATAN2 on x86.
2178 ** attempting to create a package with a colliding nickname causes
2179 correctable errors to be signalled.
2180 ** MISC.572-574: :START1 and :START2 broken for simple-base-strings.
2181 ** several x86-64 backend bugs related to sign-extension of immediate
2184 changes in sbcl-0.8.20 (0.9alpha.0?) relative to sbcl-0.8.19:
2185 * fixed inspection of specialized arrays. (thanks to Simon Alexander)
2186 * fixed disassembly of SHLD and SHRD on x86. (thanks to David
2188 * fixed loading of multiply forward-referenced layouts.
2189 (thanks to Cheuksan Wang)
2190 * fixed bug 7: less verbose COMPILE-FILE output. Additionally, the
2191 output is now directed to *STANDARD-OUTPUT* as specified by ANSI.
2192 (see COMPILE-FILE documentation for details of :PRINT option)
2193 * fixed bugs 19 and 317: fixed-format floating point printing is
2194 more accurate. This also fixes a bug reported by Adam Warner
2195 related to the ~@F format directive.
2196 * fixed bug 371: bignum print/read inconsistency. (thanks to Harald
2198 * fixed bug: SET-SYNTAX-FROM-CHAR correctly shallow-copies a
2199 dispatch table if the from-char is a dispatch macro character.
2200 * fixed bug: COUNT and EQUAL on bit vectors with lengths divisible
2201 by the wordsize no longer ignore the last word. (reported by Lutz
2203 * fixed bug in type checking of dynamic-extent variables. (reported
2205 * optimization: sequence traversal functions use their freedom to
2206 coerce function designators to functions.
2207 * optimization: code with many calls to CLOS methods specialized on
2208 CLOS classes has had redundant type checks removed.
2209 * optimization: added declarations to speed up operations that access
2210 the internal character database (for example STRING-UPCASE)
2211 * optimization: comparison operations between floats and sufficiently small
2212 fixnums no longer create extra rationals
2213 * fixed some bugs related to Unicode integration:
2214 ** portions of multibyte characters at the end of buffers for
2215 character-based file input are correctly transferred to the
2216 start of the buffer at the next read.
2217 ** COMPILE-FILE now respects any EXTERNAL-FORMAT argument given,
2218 passing it through to OPEN.
2219 ** LOAD on source files likewise passes any EXTERNAL-FORMAT
2220 argument given to internal calls to OPEN.
2221 ** the built-in comment readers (introduced by character sequences
2222 ";" and "#|") are more forgiving to encoding errors; they will
2223 STYLE-WARN and then attempt to resync the stream at a character
2224 boundary. (thanks to Teemu Kalvas)
2225 * fixed some bugs in the x86-64 port:
2226 ** Negative short int return values from c-calls are sign-extended
2228 ** The stack is aligned to 16-bytes for c-calls, as required by
2229 the ABI. (thanks to Cheuksan Wang)
2230 ** The disassembler understands more x86-64. (thanks to Cheuksan Wang)
2231 ** The regression tests use SB-ALIEN:INT instead of SB-ALIEN:INTEGER
2232 for enums. (thanks to Vincent Arkesteijn)
2233 ** Multiple small optimizations and bugfixes for floating point
2235 * fixed some bugs revealed by Paul Dietz' test suite:
2236 ** Space, Tab, Linefeed, Return and Page have the invalid
2237 secondary constituent character trait.
2238 ** SET-SYNTAX-FROM-CHAR correctly copies multiple-escape character
2240 ** WITH-INPUT-FROM-STRING should only update the index place on
2242 ** Pretty-printing backquoted forms when *PRINT-CIRCLE* is true
2243 works more reliably.
2244 ** Bit-array operations (BIT-AND and similar) worked incorrectly
2245 with one-dimensional arrays with fill pointers.
2246 ** TYPE-OF failed on a complex with an integer realpart and a
2248 ** compiler failure during type inference for the code of form
2249 (IF (EQL X (THE ...)) ...) (MISC.535).
2251 changes in sbcl-0.8.19 relative to sbcl-0.8.18:
2252 * new port: SBCL now works in native 64-bit mode on x86-64/Linux
2253 platform. The port supports 61-bit fixnums, large memory spaces
2254 and reloading shared object files.
2255 * enhancement: saving cores with foreign code loaded is now
2256 supported on ppc/Darwin in addition to the previously supported
2258 * enhancement: the statistical profiler now walks deeper into the
2259 call stack for more meaningful call-graphs and accrued time
2260 reports (x86/x86-64 only). It also now reports time spent in
2262 * enhancement: it is now possible to trace most individual methods
2263 of a generic function in addition to tracing the generic function
2265 * bug fix: invalid :DEFAULT-INITARGS are detected in compiled calls
2267 * bug fix: defaulted initargs are passed to INITIALIZE-INSTANCE and
2268 SHARED-INITIALIZE methods from compiled calls to MAKE-INSTANCE.
2269 * bug fix: COERCE to (COMPLEX FLOAT) of a complex number no longer
2270 produces an error. (thanks to Vincent Arkesteijn)
2271 * bug fix: NAMESTRING on pathnames with :WILD components in their
2272 directories works correctly. (thanks to Artem V. Andreev)
2273 * fixed bug 125: compiler preserves identity of closures. (reported
2275 * bug fixed: functions with &REST arguments sometimes failed with
2276 "Undefined function" when compiled with (DEBUG 3). (reported by
2278 * bug fix: overflow during compiling of setting element of a bit
2279 vector with constant index and value. (reported by Timmy Douglas)
2280 * build fix: fixed the dependence on *LOAD-PATHNAME* and
2281 *COMPILE-FILE-PATHNAME* being absolute pathnames.
2282 * on x86 compiler partially supports stack allocation of dynamic-extent
2284 * GO and RETURN-FROM do not check the extent of their exit points
2285 when compiled with SAFETY 0.
2286 * fixed some bugs related to Unicode integration:
2287 ** encoding and decoding errors are now much more robustly
2288 handled; it should now be possible to recover even from invalid
2289 input or output to the terminal. (thanks to Teemu Kalvas)
2290 ** provided a first cut at implementing STRING-TO-OCTETS and
2291 OCTETS-TO-STRING. (thanks to Robert J. Macomber)
2292 ** altered the SB-MD5 contributed module slightly, changing the
2293 interface just enough for it to be supportable for builds where
2294 lisp characters are not eight bits.
2295 * fixed some bugs revealed by Paul Dietz' test suite:
2296 ** the FORMATTER-generated functions for ~V[ conditionals require
2297 the correct number of arguments.
2298 ** READ-FROM-STRING returns the mandated second value when applied
2299 to displaced strings.
2300 ** the #\Rubout and #\Backspace characters are treated as invalid
2301 constituent characters by the tokenizer.
2303 changes in sbcl-0.8.18 relative to sbcl-0.8.17:
2304 * new feature: reloading changed shared object files with
2305 LOAD-SHARED-OBJECT now causes the new definitions to take effect.
2306 * new feature: references to foreign variables and functions
2307 can now be compiled and loaded before the corresponding shared
2308 object file is loaded, as long as the foreign definitions are
2309 available at runtime.
2310 * Solaris 10 (aka SunOS 5.10) on the SPARC platform is now
2311 supported. (thanks to Dan Debertin)
2312 * SB-ALIEN enums can now be represented in Lisp by any symbols, not
2313 just keywords. (thanks to Vincent Arkesteijn)
2314 * fixed bug #331: structure-class instances corresponding to
2315 DEFSTRUCT forms are now created eagerly.
2316 * fixed bug #345: backtraces from calls to undefined functions work
2317 on x86 as well. Related bug #61 is now also partially fixed on x86
2318 (backtraces from throws to unknown catch tags.)
2319 * bug fix: lambda-list parsing is now stricter vrt. order and number
2320 of lambda-list keywords.
2321 * bug fix: as specified by AMOP, an error is signalled if a
2322 class-option appears multiple times in a DEFCLASS form. (reported
2324 * bug fix: RANDOM can be compiled when the compiler derives the type
2325 of its numeric argument as a disjoint set of small integers.
2326 (reported by Paul Dietz)
2327 * bug fix: starting a core saved with shared objects loaded when
2328 those objects are not available no longer causes threaded SBCL to
2329 hang. (reported by Sean Ross)
2330 * bug fix: evaluated FUNCTION no longer bypasses encapsulation (eg.
2332 * bug fix: (SETF MACRO-FUNCTION) now accepts an optional environment
2333 argument, which must always be NIL. (reported by Kalle Niemitalo)
2334 * bug fix: printing 1.0d+23 no longer results in an error.
2335 (reported by Rolf Wester for CMUCL; bug fix from Raymond Toy)
2336 * bug fix: structure slot setters preserve evaluation order. (thanks
2338 * fixed some bugs related to Unicode integration:
2339 ** RUN-PROGRAM can allow its child to take input from a Lisp
2340 stream. (reported by Stefan Scholl)
2341 ** ASDF-INSTALL successfully downloads .tar.gz files in all
2342 locales. (reported by Ken Causey)
2343 * fixed some bugs revealed by Paul Dietz' test suite:
2344 ** INCF, DECF and REMF evaluate their place form as specified in
2346 ** FORMATTER expands ~{ ~} iteration directives with V or #
2347 parameters correctly.
2348 ** FORMATTER deals with the ~@[ ~] conditional directive where the
2349 consequent uses no arguments correctly.
2350 ** the system has a partial understanding of the (COMPLEX RATIO)
2352 ** UPGRADED-COMPLEX-PART-TYPE obeys the lattice upgrading rules
2355 changes in sbcl-0.8.17 relative to sbcl-0.8.16:
2356 * new feature: a build-time option (controlled by the :SB-UNICODE
2357 keyword feature, enabled by default) for building the system with
2358 support for the entire 21-bit character space defined by the
2360 * new feature: the system now has rudimentary external-format
2361 support; the primary user-visible change at this time is that
2362 characters with the high bit set (such as lower-case-e-acute) will
2363 print correctly to a terminal in a UTF-8 environment.
2364 * minor incompatible change: BASE-CHAR no longer names a class;
2365 however, CHARACTER continues to do so, as required by ANSI.
2366 * minor incompatible change: SB-DEBUG:*DEBUG-PRINT-FOO* variables
2367 are no longer supported, and SB-DEBUG:*DEBUG-PRINT-VARIABLE-ALIST*
2368 has been moved to the SB-EXT package (temporarily re-exported from
2370 * minor incompatible change: SB-C::*COMPILER-ERROR-PRINT-FOO* variables
2371 are no longer supported: use SB-EXT:*COMPILER-PRINT-VARIABLE-ALIST*
2373 * fixed bug #351: better error-handlind and reporting for malformed
2375 * fixed bug #350: bignum-printing is now more memory-efficient,
2376 allowing printing of very large bignums, eg. (expt 2 10000000).
2377 (reported by Bruno Haible)
2378 * fixed bug #302: better primitive-type selection for intersection
2380 * fixed bug #308: non-graphic characters now all have names, as
2381 required. (reported by Bruno Haible)
2382 * bug fix: redefining a class with different superclasses now correctly
2383 removes it from the direct-subclasses of its previous superclasses.
2384 (reported by David Morse)
2385 * bug fix: (SETF FIND-CLASS) using a FORWARD-REFERENCED-CLASS as the
2386 new value now works. (reported by Bruno Haible)
2387 * bug fix: correct canonicalization of multiple non-standard slot
2388 options in DEFCLASS as per AMOP 5.4.2. (reported by Bruno Haible)
2389 * bug fix: SB-MOP:CLASS-PROTOTYPE now signals an error if the class
2390 is not yet finalized, as required by AMOP. (reported by Bruno Haible)
2391 * bug fix: SB-MOP:ALLOCATE-INSTANCE method for instances of BUILT-IN-CLASS
2392 now exists, an signals an error.
2393 * bug fix: duplicate LOOP variable bindings now signal PROGRAM-ERROR
2394 during macroexpansion for non-iteration variables as well. (reported
2395 by Bruno Haible for CMUCL)
2396 * bug fix: Cyclic structures and unprintable objects in compiler
2397 messages no longer cause errors. (reported by Bruno Haible)
2398 * bug fix: READ, READ-PRESERVING-WHITESPACE, READ-DELIMITED-LIST,
2399 and READ-FROM-STRING all now return a primary value of NIL if
2400 *READ-SUPPRESS* is true. (reported by Bruno Haible for CMUCL)
2401 * bug fix: Default value of EOF-ERROR-P in READ-FROM-STRING is true.
2402 (reported by Bruno Haible for CMUCL)
2403 * bug fix: ERROR now signals a TYPE-ERROR if the arguments to ERROR
2404 do not designate a condition. (reported by Bruno Haible for
2406 * bug fix: UNINTERN, USE-PACKAGE, IMPORT and EXPORT all signal an
2407 SB-EXT:NAME-CONFLICT condition (subtype of PACKAGE-ERROR) in the
2408 name conflict situations in CLHS 11.1.1.2.5, and provide a restart
2409 permitting resolution in favour of any of the conflicting symbols.
2410 (reported by Bruno Haible for CMUCL)
2411 * bug fix: EQUAL compiler optimizations is less aggressive on
2412 strings which can potentially compare true despite having distinct
2413 specialized array element types.
2414 * bug fix: unit enumerations can be defined without dividing by
2415 zero. (thanks to Vincent Arkesteijn)
2416 * FORMAT compile-time argument count checking has been enhanced.
2417 (report from Bruno Haible for CMUCL)
2418 * a partial workaround for the bug 262: the compiler does not try to
2419 inline-expand a local function doing RETURN-FROM from a deleted
2420 BLOCK. (thanks to Peter Denno for the bug report and to David
2421 Wragg for the simple test case)
2422 * fixed some bugs revealed by Paul Dietz' test suite:
2423 ** RENAME-PACKAGE allows all package designators as new package
2425 ** constraint propagation and lambda variable substitution are
2426 more cautious in dealing with partially deleted code.
2427 ** compiler sometimes lost reoptimization passes.
2428 ** CERROR, given a condition as condition designator, passes its
2429 remaining arguments to the continue format control without
2431 ** Case-altering FORMAT directives work correctly on non-ASCII
2433 ** The REAL type specifier handles bounds outside the floating
2434 point ranges without signalling FLOATING-POINT-OVERFLOW.
2435 ** Functions with IR1-transformations can create intercomponent
2436 references to global functions.
2437 ** NIL parameter to the FORMAT directive ~^ means `unsupplied
2439 ** FORMAT ~R treats a nil value for its first parameter correctly.
2441 changes in sbcl-0.8.16 relative to sbcl-0.8.15:
2442 * enhancement: saving cores with foreign code loaded is now
2443 supported on x86/NetBSD and sparc/Linux in addition to the previously
2444 supported platforms.
2445 * bug fix: on some platforms repeated installations caused multiple
2446 copies of HTML documentation to be installed -- should not happen
2447 any more. (reported by Stefan Scholl)
2448 * bug fix: parsing self-recursive alien record types multiple times
2449 no longer causes infinite recursion. (reported by Thomas F. Burdick,
2450 original patch by Helmut Eller for CMUCL)
2451 * bug fix: stack-exhaustion detection works now on NetBSD as well.
2452 (thanks to Richard Kreuter)
2453 * bug fix: defining classes whose accessors are methods on existing
2454 generic functions in other (locked) packages no longer signals
2455 bogus package lock violations. (reported by François-René Rideau)
2456 * bug fix: special variables as DEFMETHOD parameters no longer have
2457 associated bogus type declarations. (reported by David Wragg and
2459 * bug fix: read-write consistency on streams of element-type
2460 (SIGNED-BYTE N) for N > 32. (reported by Bruno Haible for CMUCL)
2461 * bug fix: redefiniton of the only method of a generic function with
2462 no DEFGENERIC no longer emits a full WARNING. In addition,
2463 redefinition of generic functions with no DEFGENERIC to an
2464 incompatible lambda list now signals an error. (thanks to Zach
2466 * bug fix: DEFGENERIC now works even when there's a function of the
2467 same name in an enclosing lexical environment. (thanks to Zach
2469 * fixed compiler failure, caused by instrumenting code during
2470 IR1-optimization. (Debian bug report #273606 by Gabor Melis)
2471 * optimization: added loop analysis and improved register allocation
2472 * fixed some bugs revealed by Paul Dietz' test suite:
2473 ** POSITION on displaced vectors with non-zero displacement
2474 returns the right answer.
2475 ** (SIMPLE-STRING) is a valid type specifier for sequence
2477 ** *PRINT-LEVEL* handling for slotless structures is pedantically
2479 ** PPRINT-INDENT accepts a request for an indentation of any REAL.
2480 ** PPRINT-TAB (and the FORMAT ~T directive) now indent by the
2482 ** The justification version of the FORMAT ~< directive treats
2483 non-zero minpad parameter correctly.
2485 changes in sbcl-0.8.15 relative to sbcl-0.8.14:
2486 * incompatible change: SB-INT:*BEFORE-SAVE-INITIALIZATIONS* and
2487 SB-INT:*AFTER-SAVE-INITIALIZATIONS* have been renamed
2488 SB-EXT:*SAVE-HOOKS* and SB-EXT:*INIT-HOOKS*, and are now part of
2489 the supported interface.
2490 * new feature: Single-stepping of code compiled with DEBUG 2 or
2491 higher and (> DEBUG (MAX SPACE SPEED)) is now possible.
2492 * new feature: saving cores with foreign code loaded is now
2493 supported on x86/FreeBSD, x86/Linux, and sparc/SunOS. (based on
2494 Timothy Moore's work for CMUCL)
2495 * bug fix: DEFTYPE lambda-list parsing now binds unsupplied keyword
2496 parameters to * instead of NIL if no initform is supplied.
2497 (reported by Johan Bockgaard)
2498 * bug fix: DEFINE-COMPILER-MACRO lambda-list parsing now binds
2499 correctly when FUNCALL appears as the car of the form. Note:
2500 despite this FUNCALL forms are not currently subject to
2501 compiler-macro expansion. (port of Raymond Toy's fix for the
2502 same from CMUCL, reported by Johan Bockgaard)
2503 * bug fix: FOR ... ON ... -clauses in LOOP now work on dotted lists
2504 (thanks for Teemu Kalvas)
2505 * bug fix: in FORMAT ~^ inside ~:{ now correctly steps to the next
2506 case instead of terminating the iteration (thanks for Julian
2507 Squires, Sean Champ and Raymond Toy)
2508 * bug fix: incorrect expansion of defgeneric that caused a style
2509 warning. (thanks for Zach Beane)
2510 * bug fix: slot accessor effective method computation works properly
2511 for classes with multiple non-standard applicable methods on
2512 SB-MOP:SLOT-VALUE-USING-CLASS. (reported by Ralf Mattes)
2513 * on x86 compiler supports stack allocation of results of LIST and
2514 LIST*, bound to variables, declared DYNAMIC-EXTENT. (based on
2515 CMUCL implementation by Gerd Moellmann)
2516 * fixed some bugs revealed by Paul Dietz' test suite:
2517 ** FORMAT strings with both the ~<~:;~> form of the justification
2518 directive and pretty-printing directives cause an error.
2520 changes in sbcl-0.8.14 relative to sbcl-0.8.13:
2521 * incompatible change: the internal functions
2522 SB-KERNEL:32BIT-LOGICAL-FOO, intended for providing efficient
2523 logical operations on (UNSIGNED-BYTE 32) values, have been renamed
2524 to SB-KERNEL:WORD-LOGICAL-FOO. Modular arithmetic should be used
2525 instead of the old functions.
2526 * new feature: on platforms where "dladdr" is available foreign
2527 function names now appear in backtraces. (based on Helmut Eller's
2529 * documentation: documentation for SB-BSD-SOCKETS, SB-GRAY,
2530 SB-SIMPLE-STREAMS, and SB-PROFILE have been integrated into the user
2532 * documentation: ASDF manual is now distributed with SBCL.
2533 * bug fix: SBCL can now load its contributed modules with REQUIRE
2534 even if the system-provided entries have been removed from
2535 ASDF:*CENTRAL-REGISTRY*. Infinite recursion in REQUIRE is also
2537 * bug fix: backtraces involving undefined functions or assembly
2538 routines are more informative. (thanks to Brian Downing)
2539 * bug fix: mutually referent alien structures now work correctly.
2540 (reported by Rick Taube)
2541 * bug fix: structures defined by WITH-ALIEN can be referred to
2542 within other definitions in the same WITH-ALIEN.
2543 * bug fix: division operators (MOD, TRUNCATE and the like) with
2544 constant zero divisors and integer dividends no longer generate
2546 * bug fix: provide default methods for INPUT-STREAM-P and
2547 OUTPUT-STREAM-P specialized on SB-GRAY:FUNDAMENTAL-STREAM.
2548 * bug fix: improve the detection and resolution of MOP metacycles
2549 (where effective method computation depends on the generic function
2550 having its effective method computed). (reported by Bruno Haible)
2551 * bug fix: pass the right initargs to :BEFORE or :AFTER methods on
2552 SHARED-INITIALIZE or INITIALIZE-INSTANCE in optimized
2553 MAKE-INSTANCE constructors. (reported by Axel Schairer for cmucl)
2554 * bug fix: pathnames with relative directory components can be
2555 represented relative to default pathnames.
2556 * optimization: in taking the GCD of bignums, reduce the two bignums
2557 to approximately the same size (using Euclid's algorithm) before
2558 applying the more sophisticated binary GCD. (thanks to Juho
2560 * optimization: COUNT on bitvectors now operates word-at-a-time.
2561 * optimization: ASH with a positive, but not necessarily constant,
2562 (leftwards) shift, when in a modular context, is compiled to a
2564 * fixed some bugs revealed by Paul Dietz' test suite:
2565 ** FORMAT variable parameters ("~V<char>") are defaulted properly
2566 if the corresponding argument is NIL.
2567 ** FORMAT directives accepting a minpad parameter treat negative
2569 ** literal commas as character parameters to FORMAT directives are
2571 ** literal spaces directly after ~<Newline> directives within a
2572 format-logical-block (~:< ~@:>) do not induce :FILL-style
2573 conditional newlines.
2574 ** PRINT-UNREADABLE-OBJECT inserts spaces as specified (and only
2575 as specified: it no longer includes conditional newlines).
2576 ** PRINC-TO-STRING binds *PRINT-READABLY* to NIL (as well as
2578 ** PPRINT-LOGICAL-BLOCK signals a TYPE-ERROR if its :PREFIX or
2579 :PER-LINE-PREFIX argument does not evaluate to a string.
2581 changes in sbcl-0.8.13 relative to sbcl-0.8.12:
2582 * new feature: SB-PACKAGE-LOCKS. See the "Package Locks" section of
2583 the manual for details; for now, package locks can be disabled by
2584 removing :SB-PACKAGE-LOCKS in customize-target-features.lisp, but
2585 if no major problems are found then it is likely that they will be
2586 compiled in unconditionally.
2587 * major incompatible change: LOAD-FOREIGN and LOAD-1-FOREIGN are now
2588 unsupported operators on all platforms. To load a shared library
2589 into SBCL, use SB-ALIEN:LOAD-SHARED-OBJECT. To load a non-shared
2590 object file, link it into a shared library outside of SBCL and
2591 load it using SB-ALIEN:LOAD-SHARED-OBJECT. (incidentally fixes
2593 * bug workaround: ROOM T and the GC were not working together
2594 reliably, because invariants expected by the SB!VM:INSTANCE-USAGE
2595 reporting facility aren't preserved. That reporting has been
2596 disabled, so now until and unless someone figures out how to make it
2597 work reliably with the current GC, (ROOM T) is equivalent to (ROOM).
2598 * minor incompatible change: as threatened around sbcl-0.8.0, the
2599 home package of MOP-related symbols is now SB-MOP, not SB-PCL.
2600 The symbols are also exported from SB-PCL for backwards
2601 compatibility, but more so than before SB-PCL should be treated as
2602 an implementation-internal package.
2603 * fasl format changed: SBCL 0.8.13 fasls are incompatible with those
2605 * the SB-SPROF contrib now works on (most) non-x86 architectures.
2606 It is known as of this release not to work on the Alpha, however.
2607 * fixed bug #167: errors signalled due to illegal syntax in method
2608 bodies are now more legible.
2609 * fixed bug #338: instances of EQL-SPECIFIER are now valid type
2610 designators and can hence be used with TYPEP.
2611 * fixed bug #333: CHECK-TYPE now ensures that the type error
2612 signalled, if any, has the right object to be accessed by
2613 TYPE-ERROR-DATUM. (reported by Tony Martinez)
2614 * fixed bug #340: SETF of VALUES obeys the specification in ANSI
2615 5.1.2.3 for multiple-value place subforms. (reported by Kalle
2617 * fixed bug #334: programmatic addition of slots using specialized
2618 methods on SB-MOP:COMPUTE-SLOTS works for :ALLOCATION :INSTANCE
2619 and :ALLOCATION :CLASS slots.
2620 * fixed bug #269: SCALE-FLOAT scales floats by any integer, not just
2621 float exponents. (rereported by Peter Seibel)
2622 * fixed a bug: #\Space (and other whitespace characters) are no
2623 longer considered to be macro characters in standard syntax by
2624 GET-MACRO-CHARACTER.
2625 * fixed bug: initialization of condition class metaobjects no longer
2626 causes an instance of the condition to be created. (reported by
2628 * fixed bug: it is now possible to have slots such that
2629 SB-MOP:SLOT-DEFINITION-ALLOCATION of the effective slot
2630 description is neither :INSTANCE nor :CLASS.
2631 * fixed bug: the ctor optimization of MAKE-INSTANCE now respects
2632 user-defined methods on SLOT-BOUNDP-USING-CLASS and (SETF
2633 SLOT-VALUE-USING-CLASS), and no longer causes errors with
2634 non-standard SLOT-DEFINITION-ALLOCATION values.
2635 * fixed bugs: various race conditions handling exiting threads.
2636 CL-PPCRE's thread torture test now passes.
2637 * fixed arguably-a-bug: GC time proportional to number of threads in
2638 system even when most of them are idle
2639 * optimization: improved performance of BIT and SBIT on bit-vectors.
2640 * .fasl file incompatibility: The fasl file version number has
2641 been incremented because of changes associated with package locks.
2643 changes in sbcl-0.8.12 relative to sbcl-0.8.11:
2644 * minor incompatible change: the system no longer provides
2645 optimization hints (conditions which are TYPEP
2646 SB-EXT:COMPILER-NOTE) for conforming code in default compilation
2647 mode; these hints are still emitted when higher SPEED optimization
2649 * new contrib module: a sampling profiler (profiling by statistical
2650 sampling, rather than by instrumenting functions) is available as
2651 the SB-SPROF contrib. (thanks to Gerd Moellmann and Juho Snellman)
2652 * the behaviour of the standard function ED is now customizeable by
2653 third parties through a hook variable: see ED's documentation
2654 string for information on the protocol.
2655 * the compiler no longer emits efficiency notes for (FUNCALL X)
2656 when the type of X is uncertain under default optimization
2658 * fixed bug 276: mutating a binding of a specialized parameter to a
2659 method to something that is not TYPEP the specializer is now
2661 * fixed bugs 45d and 118: DOUBLE-FLOAT[-NEGATIVE]-EPSILON now
2662 exhibit the required behaviour on the x86 platform. (thanks to
2663 Peter van Eynde, Eric Marsden and Bruno Haible)
2664 * fixed bug 335: ATANH now computes the inverse hyperbolic tangent
2665 even for difficult arguments. (reported by Peter Graves)
2666 * fixed bug 141a: the backquote printer now descends quoted
2668 * fixed another bug in backquote printing: no more destructive
2669 modification of the form's list structure. (reported by Brian
2671 * fixed bug in INTERRUPT-THREAD: pin the function, so that it cannot
2672 move between its address being taken and the call to
2673 interrupt_thread, fixing a crashing race condition.
2674 * the SB-POSIX contrib implementation has been adjusted so that it
2675 no longer exhibits ridiculously poor performance when constructing
2676 instances corresponding to C structs.
2678 changes in sbcl-0.8.11 relative to sbcl-0.8.10:
2679 * minor incompatible change: the sb-grovel contrib now treats C
2680 structures as alien (in the sense of SB-ALIEN) objects rather than
2681 as undistinguished (simple-array (unsigned-byte 8) (*))s. This
2682 has implications for memory management of client code
2683 (sb-grovel-returned objects must now be manually managed) and for
2684 type safety (alien objects now have full types).
2685 * new feature: the SB-EXT:MUFFLE-CONDITIONS declaration should be
2686 used to control emission of compiler diagnostics, rather than the
2687 SB-EXT:INHIBIT-WARNINGS OPTIMIZE quality. See the manual for
2688 documentation on this feature. The SB-EXT:INHIBIT-WARNINGS
2689 quality should be considered deprecated.
2690 * install.sh now installs the user manual as well
2691 * (not quite a new documentable feature, but worth considering in
2692 the light of the new SB-EXT:MUFFLE-CONDITIONS declaration): the
2693 beginnings of a semantically meaningful condition hierarchy is
2694 under development, for use in SB-EXT:MUFFLE-CONDITIONS and by
2696 * fixed bug: PARSE-NAMESTRING now accepts any valid pathaname
2697 designator as the defaults argument.
2698 * fixed bug: Displaced arrays whose displaced-to array has become
2699 too small now cause ARRAY-DIMENSION to signal an error, providing
2700 for safer bounds-checking. (reported by Bruno Haible)
2701 * fixed bug: DEFCLASS slot definitions with identical :READER and
2702 :WRITER names now signal a reasonable error. (reported by Thomas
2704 * fixed bug: CLOSE :ABORT T on appending stream no longer causes
2706 * fixed bug: Invalid slot specification errors now print correctly.
2707 (thanks to Zach Beane)
2708 * fixed bug 320: Shared to local slot value transfers in class
2709 redefinitions now happen corrently from superclasses as
2710 well. (reported by Bruno Haible)
2711 * fixed bug 316: SHIFTF now accepts VALUES forms. (reported by Bruno
2713 * fixed bug 322: DEFSTRUCT :TYPE LIST type predicates now handle
2714 improper lists correctly. (reported by Bruno Haible)
2715 * fixed bug 313: source-transform for <fun-name> was erroneously
2716 applied to a call of a value of a variable with name <fun-name>.
2717 (reported by Antonio Menezes Leitao)
2718 * fixed bug 307: The obsolete instance protocol ensures that
2719 subclasses are properly obsoleted. (thanks to Nikodemus Siivola)
2720 * fixed bug 298, revealed by Paul F. Dietz' test suite: SBCL can
2721 remove dead unknown-values globs from the middle of the stack.
2722 * added a new restart to *BREAK-ON-SIGNALS* handling to make it
2723 easier to resume long computations after using *BREAK-ON-SIGNALS*
2724 to diagnose and fix failures. (thanks to Nikodemus Siivola)
2725 * fixed bug reported by PFD in lisppaste #747 (and Bruno Haible from
2726 CLISP test suite): WRITE-TO-STRING is not constant-foldable.
2727 * fixed bugs in COMPLEX type specifier: UPGRADED-COMPLEX-PART-TYPE
2728 is now consistent with (COMPLEX <x>); bugs in treatment of COMPLEX
2729 MEMBER and UNION types have likewise been fixed. (thanks to Bruno
2731 * fixed a (fairly theoretical) bug in string printing: if
2732 *PRINT-READABLY* is true, signal PRINT-NOT-READABLE if the string
2733 does not have array-element-type equal to the most general string
2735 * fixed bug: SET-PPRINT-DISPATCH does not immediately resolve
2736 function name. (thanks to Nikodemus Siivola)
2737 * fixed bug: compile-time format string checker failed on
2738 non-closed ~{. (reported by Thomas F Burdick)
2739 * fixed bug: as reported by Kalle Olavi Niemitalo on #lisp IRC,
2740 don't warn on legal loop code involving more than one aggregate
2741 boolean. (thanks to Nikodemus Siivola)
2742 * fixed bug: as reported by Peter Graves on #lisp IRC, passing a NIL
2743 in keyword position to MAKE-PACKAGE was not being reported as
2745 * fixed bug: as reported by Juan Ripoll on cmucl-imp,
2746 MULTIPLE-VALUE-BIND should be able to lexically bind lambda list
2748 * fixed bugs 280 and 312: the checking for multiple definitions in a
2749 file is less likely to become confused by uses of inline
2751 * fixed bug: the #S reader macro performs the keyword coercion
2752 specified for slot names. (reported by Kalle Niemitalo)
2753 * fixed bug: lambda lists may contain symbols whose names start with
2754 & but are not lambda-list-keywords; their occurrence triggers a
2756 * fixed bug 321: define-method-combination argument lambda lists do
2757 not cause invalid code to be generated when &optional or &aux
2758 variables have default values. (reported by Bruno Haible)
2759 * fixed bug 327: system subclasses of STRUCTURE-OBJECT or CONDITION
2760 have CLOS classes; this appears also to have fixed reported
2761 instability in stack exhaustion detection.
2762 * fixed bug: the CONTROL-ERROR from ABORT, CONTINUE and
2763 MUFFLE-WARNING when no associated restart is present is now
2765 * optimization: rearranged the expansion of various defining macros
2766 so that each expands into only one top-level form in a
2767 :LOAD-TOPLEVEL context; this appears to decrease fasl sizes by
2769 * optimization: used a previously little-used slot in symbols to
2770 cache SXHASH values, yielding a 5-10% compiler speedup. (thanks
2772 * fixed some bugs revealed by Paul Dietz' test suite:
2773 ** MAKE-INSTANCES-OBSOLETE returns the class name when called with
2775 ** Fixed an optimization bug related to inheritance of initforms
2776 from local to shared slots.
2777 ** FILE-POSITION works as specified on BROADCAST-STREAMs.
2778 ** CAST optimizer forgot to flush argument derived type.
2779 ** print/read consistency on floats is now orders of magnitude
2780 more likely. (thanks also to Bruno Haible for a similar report
2782 ** removed stack cleaning in the cleanup part of UNWIND-PROTECT.
2783 ** IMAGPART is specified (infelicitously) to return (* 0 <thing>)
2784 for objects of type REAL. Make it so.
2785 ** SXHASH is specified (infelicitously) to respect similarity,
2786 which means that (SXHASH 0.0) must equal (SXHASH -0.0). Make
2787 it so. (thanks to Markus Ziegler)
2788 ** on the Alpha, the compiler succeeds in compiling functions
2789 returning a known number of arguments greater than 63.
2790 ** fixed handling of invalid NIL arguments in keyword position
2792 ** fixed non-pretty printing of arrays with *PRINT-RADIX* being
2794 ** provided a readably-printable representation for RANDOM-STATE
2796 ** ensured that pathnames, where they have a namestring, always
2797 print using #P"..." syntax.
2799 changes in sbcl-0.8.10 relative to sbcl-0.8.9:
2800 * Support for the forthcoming 2.0 version of the NetBSD kernel
2801 running on x86 hardware has been added. (thanks to Perry
2802 E. Metzger most immediately, and others for their past work)
2803 * SBCL now runs on OpenBSD 3.4. (Thanks to Scott Parish; 3.4 is the
2804 current release version; SBCL's OpenBSD support had been broken
2805 since about the time of OpenBSD's switch to ELF binary format.)
2806 * [placeholder for DX summary]
2807 ** user code with &REST lists declared dynamic-extent, under high
2808 speed or space and low safety and debug optimization policy.
2809 * The manual has been converted to Texinfo format and the debugger
2810 chapter from the cmucl manual has been added.
2811 * A facility has been added to extract documentation strings from
2812 sbcl and store them as Texinfo-formatted snippets for inclusion in
2813 the manual (via Texinfo's @include directive)
2814 * bug fix: compiler emitted division in optimized DEREF. (thanks for
2815 the test case to Dave Roberts)
2816 * bug fix: multidimensional simple arrays loaded from FASLs had fill
2817 pointers. (reported by Sean Ross)
2818 * bug fix: PROFILE output is printed nicely even for large numerical
2819 values. (thanks to Zach Beane)
2820 * bug fix: streams with element-type (SIGNED-BYTE <N>) for <N>
2821 greater than 32 handle EOF correctly.
2822 * bug fix: on X86 an immediate argument of the IMUL instruction is
2823 correctly printed in disassembly. (thanks to Lutz Euler)
2824 * bug fix: class slots in redefined classes preserve their old
2825 values. (thanks to Bruno Haible and Nikodemus Siivola)
2826 * bug fix: compilation of funcalls of CXX+R (e.g. CDDR) now
2827 succeeds. (reported by Marco Baringer)
2828 * fixed some bugs revealed by Paul Dietz' test suite:
2829 ** READ-SEQUENCE now works on ECHO-STREAMs.
2830 ** RATIONALIZE works more according to its specification. (thanks
2833 changes in sbcl-0.8.9 relative to sbcl-0.8.8:
2834 * deprecation of old extension: *DEBUG-PRINT-LEVEL* and
2835 *DEBUG-PRINT-LENGTH* are now deprecated in favor of the new, more
2836 general SB-DEBUG:*DEBUG-PRINT-VARIABLE-ALIST* mechanism. (This
2837 should matter to you only if you rebind the printer control
2838 variables and then find you want different bindings in the
2839 debugger than in the ordinary execution of your program.)
2840 * The runtime build system has been tweaked to support building
2841 (on SPARC/SunOS) using a C compiler which invokes Sun's own
2842 assembler and linker. (thanks to Nikodemus Siivola)
2843 * Unbound, undefined, undeclared variables now trigger full
2844 WARNINGs, not just STYLE-WARNINGs, on the assumption that this is
2845 more often programmer error than deliberate exploitation of undefined
2847 * optimization: the hash algorithm for strings has changed to one
2848 that is less vulnerable to spurious collisions. (thanks to Juho
2850 * optimization: VECTOR-POP, VECTOR-PUSH-EXTEND and REPLACE do less
2851 needless bounds checking. (thanks to Juho Snellman)
2852 * optimization: implemented multiplication as a modular
2853 (UNSIGNED-BYTE 32) operation on the PPC backend.
2854 * fixed some bugs revealed by Paul Dietz' test suite:
2855 ** ADJUST-ARRAY now copies the datum in a zero rank array if
2857 ** ADJUST-ARRAY no longer adjusts non-adjustable arrays.
2858 ** MAKE-STRING-INPUT-STREAM accepts NIL as a value for :END.
2859 ** MAKE-STRING-INPUT-STREAM functions correctly for strings with
2861 ** CLEAR-INPUT accepts NIL and T for its (optional) stream
2863 ** Ratios can now be printed correctly with *PRINT-BASE* bound to
2865 ** ECHO-STREAMs no longer attempt to echo the end of file value to
2866 their output stream on EOF from read.
2867 ** CONCATENATED-STREAM-STREAMS discards constituent streams which
2868 have been read to end-of-file.
2869 ** CLOSE works as expected on the null CONCATENATED-STREAM, and on
2871 ** Printing symbols with *PRINT-CASE* :CAPITALIZE respects the
2872 description of determination of which consecutive characters
2874 ** Printing the "Space" character with escaping on now yields "#\\ ",
2875 rather than "#\\Space", as mandated by ANSI 22.1.3.2.
2876 ** Reading floating-point numbers with *READ-BASE* set to a number
2877 less than 10 works correctly.
2878 ** Reading floating-point numbers with *READ-BASE* set to a number
2879 more than 10 works correctly.
2880 ** Printing with *PRINT-READABLY* targets the standard readtable, not
2881 the readtable currently in effect.
2883 changes in sbcl-0.8.8 relative to sbcl-0.8.7:
2884 * minor incompatible change: parsing of namestrings on a physical
2885 (Unix) host has changed; numbers after the final #\. in a
2886 namestring are no longer interpreted as a version field. This is
2887 intented to be largely invisible to the user, except that the
2888 meaning of the namestring "*.*.*" has changed: it now refers to a
2889 pathname with :TYPE :WILD :NAME #<pattern "*.*">. This namestring
2890 should usually be replaced by
2891 (make-pathname :name :wild :type :wild :version :wild)
2892 with the added benefit that this is more likely to be portable.
2893 As a consequence of this change, the :IF-EXISTS :NEW-VERSION
2894 option to OPEN now signals an error if the file being opened
2895 exists; this may have an impact on existing code.
2896 * fixed bug 190: RUN-PROGRAM should now work properly, respecting
2897 signals received, on the PowerPC platforms (both Linux and
2898 Darwin). (thanks to Pierre Mai for pointing out the location of
2900 * several fixes on OS X: The system now builds and runs cleanly on
2901 Panther (10.3), and works around sigreturn bug (no more SIGFPEs).
2902 (thanks to Brian Mastenbrook)
2903 * bug fix: DECODE-UNIVERSAL-TIME now accepts timezone arguments with
2904 second-resolution: integer multiples of 1/3600 between -24 and 24.
2905 (thanks to Vincent Arkesteijn)
2906 * bug fix: functions =, /=, <, <=, >, >= did not check the argument
2907 type when called with 1 argument; PEEK-CHAR checked type of
2908 PEEK-TYPE only after having read first character from a
2909 stream. (reported by Peter Graves)
2910 * bug fix: the garbage collector now has much better locality
2911 behaviour, and in particular no longer treats all memory as being
2912 exhausted when many small objects point to each other in a deeply
2914 * bug fix: arrays specialized on (UNSIGNED-BYTE 15) are now
2915 recognized as being TYPEP their class.
2916 * bug fix: the PUSHNEW documentation string has been corrected.
2917 (thanks to Vincent Arkesteijn)
2918 * bug fix: defaulting of the value for the last of an atypically
2919 large number of multiple values being bound was not being
2920 performed correctly on the Alpha or PPC platforms
2921 * optimization: implemented multiplication as a modular
2922 (UNSIGNED-BYTE 32) operation on the x86 backend.
2923 * optimization: SEARCH on simple-base-strings can now be open-coded.
2924 (see also contrib/compiler-extras.lisp for inspiration for
2925 teaching the compiler about the Boyer-Moore algorithm).
2926 * value, returned by MAX (and MIN) called with several EQUALP, but
2927 not EQL, arguments now does not depend on compiler settings.
2928 * fixed some bugs revealed by Paul Dietz' test suite:
2929 ** in stack analysis liveness information is propagated from
2930 non-local entry points.
2931 ** pathwise CAST removing failed when the CAST node did not start
2933 ** INPUT-STREAM-P, OUTPUT-STREAM-P, STREAM-ELEMENT-TYPE and
2934 OPEN-STREAM-P signal a TYPE-ERROR if their argument is not a
2936 ** LOAD-LOGICAL-PATHNAME-TRANSLATIONS returns NIL if the logical
2937 host is already defined.
2938 ** RENAME-FILE works on streams instead of signalling an internal
2940 ** PEEK-CHAR uses the current readtable when determining whether
2941 or not a character is whitespace.
2942 ** MERGE-PATHNAMES handles the case when the pathname does not
2943 specify a name while the default-pathname specifies a version.
2944 ** Pathnames now stand a better chance of respecting print/read
2946 ** Attempting to use standardized file system operators with a
2947 pathname with invalid :DIRECTORY components signals a
2949 ** OPEN :DIRECTION :IO no longer fails to work on non-existent
2951 ** DIRECTORY on logical pathnames is more correct.
2952 ** CLEAR-INPUT, CLEAR-OUTPUT, FINISH-OUTPUT and FORCE-OUTPUT
2953 signal a TYPE-ERROR if their argument is not a stream.
2954 ** READ-BYTE and WRITE-BYTE signal a TYPE-ERROR if their stream
2955 designator argument does not designate a stream.
2956 ** OPEN-STREAM-P and INPUT-STREAM-P on synonym streams work by
2957 examining the synonym.
2958 ** STREAM-ELEMENT-TYPE and FRESH-LINE on broadcast-streams now
2960 ** OPEN and WITH-OPEN-STREAM allow opening streams with
2961 element-type larger than ([UN]SIGNED-BYTE 32).
2963 changes in sbcl-0.8.7 relative to sbcl-0.8.6:
2964 * When built with the :SB-FUTEX feature, threaded builds now take
2965 advantage of the "fast userspace mutex" facility in Linux kernel 2.6
2966 for faster/more reliable mutex and condition variable support.
2967 * Incompatible change (but one you probably shouldn't have been using
2968 anyway): the interface and code for arbitrating between multiple
2969 threads in the same user session has been redesigned.
2970 * bug fix: GET-SETF-EXPANSION no longer throws an internal type
2971 error when called without an explicit environment argument.
2972 (thanks to Nikodemus Siivola)
2973 * bug fix: buffered :DIRECTION :IO streams are less likely to become
2974 confused about their position. (thanks to Adam Warner and Gerd
2976 * bug fix: Pretty printing backquoted forms with unquotations in the
2977 argument list position of various code constructs such as LAMBDA
2978 now works correctly. (reported by Paul Dietz)
2979 * bug fix: Pretty printing unquotations no longer loses all
2980 stream position information.
2981 * optimization: performance of string output streams is now less
2982 poor for multiple small sequence writes.
2983 * optimization: performance of CSUBTYPEP in the presence of complex
2984 expressions involving CONS and NOT many times has been improved.
2985 (reported by Paul Dietz)
2986 * ASDF-INSTALL bug fix: now parses *PROXY* properly. (thanks to
2988 * SB-SIMPLE-STREAMS enhancement: simple-streams can now be used as
2989 streams for the REPL, for the debugger, and so on. (thanks to
2991 * DEFINE-CODITION is more efficient. (thanks to Brian Mastenbrook)
2992 * fixed some bugs revealed by Paul Dietz' test suite:
2993 ** the value of the :REHASH-THRESHOLD argument to MAKE-HASH-TABLE
2994 is ignored if it is too small, rather than propagating through
2995 to cause DIVIDE-BY-ZERO or FLOATING-POINT-OVERFLOW errors.
2996 ** extremely complex negations of CONS types were not being
2997 sufficiently canonicalized, leading to inconsistencies in
2999 ** VALUES tranformer lost derived type.
3001 changes in sbcl-0.8.6 relative to sbcl-0.8.5:
3002 * fixed a bootstrapping bug: the build process no longer assumes
3003 that the various BOOLE-related constants have the same value in
3004 host and target lisps. (noted by Paul Dietz' test suite on an
3005 SBCL binary built from CLISP)
3006 * The system can now be dynamically linked on the MIPS platform,
3007 which enables dynamic loading of foreign code from Lisp. (thanks
3008 to Ralf Baechle for discussions on the MIPS ABI)
3009 * The system now records debugging information for its own source
3010 files in a filesystem-position-independent manner, relative to
3011 the "SYS" logical host.
3012 * fixed a compiler bug: MV-LET convertion did not check references
3013 to the "max args" entry point. (reported by Brian Downing)
3014 * tweaked disassembly notes to be less confident about proclaiming
3015 some instruction as an LRA. (thanks to Brian Downing)
3016 * contrib update: SB-ACLREPL is now threadsafe; multiple listeners
3017 now each have their own history, command character, and other
3018 characteristics. (thanks to David Lichteblau)
3019 * fixed some bugs revealed by Paul Dietz' test suite:
3020 ** compiler failure in compiling LOGAND expressions including a
3022 ** Implementation of ASH-MOD32 on X86 and PPC did not work for the
3023 shift greater than 32.
3024 ** FLUSH-DEST did not mark blocks for type check regeneration.
3025 ** HANDLER-CASE failed to accept declarations in handler clauses
3026 in some circumstances.
3028 changes in sbcl-0.8.5 relative to sbcl-0.8.4:
3029 * New code in contrib/sb-introspect (still probably not entirely
3030 stable yet) provides some support for smart Lisp development
3031 environments like SLIME.
3032 * The conditions signalled for errors occurring when loading .fasl
3033 files have been systematized (inheriting from SB-EXT:INVALID-FASL)
3034 in a way which should help ASDF recover gracefully.
3035 * The REQUIRE/PROVIDE behavior of *MODULE-PROVIDER-FUNCTIONS*
3036 stuff has been cleaned up. If you code contrib/ stuff, this might
3037 affect you, and you can look at contrib/README, contrib/STANDARDS,
3038 and/or the 0.8.4.27 diff to check.
3039 * In full calls the compiler now does not generate checks for declared
3040 argument types for all arguments.
3041 * various threading fixes
3042 ** and some experimental patches which didn't make it into
3043 the main tree for this release, but which are shipped in
3044 contrib/experimental-thread.patch as a possible fix for some
3045 failures (deadlock, spinning...) in GC-intensive multithreaded
3047 * fixed PPC build problem (source code incompatibility of different
3048 library versions): added offsetof() hackery which attempts to divine
3049 where glibc maintainers put uc_mcontext today
3050 * fixed bug 282: compiler does not trust type assertions while passing
3051 arguments to a full call.
3052 * fixed bug 261: compiler allows NIL or "no value" to be accepted for
3053 &OPTIONAL VALUES type parameter.
3054 * fix bug 214: algorithm for noting rejected templates is now more
3055 similar to that of template seletion. (also reported by rydis on
3057 * fixed bug 141b: printing backquoted information readably and prettily
3058 inserts a space where necessary.
3059 * bug fix: obviously wrong type specifiers such as (FIXNUM 1) or
3060 (CHARACTER 10) are now reported as errors, rather than propagated
3061 as unknown types. (reported by piso on #lisp)
3062 * bug fix: the :IF-EXISTS argument to OPEN now behaves correctly
3063 with values NIL and :ERROR. (thanks to Milan Zamazal)
3064 * fixed bug 191c: CLOS now does proper keyword argument checking as
3065 described in CLHS 7.6.5 and 7.6.5.1.
3066 * bug fix: LOOP forms using NIL as a for-as-arithmetic counter no
3067 longer raise an error; further, using a list as a for-as-arithmetic
3068 counter now raises a meaningful error.
3069 * fixed bug 213a: even fairly unreasonable CONS type specifiers are
3070 now understood by sequence creation functions such as MAKE-SEQUENCE
3072 * fixed bug 46k: READ-BYTE now signals an error when asked to read from
3073 a STRING-INPUT-STREAM.
3074 * compiler enhancement: SIGNUM is now better able to derive the type
3076 * type declarations inside WITH-SLOTS are checked. (reported by
3078 * fixed some bugs revealed by Paul Dietz' test suite:
3079 ** incorrect optimization of TRUNCATE for a positive first
3080 argument and negative second.
3081 ** compiler failure in let-convertion during flushing dead code.
3082 ** compiler failure while deriving type of TRUNCATE on an
3083 interval, containing 0.
3084 ** ASH of a negative bignum by a negative bignum count now returns
3086 ** intersection of CONS types now canonicalizes properly, fixing
3087 inconsistencies in SUBTYPEP.
3089 changes in sbcl-0.8.4 relative to sbcl-0.8.3:
3090 * incompatible change: The --disable-debugger command line
3091 option now clobbers the debugger at a more fundamental
3092 level, by redefining #'INVOKE-DEBUGGER instead of by
3093 rebinding *DEBUGGER-HOOK*. The main difference is that BREAK
3094 is specified by ANSI to ignore *DEBUGGER-HOOK* and
3095 INVOKE-DEBUGGER regardless. Under the old system, BREAK would
3096 enter the debugger REPL and then suffer recursive errors
3097 because *DEBUG-IO* is also messed up in --disable-debugger mode;
3098 while under the new system, BREAK in --disable-debugger mode
3099 terminates the system just as an unhandled error would.
3100 * fixed compiler performance when processing loops with a step >1;
3101 * bug fix: DOCUMENTATION now retrieves generic function
3102 documentation. Also, DOCUMENTATION and (SETF DOCUMENTATION)
3103 support has been systematized, and now supports the methods
3104 specified by ANSI, along with a default method and a method for
3105 slot documentation. (reported by Nathan Froyd)
3106 * bug fix: effective methods associated with a generic function are
3107 no longer cached over a change of that generic function's method
3108 combination. (reported by Andreas Fuchs)
3109 * bug fix: RUN-PROGRAM now does not fail if some element in $PATH
3110 names a non-existent directory. (thanks to Andreas Fuchs)
3111 * bug fix: ROUND and TRUNCATE could, under certain circumstances on
3112 the PPC platform, lead to stack corruption; this has been fixed.
3113 (reported by Rainer Joswig)
3114 * bug fix: ASH on an (UNSIGNED-BYTE 32) with a shift of -32 or lower
3115 no longer ever returns 1 instead of 0. (thanks to Lars Brinkhoff)
3116 * fixed bug 285: TRUNCATE on bignum arguments, and indeed bignum
3117 arithmetic in general, is now much more reliable on the PPC
3119 * bug fix: LOGCOUNT on (UNSIGNED-BYTE 64) objects on the Alpha
3120 platform now returns the right answer.
3121 * optimization: restored some effective method precomputation in
3122 CLOS (turned off by an ANSI fix in sbcl-0.8.3); the amount of
3123 precomputation is now tunable.
3124 * optimization: compiler-internal data structure use has been
3125 reviewed, and changes have been made that should improve the
3126 performance of the compiler by about 20%.
3127 * optimization: performance of FILL (and :INITIAL-ELEMENT) on
3128 simple-base-strings and simple-bit-vectors is improved.
3129 * optimization: the optimization of 32-bit logical and arithmetic
3130 functions introduced in version 0.8.3 on the x86 has been
3131 implemented on the mips, ppc and sparc platforms; an
3132 implementation of the same facility, but for 64-bit arithmetic,
3133 has been added for the alpha.
3134 * microoptimization: the compiler is better able to make use of the
3135 x86 LEA instruction for multiplication by constants.
3136 * bug fix: in some situations compiler did not report usage of
3137 generic arithmetic in (SPEED 3) policy.
3138 * bug 145b fix: compiler used wrong type specifier while converting
3139 MEMBER-types to numeric.
3140 * bug fix: COMPILE-FILE must bind *READTABLE*. (reported by Doug
3142 * bug fix: (SETF AREF) on byte-sized-element arrays with constant
3143 index argument now works properly on the Alpha platform.
3144 * bug fix: floating point exception treatment on the Alpha platform
3146 * bug fix: FILE-POSITION works much better on string input and
3147 output streams. (thanks to Nikodemus Siivola)
3148 * bug fix: many threading/garbage collection symptoms sorted.
3149 SB-THREAD:INTERRUPT-THREAD now safe to call on a thread that
3150 might be pseudo-atomic.
3151 * internal change: Stopping for GC is now done with signals not
3152 ptrace. GC is now done in whichever thread wanted it, instead of
3154 * bug fix: GC hooks (missing since 0.8) reinstated, so finalizers
3156 * bug fix: result form in DO is not contained in the implicit
3158 * incompatible change: ICR structure is changed; the value part of
3159 CONTINUATION is now called LVAR; corresponding functions are
3160 renamed (e.g. SB-C::CONTINUATION-TYPE has become SB-C::LVAR-TYPE).
3161 * added type deriver for ISQRT (thanks to Robert E. Brown).
3162 * bug fix: better support for loading from the command line when an
3163 initialization file sets (READTABLE-CASE *READTABLE*). (thanks
3165 * fixed some bugs revealed by Paul Dietz' test suite:
3166 ** the RETURN clause in LOOP is now equivalent to DO (RETURN ...).
3167 ** ROUND and FROUND now give the right answer when given very
3168 small float arguments.
3169 ** (FLOAT X) for X of type DOUBLE-FLOAT now returns X in all
3171 ** optimizer for (EXPT X 0) did not work for X not of type FLOAT.
3172 ** (GCD 0 <negative-integer>) returned <negative-integer>.
3173 ** LCM should return a non-negative integer.
3174 ** PARSE-INTEGER returned the index of a terminator instead of the
3175 upper bounding index of a substring in case :JUNK-ALLOWED NIL.
3176 ** PARSE-INTEGER returned an incorrect index being applied to a
3178 ** LCM with two arguments of 0 returns 0 rather than signalling
3180 ** unsigned addition of a 32-bit constant with the high bit set no
3181 longer causes an internal compiler error on the x86.
3182 ** LOGBITP accepts a non-negative bignum as its INDEX argument.
3183 ** compiler incorrectly derived types of DPB and DEPOSIT-FIELD
3184 with negative last argument.
3185 ** byte specifiers with zero size and position no longer cause
3186 an error during type derivation.
3187 ** bignum multiplication on the Alpha platform now returns the
3189 * porting: The system now builds on SuSE AMD64, although it still
3190 generates a 32-bit binary.
3191 * .fasl file incompatibility: The fasl file version number has
3192 been incremented (because of the changes to internal compiler
3193 data structures referred to above).
3195 changes in sbcl-0.8.3 relative to sbcl-0.8.2:
3196 * SBCL now builds and runs on MacOS X (version 10.2), or perhaps
3197 more accurately, on the Darwin kernel running on PowerPC hardware.
3198 (thanks to Brian Mastenbrook, Pierre Mai and Patrik Nordebo)
3199 * Compiler code deletion notes now signal a condition of type
3200 SB-EXT:CODE-DELETION-NOTE (a subtype of SB-EXT:COMPILER-NOTE) with
3201 an associated MUFFLE-WARNING restart.
3202 * The compiler now performs limited argument count validation of
3203 constant format strings in FORMAT, and where appropriate in ERROR,
3204 CERROR and WARN. (thanks to Gerd Moellmann)
3205 * New ASDF-INSTALL contrib can be used for automatic download and
3206 installation of third-party Lisp code from CCLAN or other sites
3208 * Threaded builds (:SB-THREAD) now support SB-THREAD:INTERRUPT-THREAD,
3209 which forces another thread to execute a function supplied by the
3211 * bug 75 fix: WITH-OUTPUT-TO-STRING (and MAKE-STRING-OUTPUT-STREAM)
3212 now accept and act upon their :ELEMENT-TYPE keyword argument.
3213 (reported by Martin Atzmueller, Edi Weitz)
3214 * bug fix: FILE-POSITION now accepts position designators up to
3215 ARRAY-DIMENSION-LIMIT or the extreme of the off_t range, whichever
3216 is the greater. (thanks to Patrik Nordebo)
3217 * bug fix: MAKE-ARRAY ignored :INITIAL-CONTENTS NIL. (reported by
3218 Kalle Olavi Niemitalo)
3219 * bug fix: the CLASS-PROTOTYPE of the GENERIC-FUNCTION class is now
3220 printable. (reported by Eric Marsden)
3221 * bug fix in sb-posix: mmap() now works on systems with a 64-bit
3222 off_t, including Darwin and FreeBSD. (thanks to Andreas Fuchs)
3223 * x86 bug fix in control stack exhaustion checking: now shows backtrace
3224 * bug fix in WITH-TIMEOUT: now the body can have more than one form.
3225 (thanks to Stig Sandoe)
3226 * bug fix in READ-SEQUENCE: READ-SEQUENCE following PEEK-CHAR or
3227 UNREAD-CHAR now correctly includes the unread character in the
3228 target sequence. (thanks to Gerd Moellmann)
3229 * bug fix in threaded builds: the system can now be suspended and
3230 resumed by shell job control with minimal disruption.
3231 * bug fixes in times and timezones >2038AD
3232 * better handling of "where is GNU make?" problem in build scripts
3233 (thanks to Nikodemus Siivola)
3234 * new optimization: inside a named function any reference to a
3235 function with the same name is considered to be a self-reference;
3236 this behaviour is controlled with SB-C::RECOGNIZE-SELF-CALLS
3237 optimization quality.
3238 * new optimization on x86: logical functions and + now have
3239 optimized (UNSIGNED-BYTE 32) versions, which are automatically
3240 used when the result is truncated to 32 bits.
3241 * VALUES declaration is partially enabled.
3242 * fixes in SB-GROVEL (thanks to Andreas Fuchs)
3243 * fixed some bugs revealed by Paul Dietz' test suite:
3244 ** The system now obeys the constraint imposed by
3245 UPGRADED-ARRAY-ELEMENT-TYPE that the upgraded array element
3246 types form a lattice under type intersection.
3247 ** FFLOOR, FTRUNCATE, FCEILING and FROUND work with integers.
3248 ** ASSOC now ignores NIL elements in an alist.
3249 ** CEILING now gives the right answer with MOST-NEGATIVE-FIXNUM
3250 and (1+ MOST-POSITIVE-FIXNUM) answers.
3251 ** The addition of a method with invalid qualifiers to a generic
3252 function does not cause an error to be signalled immediately;
3253 a warning is signalled, and the error is generated only on
3254 calling the generic function.
3255 * changed .fasl file version number, in part to add type codes for
3256 new array subtypes UNSIGNED-BYTE 7, 15, 29, and 31 mandated by
3257 obscure ANSI requirements
3259 changes in sbcl-0.8.2 relative to sbcl-0.8.1:
3260 * fixed bug 148: failure to inline-expand a local function left
3261 garbage, confusing the compiler.
3262 * fixed bugs 3cd: structure slot readers perform type check if the
3263 slot can have an invalid value (i.e. it is either not initialized
3264 or can be written with a less specific slot writer).
3265 * bug fix: the compiler now traps array references to elements off
3266 the end of an array; previously, the bounds checking in some
3267 circumstances could go off-by-one.
3268 * improved MACHINE-VERSION, especially on Linux (thanks to Lars
3270 * type declarations for array element types now obey the description
3271 on the CLHS page "Declaration TYPE", as per discussions on
3272 sbcl-help around 2003-05-08. This means that a declaration
3273 (TYPE (ARRAY FOO) BAR) means that, within the scope of the
3274 declaration, all references to BAR will be asserted or assumed
3275 (with THE, so dependent on compiler policy) to involve objects of
3276 type FOO. Note that no such declaration is implied in
3277 (MAKE-ARRAY .. :ELEMENT-TYPE 'FOO).
3278 * declared types of functions from the "Conditions"
3279 chapter. (reported by Paul Dietz)
3280 * bug fix: CERROR accepts a function as its first argument.
3281 * bug fix: NTH an NTHCDR accept a bignum as index
3282 arguments. (reported by Adam Warner)
3283 * optimization: character compare routines now optimize comparing
3284 against a constant character. (reported by Gilbert Baumann)
3285 * bug fix: (SETF AREF) on byte-sized-element arrays with constant index
3286 argument now works properly on the MIPS platform.
3287 * fixed compiler failure on (TYPEP x '(NOT (MEMBER 0d0))).
3288 * repeated evaluation of the same DEFSTRUCT, a slot of which is
3289 declared to have a functional type, does not cause an error
3291 * fixed bug: sometimes MAKE-INSTANCE did not work with classes with
3292 many :DEFAULT-INITARGS. (reported by Istvan Marko)
3293 * fixed bug: if last continuation of a deleted block has a
3294 destination, this destination should be deleted too. (reported by
3296 * fixed a bug in the bootstrap process: the host compiler's values
3297 of ARRAY-DIMENSION-LIMIT and ARRAY-TOTAL-SIZE-LIMIT no longer leak
3298 into the newly-built SBCL. (reported by Eric Marsden on #lisp,
3299 test case from Patrik Nordebo)
3300 * improved the ability of the disassembler on the PPC platform to
3301 provide helpful disassembly notes.
3302 * SB-MOP:CLASS-PROTOTYPE on built-in-classes returns an instance of
3303 the class in more cases than previously.
3304 * bug fix: FILE-POSITION now understands :START and :END for
3305 STRING-INPUT-STREAMs. (thanks to Nikodemus Siivola)
3306 * bug fix: (SIGNED-BYTE 8) streams no longer return (UNSIGNED-BYTE
3307 8) data. (thanks to David Lichteblau)
3308 * bug fix: it is possible to add a method to a generic function
3309 without lambda list.
3310 * bug fix: reader failed to signal END-OF-FILE inside an
3311 object representation. (reported by Nikodemus Siivola)
3312 * fixed some bugs revealed by Paul Dietz' test suite:
3313 ** LAST and [N]BUTLAST should accept a bignum.
3314 ** condition slot accessors are methods.
3315 ** (VECTOR NIL) is a subtype of STRING.
3317 changes in sbcl-0.8.1 relative to sbcl-0.8.0:
3318 * minor incompatible change: some nonsensical specialized lambda
3319 lists (used in DEFMETHOD) which were previously ignored now signal
3321 * minor incompatible change: the system is now aware of the types of
3322 variables in the COMMON-LISP package, and will signal errors for
3323 most violations of these type constraints (where previously they
3324 were silently accepted).
3325 * minor incompatible change: COMPILE-FILE now uses the freedom
3326 afforded (ANSI 3.2.2.3) to use derived function types for
3327 functions defined in the same file. This also permits the system
3328 to warn on static type mismatches and function
3329 redefinition. (Currently it does not work with high DEBUG level.)
3330 * minor incompatible change: VALUES declaration is disabled.
3331 * When issuing notes, the compiler now signals a condition of type
3332 SB-EXT:COMPILER-NOTE, and provides an associated MUFFLE-WARNING
3333 restart for use in user handlers. It is expected that the
3334 COMPILER-NOTE condition will eventually become a condition
3335 supertype to a hierarchy of note types, which will then be
3336 handleable in a similar fashion. However, at the moment, no such
3337 note subtypes yet exist. (SB-INT:SIMPLE-COMPILER-NOTE exists,
3338 but it's an implementation detail, not a classification for the
3340 * Changes in type checking closed the following bugs:
3341 ** type checking of unused values (192b, 194d, 203);
3342 ** template selection based on unsafe type assertions (192c, 236);
3343 ** type checking in branches (194bc).
3344 * A short form of VALUES type specifier has ANSI meaning (it has
3345 increased the number of situations when SBCL cannot perform type
3347 * fixed bug in DEFSTRUCT: once again, naming structure slots with
3348 keywords or constants is permissible.
3349 * STREAM-READ-SEQUENCE and STREAM-WRITE-SEQUENCE now have methods
3350 defined on the relevant FUNDAMENTAL-BINARY-{INPUT,OUTPUT}-STREAM
3351 classes. (thanks to Antonio Martinez)
3352 * improved ANSIness in DESCRIBE: The DESCRIBE function no longer
3353 outputs FRESH-LINE or TERPRI, and no longer converts its stream
3354 argument to a pretty-print stream. Instead, it leaves any such
3355 operations to DESCRIBE-OBJECT methods.
3356 * bug fix: APROPOS now respects the EXTERNAL-ONLY flag. (reported
3358 * bug fix: NIL is now a valid destructuring argument in DEFMACRO
3359 lambda lists. (thanks to David Lichteblau)
3360 * bug fix: Defining a generic function with a :METHOD-CLASS being a
3361 subclass of STANDARD-METHOD no longer causes stack exhaustion.
3362 (thanks to Gerd Moellmann)
3363 * fixed bug 246: increased compilation speed of long
3364 MULTIPLE-VALUE-BIND (and likewise of NTH-VALUE with a constant
3366 * a contributed module implementing COMPILER-LET and MACROEXPAND-ALL
3368 * DEFCONSTANT now throws a condition of type
3369 SB-EXT:DEFCONSTANT-UNEQL if it is being asked to redefine a
3370 constant to a non-EQL value; CONTINUE and ABORT restarts
3371 respectively change and preserve the value.
3372 * fixed bug 63: The code walker, part of the implementation of CLOS,
3373 is now better at handling symbol macros.
3374 * bug fix: There is no longer an internal implementation type named
3375 CL:LENGTH. (reported by Raymond Toy)
3376 * bug fix: In macro-like defining macros/special operators the
3377 implicit block does not enclose the lambda list.
3378 * fixed bugs 10 and 43: Bare VALUES, AND, OR and MEMBER symbols (not
3379 enclosed in parentheses) are not suitable as type specifiers, and
3380 their use properly signals an error now.
3381 * bug fix: An argument count mismatch for a type specifier in code
3382 being compiled no longer causes an unhandled error at compile
3383 time, but signals a compile-time warning.
3384 * fixed simple vector readable printing
3385 * bug fix: DESCRIBE takes more care over whether the class
3386 precedence list slot of a class is bound before accessing it.
3387 (reported by Markus Krummenacker)
3388 * bug fix: FORMATTER can successfully compile pretty-printer format
3389 strings which use variants of the ~* directive inside.
3390 * bug fix: SEARCH now applies its TEST predicate to the elements of
3391 the arguments in the correct order. (thanks to Wolfhard Buss)
3392 * fixed bug 235b: The compiler uses return types of MAPCAR and friends
3393 in type inference. (thanks to Robert E. Brown)
3394 * bug fix: Reading in symbols with an explicit package name of ""
3395 (e.g. '||::FOO) now works correctly. (reported by Henrik Motakef)
3396 * fixed some bugs revealed by Paul Dietz' test suite:
3397 ** NIL is now allowed as a structure slot name.
3398 ** Arbitrary numbers, not just REALs, are allowed in certain
3399 circumstances in LOOP for-as-arithmetic clauses.
3400 ** Multiple class redefinitions before slot access no longer
3401 causes a type error.
3402 ** (SETF FIND-CLASS) now accepts NIL as an argument to remove the
3403 association between the name and a class.
3404 ** Generic functions with non-standard method-combination and over
3405 five methods all of which return constants no longer return NIL
3406 after the first few invocations. (thanks to Gerd Moellmann)
3407 ** CALL-NEXT-METHOD with no arguments now passes the original
3408 values of the arguments, even in the presence of assignment.
3409 ** Functions [N]SUBST*, LAST, NRECONC, [N]SUBLIS may return any
3411 ** DISASSEMBLE works with closures and funcallable instances.
3412 ** ADD-METHOD now returns the generic function, not the new method.
3413 ** FIND-METHOD signals an error if the lengths of the specializers
3414 is incompatible with the generic function, even if the ERRORP
3416 ** TYPE-OF returns recognizeable subtypes of all built-in-types of
3417 which its argument is a member.
3418 ** DEFCLASS only redefines the class named by its class-name
3419 argument if that name is the proper name of the class;
3420 otherwise, it creates a new class.
3421 ** SLOT-UNBOUND now correctly initalizes the CELL-ERROR-NAME slot
3422 of the UNBOUND-SLOT condition to the name of the slot.
3423 ** (SETF (AREF bv 0) ...) did not work for bit vectors.
3424 ** SLOT-UNBOUND and SLOT-MISSING now have their return values
3425 treated by SLOT-BOUNDP, SLOT-VALUE, (SETF SLOT-VALUE) and
3426 SLOT-MAKUNBOUND in the specified fashion.
3428 changes in sbcl-0.8.0 relative to sbcl-0.8alpha.0
3429 * SBCL now builds using CLISP (version of late April 2003 from CVS) as
3430 cross-compilation host. As a consequence, we can now bootstrap our
3431 way up to SBCL starting with a bare gcc toolchain and human-readable
3432 source code (first the source to CLISP, then the source to SBCL).
3433 * A contributed module containing a partial implementation of the
3434 simple-streams interface has been included. (thanks to Rudi
3436 * A contributed module implementing the RFC1321 Message Digest
3437 Algorithm, known as MD5, has been included.
3438 * minor incompatible change: The :NEGATIVE-ZERO-IS-NOT-ZERO feature
3439 no longer has any effect, as the code controlled by this feature
3440 has been deleted. (As far as we know, no-one has ever built using
3441 this feature, and its semantics were confused in any case).
3442 * minor incompatible change: As a consequence of making SLOT-EXISTS-P
3443 work on conditions (as required by the ANSI specification),
3444 SLOT-VALUE, (SETF SLOT-VALUE) and SLOT-BOUNDP likewise have the
3445 expected behaviour on conditions. Users should note, however,
3446 that such behaviour is not required by the ANSI specification,
3447 and so use of this behaviour may render their code unportable.
3448 * fixed some bugs revealed by Paul Dietz' test suite:
3449 ** the GENERIC-FUNCTION type is no longer disjoint from FUNCTION
3451 ** &ENVIRONMENT parameter in macro lambda list is bound first.
3452 ** SXHASH on condition objects no longer returns NIL.
3453 ** :ALLOCATION :CLASS slots are better treated; their values are
3454 updated on class redefinition, and initforms inherited from
3455 superclasses are applied.
3456 ** REMOVE-METHOD returns its generic function argument even when
3457 no method was removed.
3458 ** SHARED-INITIALIZE now initializes the values of the requested
3459 slots, including those with :ALLOCATION :CLASS.
3460 ** ALLOCATE-INSTANCE now works on structure classes defined via
3461 DEFSTRUCT (and not just by those from DEFCLASS :METACLASS
3463 ** SLOT-EXISTS-P now works on conditions, as well as structures
3465 ** MAKE-LOAD-FORM now has the required methods on
3466 STRUCTURE-OBJECT, CONDITION and STANDARD-OBJECT.
3467 ** MAKE-LOAD-FORM-SAVING-SLOTS no longer returns a special
3468 keyword, and now implements the SLOT-NAMES argument.
3469 ** methods with &OPTIONAL arguments no longer allow too many
3470 arguments to be passed in the call without error.
3471 ** DEFGENERIC now checks that the :ARGUMENT-PRECEDENCE-ORDER
3472 option is consistent with the required arguments of the generic
3473 function lambda list.
3474 * bug fix: REQUIRE accepts a string designator. (Thanks to
3476 * bug fix: SB-MOP:DIRECT-SLOT-DEFINITION-CLASS and
3477 SB-MOP:EFFECTIVE-SLOT-DEFINITION-CLASS now have the
3478 specified-by-AMOP lambda list of (CLASS &REST INITARGS).
3479 * bug fix: The compiler now checks for duplicated variables in macro
3481 * bug fix: SETQ on globals returns the correct value.
3482 * fixed bug 47.d: (DEFGENERIC IF (X)) now signals a PROGRAM-ERROR,
3483 not a COMPILER-ERROR (followed by some other strange error on
3484 choosing the CONTINUE restart).
3485 * bug fix: make.sh and friends are now more consistent in the way that
3486 they look for GNU "make".
3488 changes in sbcl-0.8alpha.0 relative to sbcl-0.7.14
3489 * experimental native threads support (on x86 Linux >=2.4 only).
3490 This is not compiled in by default: you need to add :SB-THREAD to
3491 the target features. See the "Beyond ANSI" chapter of the manual
3493 * fix for longstanding nonANSIism: The old distinction between
3494 CL:CLASS objects and SB-PCL:CLASS objects has been eliminated.
3495 The return value from CL:FIND-CLASS is now a CLOS class, and
3496 likewise that of CL:CLASS-OF; CL:BUILT-IN-CLASS,
3497 CL:STRUCTURE-CLASS and CL:STANDARD-CLASS name CLOS classes.
3498 * An interface to the MetaObject Protocol, as described in Kiczales,
3499 des Rivieres and Bobrow's "The Art of the Metaobject Protocol",
3500 MIT Press, 1991, is available from the SB-MOP package.
3501 * incompatible change: the SB-PCL package should now be considered
3502 a private implementation detail, and no longer a semi-private MOP
3504 * minor incompatible change: due to rearrangement for threads, the
3505 control stack and binding stack are now allocated at arbitrary
3506 addresses instead of being hardcoded per-port. Users affected by
3507 this probably have to be doing advanced things with shared
3508 libraries, and will know who they are.
3509 * minor incompatible change: Previously, all --eval forms used were
3510 processed with READ before any of them were processed with EVAL.
3511 Now each --eval form is processed with both READ and EVAL before
3512 the next --eval form is processed. (Thus package operations like
3513 sbcl --eval "(defpackage :foo)" --eval "(print 'foo::bar)" now
3514 work as the user might reasonably expect.)
3515 * minor incompatible change: *STANDARD-INPUT* is now only an
3516 INPUT-STREAM, not a BIDIRECTIONAL-STREAM. (thanks to Antonio
3518 * minor incompatible change: Y-OR-N-P is now character-oriented,
3519 not line oriented. Also, YES-OR-NO-P now works without errors.
3520 (thanks to Antonio Martinez)
3521 * sb-aclrepl module improvements: an integrated inspector, added
3522 repl features, and a bug fix to :trace command.
3523 * Known functions, which cannot be open coded by the backend, are
3524 considered to be able to check types of their arguments. (fixing
3525 a bug report by Nathan J. Froyd)
3526 * fixed a bug in computing method discriminating functions: It is
3527 now possible to define methods specialized on classes which have
3528 forward-referenced superclasses. (thanks to Gerd Moellmann)
3529 * fixed evaluation order in optional entries (reported by Gilbert
3531 * SB-MOP:ENSURE-CLASS-USING-CLASS now takes its arguments in the
3532 specified-by-AMOP order of (CLASS NAME &REST ARGS &KEY).
3533 * SB-MOP:COMPUTE-EFFECTIVE-SLOT-DEFINITION now takes the
3534 required-by-AMOP NAME argument, as well as CLASS and
3535 DIRECT-SLOT-DEFINITIONS. (thanks to Kevin Rosenberg)
3536 * fixed bug 20: DEFMETHOD can define methods using names that are
3537 not the proper names of classes to designate class specializers.
3538 * bug fix: INTERACTIVE-STREAM-P now works on streams associated with
3539 Unix file descriptors, instead of blowing up. (thanks to Antonio
3541 * Garbage collection refactoring: user-visible change is that a
3542 call to the GC function during WITHOUT-GCING will not do garbage
3543 collection until the end of the WITHOUT-GCING. If you were doing
3544 this you were probably losing anyway.
3545 * fixed bug in MEMBER type: (MEMBER 0.0) is not the same as
3546 (SINGLE-FLOAT 0.0 0.0), because of the existence of -0.0 which is
3547 TYPEP the latter but not the former.
3548 * The compiler issues a full WARNING for calls to undefined functions
3549 with names from the CL package.
3550 * MAP-INTO for a vector destination is open coded. (reported by
3551 Brian Downing on c.l.l)
3552 * bug fix: the long form of DEFINE-METHOD-COMBINATION now accepts a
3553 documentation string.
3554 * fixed some bugs revealed by Paul Dietz' test suite:
3555 ** COPY-ALIST now signals an error if its argument is a dotted
3557 ** Condition slots are now accessed more correctly in the presence
3558 of multiple initargs for a given slot.
3559 ** The USE-VALUE, CONTINUE and STORE-VALUE functions now correctly
3560 exclude restarts of the same name associated with a different
3562 ** DEFCLASS of forward-referenced classes with another
3563 forward-referenced class in the superclasses list no longer
3565 ** Condition slots are now initialized once each, not multiple
3566 times. (thanks to Gerd Moellmann)
3567 ** CONVERT-MORE-CALL failed on a lambda list (&KEY). (thanks to
3569 ** &WHOLE and &REST arguments in macro lambda lists are patterns.
3570 ** NSET-EXCLUSIVE-OR does not return extra elements when its
3571 arguments contain duplicated elements.
3572 ** RESTART-CASE understands local macros.
3573 ** RESTART-CASE associates exactly its own restarts with a condition.
3574 ** ENDP in safe mode checks its argument to be of type LIST.
3575 ** COPY-SYMBOL in a threaded build no longer fails when the symbol
3576 in question is unbound.
3577 ** Optimized MAKE-INSTANCE functions no longer cause internal
3578 assertion failures in the presence of duplicate initargs.
3579 ** SLOT-MAKUNBOUND returns the instance acted upon, not NIL.
3580 ** Side-effectful :DEFAULT-INITARGS have their side-effects
3581 propagated even in the ctor optimized implementation of
3583 ** :ALLOW-OTHER-KEYS NIL is now accepted in an initarg list.
3585 changes in sbcl-0.7.14 relative to sbcl-0.7.13:
3586 * a better implementation of SXHASH on (simple) bit vectors,
3587 measured both in execution speed and in distribution of results
3588 over the positive fixnums, has been installed. Likewise, a better
3589 implementation of EQUAL for simple bit vectors is now available.
3590 * fixed CEILING optimization for a divisor of form 2^k.
3591 * fixed bug 240 (emitting extra style warnings "using the lexical
3592 binding of the symbol *XXX*" for &OPTIONAL arguments). (reported
3593 by Antonio Martinez)
3594 * fixed SXHASH, giving different results for NIL depending on type
3595 declarations (SYMBOL or LIST). (thanks to Gerd Moellmann)
3596 * fixed bug in DEFPARAMETER and DEFVAR: they could assign a lexical
3597 variable. (found by Rolf Wester)
3598 * SBCL does not ignore type declarations for special
3599 variables. (reported by rif on c.l.l 2003-03-05)
3600 * some bug fixes in contrib/sb-aclrepl/
3601 * fixed some bugs revealed by Paul Dietz' test suite:
3602 ** a bug in the CONS type specifier, whereby the CAR and CDR
3603 types got intertwined, has been fixed;
3604 ** the type system is now able to reason about the interaction
3605 between INTEGER and RATIO types more completely;
3606 ** APPEND, [N]REVERSE and NRECONC check that those their
3607 arguments, which must be proper lists, are really so;
3608 ** An array specialized to be unable to hold elements has been
3609 implemented, as required -- yes, really -- by ANSI;
3610 ** GETF and GET-PROPERTIES throw a TYPE-ERROR, not a SIMPLE-ERROR,
3611 on malformed property lists;
3613 changes in sbcl-0.7.13 relative to sbcl-0.7.12:
3614 * incompatible packaging change: in line with Unix convention,
3615 SBCL now looks for its core file in /usr/{local/,}lib/sbcl/sbcl.core
3616 if it's not in $SBCL_HOME. It also sets SBCL_HOME to match.
3617 * REQUIRE and PROVIDE are now optionally capable of doing something
3618 useful. See the documentation string for REQUIRE.
3619 * infrastructure for a managed SBCL contrib system: contributed
3620 modules in this release include:
3621 ** the ASDF system definition facility;
3622 ** an interface to the BSD Sockets API;
3623 ** an ACL-like convenience interface to the repl;
3624 (thanks to Kevin Rosenberg)
3625 ** an implementation of ROTATE-BYTE, with efficient implementation
3627 * fixed a bug in LOG, so that LOG of a rational argument near 1 now
3628 gives a closer approximation to the right answer than previously.
3629 (thanks to Raymond Toy)
3630 * fixed bug 157: TYPEP, SUBTYPEP, UPGRADED-ARRAY-ELEMENT-TYPE and
3631 UPGRADED-COMPLEX-PART-TYPE now take (ignored, in all situations)
3632 optional environment arguments, as required by ANSI.
3633 * fixed bugs in other functions taking environment objects, allowing
3634 calls with an explicit NIL environment argument to be compiled
3636 * fixed bug 228: primary return values from
3637 FUNCTION-LAMBDA-EXPRESSION are either NIL or suitable for input to
3638 COMPILE or FUNCTION.
3639 * fixed a bug in DEFSTRUCT: predicates for :NAMED structures with
3640 :TYPE will no longer signal errors on innocuous objects.
3641 * fixed bug 231b: SETQ is better at respecting type declarations in
3642 the lexical environment.
3643 * fixed a bug in DEFCLASS: classes named by symbols with no or
3644 unprintable packages can now be defined.
3645 * fixed a bug in RESTART-BIND: The :TEST-FUNCTION option had been
3646 carelessly renamed to :TEST-FUN. (thanks to Robert E. Brown)
3647 * fixed compiler failure related to checking types of functions.
3648 (reported by Robert E. Brown)
3649 * the compiler is now much more consistent in its error-checking
3650 treatment of bounding index arguments to sequence functions: in
3651 (SAFETY 3) code, errors will be signalled in almost all cases if
3652 invalid sequence bounding indices are passed to functions defined
3653 by ANSI to operate on sequences.
3654 * fixed a bug in the build procedure: documentation of SBCL-specific
3655 packages is now preserved and available in the final Lisp image.
3656 * lifted FDEFINITION lookup out of loops in the implementation of
3657 many list operations. (thanks to Robert E. Brown)
3658 * fixed a bug in the reader: the #n# reader macro now works for
3659 objects of type STANDARD-OBJECT. (reported by Tony Martinez)
3660 * the compiler is now aware that SYMBOL-FUNCTION returns a FUNCTION
3661 and that READ-DELIMITED-LIST returns a LIST. (thanks to Robert
3662 E. Brown and Tony Martinez respectively)
3663 * PCL is now smarter about SLOT-VALUE, (SETF SLOT-VALUE) and
3664 SLOT-BOUNDP: in particular, it is now able to optimize them much
3665 better, and is now not vulnerable to having packages renamed.
3666 Furthermore, a compliance bug has been fixed: SLOT-MISSING is now
3667 always called when a slot is not present in an instance. (thanks
3669 * fixed a bug related to CONCATENATED-STREAMs: PEEK-CHAR will no
3670 longer signal an error on unreading a character following EOF on
3671 the previous constituent stream. (thanks to Tony Martinez)
3672 * fixed some bugs revealed by Paul Dietz' test suite:
3673 ** ARRAY-IN-BOUNDS-P now allows arbitrary integers as arguments,
3674 not just nonnegative fixnums;
3675 ** the logical bit-array operators such as BIT-AND now accept an
3676 explicit NIL for their "opt-arg" argument (to indicate a
3677 freshly-consed result bit-array);
3678 ** ELT now signals an error on an invalid sequence index in safe
3680 ** the type system is now cleverer about negations of numeric
3681 types, and consequently understands the BIGNUM and RATIO types
3683 ** the type system is now cleverer about the interaction between
3684 INTEGER and RATIO types: while bugs still remain, many more
3685 cases are accurately computed;
3686 ** in TYPECASE, OTHERWISE now only introduces an otherwise-clause
3687 if it is in the last clause;
3688 ** CONSTANTLY now correctly returns a side-effect-free function in
3690 ** DECLARE is no longer treated as a special-operator; in
3691 particular, SPECIAL-OPERATOR-P no longer returns T for DECLARE;
3692 * incremented fasl file version number due to the change in the
3693 DEFSTRUCT-SLOT-DESCRIPTION structure.
3695 changes in sbcl-0.7.12 relative to sbcl-0.7.11:
3696 * minor incompatible change: code processed by the "interpreter" or
3697 EVAL now has a compilation optimization policy of (DEBUG 2)
3698 (changed from (DEBUG 1)) to improve debuggability of interactive
3699 development, and to allow the use of the debug RETURN command in
3701 * an experimental implementation of the RETURN command for the
3702 debugger has been included. (thanks to Frederik Kuivinen)
3703 * fixed bug 62: constraints were not propagated into a loop.
3704 * fixed bug in embedded calls of SORT (reported and investigated by
3706 * fixed some bugs revealed by Paul Dietz' test suite:
3707 ** printing and reading of arrays with some dimensions having
3708 length 0 (thanks to Gerd Moellmann);
3709 ** BOA constructor with &AUX argument without a default value does
3710 not cause a type error;
3711 ** CONSTANTP now returns true for all self-evaluating objects.
3713 changes in sbcl-0.7.11 relative to sbcl-0.7.10:
3714 * fixed bug 127: DEFSTRUCT now does not clobber old structure
3715 accessors that are related by inheritance, as specified in the
3716 :CONC-NAME section of the specification of DEFSTRUCT. (thanks to
3717 Valtteri Vuorikoski)
3718 * The compiler is now able to inline functions that were defined in
3719 a complex lexical environment (e.g. inside a MACROLET).
3720 * fixed bug in DESCRIBE, which now works on rank-0 arrays. (thanks
3722 * Support for the upcoming FreeBSD-5.0 release has been included.
3723 (thanks to Dag-Erling Smorgrav)
3724 * fixed bug 219: DEFINE-COMPILER-MACRO no longer has compile-time
3725 effect when it is not in a toplevel context.
3726 * fixed bug 222: DEFMETHOD and SYMBOL-MACROLET interactions now
3727 stand a better chance of being correct. (thanks to Gerd
3729 * fixed bug in COERCE, which now signals an error on coercing a
3730 rational to a bounded real type which excludes the expected
3732 * The compiler is now able to derive types more accurately from the
3733 COERCE and COMPILE functions.
3734 * fixed bug 223: functional binding is considered to be constant
3735 only for symbols in the CL package.
3736 * fixed bug 231: SETQ did not check the type of a variable being set
3737 (reported by Robert E. Brown)
3738 * A new optimization for MAKE-INSTANCE has been included, fixing
3739 various bugs (including relating to :ALLOCATION :CLASS slots and
3740 :DEFAULT-INITARGS over-eager evalueation). (thanks to Gerd
3742 * fixed some LOOP bugs revealed by Paul Dietz' test suite:
3743 ** As required by ANSI, LOOP now disallows anonymous collection
3744 clauses such as COLLECT I in conjunction with aggregate boolean
3745 clauses such as THEREIS (= I 1);
3746 ** LOOP now signals an error when any variable is reused in the
3747 same loop (including the potentially useful construct analogous
3748 to WITH A = 1 WITH A = (1+ A);
3749 ** IT is only a special loop symbol within the first clause of a
3750 conditional loop clause;
3751 ** LOOP with a typed iteration variable over a hashtable now
3752 signals a type error iff it should.
3753 * fixed some other bugs revealed by Paul Dietz' test suite:
3754 ** FILE-STREAM now names the class previously known as FD-STREAM;
3755 ** in DEFSTRUCT, a bare :CONC-NAME (or a :CONC-NAME with no
3756 argument) no longer signals an error;
3757 ** likewise in DEFSTRUCT, :CONC-NAME NIL now respects the package
3758 of the slot symbol, rather than using the current package
3759 ((:CONC-NAME "") continues to intern the slot's name in the
3761 * incremented fasl file version number, because of the incompatible
3762 change to the DEFSTRUCT-DESCRIPTION structure, and again because
3763 of the new implementation of DEFINE-COMPILER-MACRO.
3765 changes in sbcl-0.7.10 relative to sbcl-0.7.9:
3766 * Support for building SBCL for MIPS platforms running in
3767 little-endian mode has now been checked in, and basic
3768 functionality on said platforms verified.
3769 * minor incompatible change: PCL now records the pathname of a file
3770 in which methods and the like are defined, rather than its
3772 * minor incompatible change: TRUENAME now considers the truename of
3773 a file naming a directory to be the pathname with :DIRECTORY
3774 component indicating that directory.
3775 * minor incompatible change: a NAMED clause in the extended form of
3776 LOOP no longer causes a BLOCK named NIL to surround the LOOP. The
3777 reason for the previous behaviour is unclear.
3778 * more systematization and improvement of CLOS and MOP conformance
3779 in PCL (thanks to Gerd Moellman and Pierre Mai):
3780 ** the standard ANSI CL generic function NO-NEXT-METHOD is now
3782 ** DEFINE-METHOD-COMBINATION no longer signals an error for
3783 primary methods with no specializers;
3784 ** the MOP generic function GENERIC-FUNCTION-DECLARATIONS is now
3786 ** the Readers for Class Metaobjects methods CLASS-DIRECT-SLOTS
3787 and CLASS-DIRECT-DEFAULT-INITARGS have been implemented for
3788 FORWARD-REFERENCED-CLASSes; error reporting on
3789 CLASS-DEFAULT-INITARGS, CLASS-PRECEDENCE-LIST and CLASS-SLOTS
3791 ** SXHASH on CLOS instances now uses a slot internal to the
3792 instance to return different numbers on distinct instances,
3793 while preserving the same return value through invocations of
3795 ** DEFMETHOD signals errors when methods with longer incongruent
3796 lambda lists are added to generic functions;
3797 ** COMPUTE-CLASS-PRECEDENCE-LIST now has a method specialized on
3798 CLASS, as specified in AMOP;
3799 ** COMPUTE-SLOTS :AROUND now assigns locations sequentially based
3800 on the order returned by the primary method for classes of
3801 class STANDARD-CLASS;
3802 ** DEFINE-METHOD-COMBINATION now works with the :ARGUMENTS option.
3803 * fixed some bugs shown by Paul Dietz' test suite:
3804 ** DOLIST puts its body in TAGBODY;
3805 ** SET-EXCLUSIVE-OR sends arguments to :TEST function in the
3807 ** MULTIPLE-VALUE-SETQ evaluates side-effectful places before
3808 value producing form;
3809 ** if more variables are given to PROGV than values, extra
3810 variables are bound and made to have no value;
3811 ** NSUBSTITUTE on list arguments gets the right answer with
3813 ** ELT signals an error of type TYPE-ERROR when the index argument
3814 is not a valid sequence index;
3815 ** LOOP signals (at macroexpansion time) an error of type
3816 PROGRAM-ERROR when duplicate variable names are found;
3817 ** LOOP supports DOWNTO and ABOVE properly; (thanks to Matthew Danish)
3818 ** FUNCALL of special-operators now cause an error of type
3820 ** PSETQ now works as required in the presence of side-effecting
3821 symbol-macro places;
3822 ** NCONC accepts any object as its last argument;
3823 ** :COUNT argument to sequence functions may be BIGNUM; (thanks to
3825 ** loop-for-as-package does not require a package to be explicitely
3827 ** LOOP WITH now treats NIL in the d-var-spec correctly as an
3829 * fixed bug 166: compiler preserves "there is a way to go"
3830 invariant when deleting code.
3831 * fixed bug 172: macro lambda lists with required arguments after
3832 &REST arguments now cause an error to be signalled. (thanks to
3834 * fixed Entomotomy PEEK-CHAR-WRONGLY-ECHOS-TO-ECHO-STREAM
3835 bug. (thanks to Matthew Danish)
3836 * fixed bug 225: STRING-STREAM is now a class. (reported by Gilbert
3838 * fixed bug 136: CALL-NEXT-METHOD no longer gets confused when
3839 arguments are lexically rebound. (thanks to Gerd Moellmann and
3841 * fixed bug 194: error messages are now more informative when there
3842 is no primary method applicable in a call to a generic
3843 function. (thanks to Gerd Moellmann)
3844 * fixed bug in command line argument checking (thanks to Julian
3846 * fixed bug in COUNT-IF, making it handle :FROM-END correctly
3847 (thanks to Matthew Danish)
3848 * incremented fasl file version number, because of the
3849 SXHASH-related changes in the layout of CLOS data structures
3851 changes in sbcl-0.7.9 relative to sbcl-0.7.8:
3852 * minor incompatible change: The runtime (the Unix executable named
3853 "sbcl") is now much pickier about the .core files it will load.
3854 Essentially it now requires .core files to descend from the same
3855 build (not just the same sources or LISP-IMPLEMENTATION-VERSION)
3856 as the runtime does. (The intent is to prevent the crashes which
3857 can occur, and which can even be reported as mysterious failures,
3858 when people patch the sources or change the build parameters
3859 without changing LISP-IMPLEMENTATION-VERSION, then mix and match
3860 sbcl and .core files.)
3861 * fixed bug: VALUES-LIST is no longer optimized away.
3862 * fixed bug 142: The FFI conversion of C string values to Lisp
3863 string values no longer conses excessively. (thanks to Nathan
3864 Froyd porting Raymond Toy's fix to CMU CL)
3865 * began to systematize and improve MOP conformance in PCL (thanks to
3866 Nathan Froyd, Gerd Moellman and Pierre Mai):
3867 ** SLOT-DEFINITION-ALLOCATION now returns :CLASS, not the class
3869 ** GENERIC-FUNCTION-ARGUMENT-PRECEDENCE-ORDER is now implemented;
3870 ** FINALIZE-INHERITANCE is now called on class finalization;
3871 ** DOCUMENTATION and (SETF DOCUMENTATION) now have the correct
3872 argument precedence order.
3873 * fixed bug 202: The compiler no longer fails on functions whose
3874 derived types contradict their declared type.
3875 * DEFMACRO is implemented via EVAL-WHEN instead of IR1 translation,
3876 so it can be non-toplevel.
3877 * The fasl file version number has changed (because of the new
3878 implementation of DEFMACRO).
3879 * (mostly) fixed bugs 46b and 46c: sequence functions now check, in
3880 safe code, that any length requirement by their type-specifier
3881 argument is valid. The exceptions to this are described in bug
3883 * fixed bugs 46h and 46i: TWO-WAY- and CONCATENATED-STREAM creation
3884 functions now check the types of their inputs as required by ANSI.
3885 * fixed bug 48c: SYMBOL-MACROLET signals PROGRAM-ERROR when an
3886 introduced symbol is DECLAREd to be SPECIAL.
3887 * fixed reading of (COMPLEX DOUBLE-FLOAT) literals from fasl files
3888 * fixed bug: :COUNT argument to sequence functions may be negative
3889 * fixed bug: body of DO-SYMBOLS may contain declarations
3890 * fixed bug: PUSHNEW now evaluates its arguments from left to right
3891 (reported by Paul F. Dietz, fixed by Gerd Moellman)
3892 * fixed bug: PUSH, PUSHNEW and POP now evaluate a place given by a
3893 symbol macro only once
3894 * fixed printing of call frame when argument list is unavailable
3895 * fixed bug: :ALLOW-OTHER-KEYS is an allowed keyword name
3896 * compiler no longer signals WARNING on unknown keyword
3899 changes in sbcl-0.7.8 relative to sbcl-0.7.7:
3900 * A beta-quality port to the mips architecture running Linux,
3901 based on the old CMUCL backend, has been made. It has been tested
3902 on a big-endian kernel, and works sufficiently well to be able to
3903 rebuild itself; it has not been tested in little-endian mode.
3904 * fixed an inconsistency between gencgc.c and purify.c which made
3905 dumping/loading .core files unreliable
3906 * fixed bug 120a: The compiler now deals correctly with IFs where
3907 the consequent is the same as the alternative, instead of
3908 misderiving the return type. (thanks to Alexey Dejneka)
3909 * fixed bug 113: Logical pathnames are now dumpable (the logical
3910 host is resolved at load-time, throwing an error if it is not
3912 * fixed bug 174: FORMAT's error message is slightly clearer when a
3913 non-printing character is used in a format directive.
3914 * fixed several bugs in compiler checking of type declarations, i.e.
3915 violations of the Python "declarations are assertions" principle
3916 (thanks to Alexey Dejneka)
3917 * fixed several bugs in PCL's error checking (thanks to Gerd
3919 * fixed bug: printing of FILE-ERROR (thanks to Antonio
3921 * fixed bug in compilation of functions as first class values
3922 (thanks to Antonio Martinez-Shotton)
3923 * The compiler's handling TYPE-ERRORs which it can prove will
3924 inevitably happen at runtime has been cleaned up and corrected
3925 in several ways. (thanks to Alexey Dejneka)
3926 * improved argument type checking for various basic arithmetic
3927 operations (MAX, +, LOGXOR, etc.) which have had so much TLC
3928 lavished on them in the past that they can be compiled in many
3929 ways in different special cases
3930 * fixed bug 181: compiler checks validity of user supplied type
3932 * cleaned up code flushing in optimization: Function calls which
3933 should signal errors for safety purposes (e.g. which ANSI says
3934 should signal errors when their arguments are of incorrect type)
3935 are no longer optimized away.
3936 * added new extension: SB-DEBUG:BACKTRACE-AS-LIST
3937 * incremented fasl file version number, because changes in the
3938 implementation of sequence functions like COERCE caused
3939 internal utility functions like COERCE-TO-SIMPLE-VECTOR (used
3940 in old inline expansions) to become undefined. (Actually these
3941 changes were later undone, so we might very well be binary
3942 compatible with 0.7.7 after all, but leaving the version number
3943 incremented seemed like the simplest and most conservative
3946 changes in sbcl-0.7.7 relative to sbcl-0.7.6:
3947 * An alpha-quality port to the parisc architecture running Linux,
3948 based on the old CMUCL backend, has been made. This, even more so
3949 than the other backends, should be considered still a work in
3950 progress; known problems include that the Linux kernel in 64-bit
3951 mode does not propagate the correct sigcontext structure to
3952 userspace, and consequently SBCL on a parisc64 kernel will not
3954 * fixed bug 189: The compiler now respects NOTINLINE declarations for
3955 functions declared in FLET and LABELS. (I.e. "LET conversion" is
3956 suppressed.) Also now that the compiler is looking at declarations
3957 in the environment, it checks optimization declarations as well,
3958 and suppresses inlining when (> DEBUG SPEED).
3959 * More fixes have been made to treatment of floating point exception
3960 treatment and other Unix signals. In particular, floating point
3961 exceptions no longer cause Bus errors on the SPARC/Linux platform.
3962 * The detection and handling of control stack exhaustion (infinite
3963 or very deeply nested recursion) has changed. Stack exhaustion
3964 detection is now done by write-protecting pages at the OS level
3965 and applies at all optimization settings; when found, a
3966 SB-KERNEL:CONTROL-STACK-EXHAUSTED condition (subclass of
3967 STORAGE-CONDITION) is signalled, so stack exhaustion can no longer
3968 be caught using IGNORE-ERRORS.
3969 * Bugs 65, 70, and 109 fixed: The compiler now preserves invariants
3970 correctly when transforming recursive LABELS functions to LETs.
3971 (thanks to Alexey Dejneka)
3972 * Bug 48a./b. fixed: SYMBOL-MACROLET now refuses to bind symbols
3973 that are names of constants or global variables.
3974 * Bug fix: DEFINE-ALIEN-ROUTINE now declaims the correct FTYPE for
3975 alien routines with docstrings.
3976 * Bug 184 fixed: Division of ratios by the integer 0 now signals an
3977 error of type DIVISION-BY-ZERO. (thanks to Wolfhard Buss and
3979 * Bug fix: Errors in PARSE-INTEGER are now of type PARSE-ERROR.
3980 (thanks to Eric Marsden)
3981 * Bug fix: COERCE to (COMPLEX FLOAT) of rationals now returns an
3982 object of type (COMPLEX FLOAT). (thanks to Wolfhard Buss)
3983 * Bug fix: The SPARC backend can now compile functions involving
3984 LOGAND and stack-allocated arguments. (thanks to Raymond Toy)
3985 * Bug fix: We no longer segfault on passing a non-FILE-STREAM stream
3986 to a functions expecting a PATHNAME-DESIGNATOR.
3987 * Bug fix: DEFGENERIC now enforces the ANSI restrictions on its
3988 lambda lists. (thanks to Alexey Dejneka)
3989 * Bug fix: changed encoding of PCL's internal MAKE-INSTANCE
3990 functions so that EXPORTing the name of the class doesn't cause
3991 MAKE-INSTANCE functions from earlier DEFCLASSes to get lost (thanks
3992 to Antonio Martinez for reporting this)
3993 * Bug 192 fixed: The internal primitive DATA-VECTOR-REF can now be
3994 constant-folded without failing an assertion. (thanks to Einar
3995 Floystad Dorum for reporting this)
3996 * Bugs 123 and 165 fixed: array specializations on as-yet-undefined
3997 types are now dealt with more correctly by the compiler.
3998 * Minor incompatible change: COMPILE-FILE-PATHNAME now merges its
3999 OUTPUT-FILE argument with its INPUT-FILE argument, resulting in
4000 behaviour analogous to RENAME-FILE. This puts its behaviour more
4001 in line with ANSI's wording on COMPILE-FILE-PATHNAME. (thanks to
4003 * The fasl file version number has changed again. (because of the
4004 bug fix involving the names of PCL MAKE-INSTANCE functions)
4006 changes in sbcl-0.7.6 relative to sbcl-0.7.5:
4007 * bug fix: Floating point exceptions are treated much more
4008 consistently on the x86/Linux and PPC/Linux platforms.
4009 * Array initialization with :INITIAL-ELEMENT is now much faster for
4010 cases when the compiler cannot open code the array creation, but
4011 does know what the UPGRADED-ARRAY-ELEMENT-TYPE will be. General
4012 array accesses have also seen a speed increase.
4013 * bug fix: LOAD :IF-DOES-NOT-EXIST NIL now works when file type is
4014 specified. (This was at the root of some bad interactions between
4015 SBCL and ILISP: thanks to Gregory Wright for diagnosing this and
4017 * bug fix: Internal error arguments for undefined functions are now
4018 computed correctly on the PPC/Linux platform.
4019 * bug fix: Bad &REST syntax is now checked correctly. (thanks to
4020 Raymond Toy's patch for CMU CL)
4021 * Support for the Solaris 9 operating environment has been included
4022 (thanks to Daniel Merritt)
4023 * A very ugly but hopefully complete draft of the missing FFI chapter
4024 of the manual has been created by reformatting the corresponding
4025 CMU CL manual chapter into (currently very ugly and incoherent)
4026 DocBook and bringing it up to date for SBCL behavior. Thus, the
4027 manual is now essentially complete, at least by my extreme
4028 once-and-only-once standards, whereby it's acceptable to refer to
4029 the doc strings of SB-EXT functions as the primary documentation.
4030 * The fasl file version number has changed again, due to cleanup of
4031 (user-invisible) bitrotted stuff. (E.g. *!INITIAL-FDEFN-OBJECTS*
4032 is no longer a static symbol.)
4034 changes in sbcl-0.7.5 relative to sbcl-0.7.4:
4035 * SBCL now builds with OpenMCL (version 0.12) as the
4036 cross-compilation host; also, more progress has been made toward
4037 bootstrapping under CLISP.
4038 * SBCL now runs on the Tru64 (aka OSF/1) operating system on the
4040 * bug 158 fixed: The compiler can now deal with integer loop
4041 increments different from 1; fixing this turned out also to fix
4043 * bug 169 fixed: no more bogus warnings about using lexical bindings
4044 despite the presence of perfectly good SPECIAL declarations (thanks
4045 to David Lichteblau)
4046 * bug 175 fixed: CHANGE-CLASS is now more ANSI-conforming,
4047 accepting initargs. (thanks to Espen Johnsen and Pierre Mai)
4048 * bug 179 fixed: DIRECTORY can now deal with filenames with pattern
4050 * bug 180 fixed: Method combination specifications no longer ignore
4051 the :MOST-SPECIFIC-LAST option. (thanks to Pierre Mai)
4052 * bug fix: Structure type predicate functions now check their argument
4053 count as they should.
4054 * bug fix: Classes with :METACLASS STRUCTURE-CLASS now print
4055 correctly. (thanks to Pierre Mai)
4056 * minor incompatible change: The --noprogrammer option is deprecated
4057 in favor of the new --disable-debugger option, which is very similar.
4058 (The major difference is that it takes effect at a slightly different
4059 time at startup, causing handling of errors in --sysinit and
4060 --userinit files will be affected differently.) The
4061 SB-EXT:DISABLE-DEBUGGER and SB-EXT:ENABLE-DEBUGGER functions have
4062 been added to allow this functionality to be controlled from ordinary
4063 Lisp code. (ENABLE-DEBUGGER should help people like the Debian
4064 maintainers, who might want to run non-interactive scripts to
4065 build SBCL cores which will later be used interactively.)
4066 * minor incompatible change: The LOAD function no longer, when given
4067 a wild pathname to load, loads all files matching that pathname.
4068 Instead, an error of type FILE-ERROR is signalled.
4070 changes in sbcl-0.7.4 relative to sbcl-0.7.3:
4071 * bug 147 fixed: The compiler preserves its block link/count
4072 invariants more correctly now so that it doesn't crash. (thanks
4074 * Dynamic loading of object files in OpenBSD is now supported. (thanks
4076 * COMPILE now works correctly on macros. (thanks to Matthias Hoelzl)
4077 * GET-MACRO-CHARACTER and SET-MACRO-CHARACTER now represent
4078 no-value-for-this-character as NIL (as specified by ANSI).
4079 * HOST-NAMESTRING on physical pathnames now returns a string that is
4080 valid as a host argument to MERGE-PATHNAMES and to MAKE-PATHNAME.
4081 (thanks to Christophe Rhodes)
4082 * The Alpha port handles icache flushing more correctly. (thanks to
4084 * More progress has been made toward bootstrapping under CLISP. (thanks
4085 to Christophe Rhodes)
4086 * The fasl file format has changed again, because dynamic loading
4087 on OpenBSD (which has non-ELF object files) motivated some cleanups
4088 in the way that foreign symbols are transformed and passed around.
4089 * minor incompatible change: The ASCII RUBOUT character, (CHAR-CODE 127),
4090 is no longer treated as whitespace by the reader, but instead as
4091 an ordinary character. Thus e.g. (READ-FROM-STRING "A
\7fB") returns
4092 |A
\7fB|, instead of A as it used to.
4094 changes in sbcl-0.7.3 relative to sbcl-0.7.2:
4095 * ANSI's DEFINE-SYMBOL-MACRO is now supported. (thanks to Nathan
4096 Froyd porting CMU CL code originally by Douglas Thomas Crosher)
4097 * SBCL now runs on the PPC archtiecture under Linux. It actually did
4098 this as of 0.7.1.45, but was left out of the previous news section
4099 (thanks to Dan Barlow)
4100 * SBCL now runs on the Solaris operating system on SPARC architectures
4101 (thanks to Christophe Rhodes's port of the CMUCL runtime)
4102 * cleanups to the runtime on SPARC, both Linux and Solaris, and for
4103 gcc>=3 (thanks to Nathan Froyd and Ingvar Mattsson)
4104 * SPARC backend cleanups, allowing builds of cores optimized for V8
4105 and V9 SPARCS, and also emission of code targeted to a particular
4106 backend chosen at runtime (thanks to Christophe Rhodes and Raymond
4108 * SBCL is closer to bootstrapping under CLISP, thanks to various
4109 fixes by Christophe Rhodes.
4110 * The fasl file format has changed again, to allow the compiler's
4111 INFO database to support symbol macros.
4112 * The user manual (in doc/) is formatted into HTML more nicely.
4113 (thanks to coreythomas)
4114 * The system is smarter about SUBTYPEP relationships, especially
4115 those involving NOT types (including types such as ATOM which are
4116 represented internally using NOT types). Thus SUBTYPEP is less
4117 likely to return (VALUES NIL NIL) in general, and in particular
4118 bugs 58 and (the remaining bits of) bug 50 are fixed. (thanks to
4120 * The fasl file format has changed again, because the internal
4121 representation of types now includes a new slot to support the new
4122 SUBTYPEP-of-NOT-types logic.
4123 * (not a change in the main branch of SBCL, but a related prototype
4124 which can hopefully be merged into the main branch of SBCL in the
4125 future:) Brian Spilsbury has produced a Unicode-enabled variant of
4126 sbcl-0.7.0, available as a patch against sbcl-0.7.0 at
4127 <http://designix.com.au/brian/SBCL/sbcl-0.7.0-unicode.p0.gz>.
4128 * Bug 151 fixed: GET-DISPATCH-MACRO-CHAR now returns NIL for undefined
4129 dispatch macro character combinations. (thanks to Alexey Dejneka)
4130 * Bugfix in PARSE-NAMESTRING: we now correctly parse unix namestrings
4131 that superficially look like logical namestrings correctly.
4132 * USER-HOMEDIR-PATHNAME now returns a (physical) pathname that SBCL
4134 * Bugfix in DEFSTRUCT: BOA constructor lambda lists now accept (name
4135 default supplied-p) for &optional and &key arguments. (thanks to
4138 changes in sbcl-0.7.2 relative to sbcl-0.7.1:
4139 * incompatible change: The compiler is now less aggressive about
4140 tail call optimization, doing it only when (> SPACE DEBUG) or
4141 (> SPEED DEBUG). (This is an incompatible change because there are
4142 programs which relied on the old CMU-CL-style behavior to optimize
4143 away their unbounded recursion which will now die of stack overflow.)
4144 * minor incompatible change: The default BYTES-CONSED-BETWEEN-GCS
4145 for non-GENCGC systems has been increased to 20M (since that
4146 seems much closer to the likely performance optimum for modern
4147 systems than the old 4M value was)
4148 * minor incompatible change: new larger values for *DEBUG-PRINT-LENGTH*
4149 and *DEBUG-PRINT-LEVEL*
4150 * SBCL runs on SPARC systems now. (thanks to Christophe Rhodes' port
4151 of CMU CL's support for SPARC, and various endianness and other
4152 SBCL portability fixes due to Christophe Rhodes and Dan Barlow)
4153 * new syntactic sugar for the Unix command line: --load foo.bar is now
4154 an alternate notation for --eval '(load "foo.bar")'.
4156 ** The system now detects stack overflow and handles it gracefully,
4157 at least for (OR (> SAFETY (MAX SPEED SPACE)) (= SAFETY 3))
4158 optimization settings. (This is a good thing in general, and
4159 its introduction in this version should be particularly timely
4160 for anyone whose code fails because of suppression of tail
4162 ** The system now hunts for the C variable "environ" in a more
4163 devious way, to avoid segfaults when the C library version differs
4164 between compile time and run time. (thanks to Christophe Rhodes)
4165 ** INTEGER-valued CATCH tags now work. (thanks to Alexey Dejneka,
4166 and also to Christophe Rhodes for porting the fix to non-X86 CPUs)
4167 ** The compiler no longer issues bogus style warnings for undefined
4168 classes in the same source file as the DEFCLASSes which defined
4169 them. (thanks to Stig E Sandoe for reporting and Martin Atzmueller
4171 ** fixes in CONDITION class precedence list for undefined function
4172 errors (thanks to Alexei Dejneka)
4173 ** *DEFAULT-PATHNAME-DEFAULTS* is used more consistently and
4174 correctly. (thanks to Dan Barlow)
4175 ** portability fixes aiming at bootstrapping under CLISP (thanks
4176 to Dave McDonald and Christophe Rhodes)
4177 ** FORMAT fixes (thanks to Robert Strandh and Dan Barlow)
4178 ** fixes in type translation and and type inference (thanks to
4180 ** fixes to optimizer internal errors (thanks to Alexei Dejneka)
4181 ** various fixes in the new ports (thanks to Dan Barlow)
4182 * several changes related to debugging:
4183 ** suppression of tail recursion, as noted above
4184 ** stack overflow detection, as noted above
4185 ** The default implementation of TRACE has changed. :ENCAPSULATE T
4186 is now the default. (For some time encapsulation has been more
4187 reliable than the breakpoint-based :ENCAPSULATE NIL
4188 implementation, at least on X86 systems; and I just noticed that
4189 encapsulation also seems closer to the spirit of the ANSI
4192 changes in sbcl-0.7.1 relative to sbcl-0.7.0:
4194 ** SB-ALIEN:LOAD-FOREIGN and SB-ALIEN:LOAD-1-FOREIGN are set
4195 up properly again. (There was a packaging bug in 0.7.0 which
4196 left their definitions in SB-SYS::LOAD-FOREIGN and
4197 SB-SYS::LOAD-1-FOREIGN. LOAD-FOREIGN and LOAD-1-FOREIGN are
4198 vital for most things which interface to C-level interfaces,
4199 like extensions working with sockets or databases or
4200 Perl-compatible regexes or whatever, and the need to fix
4201 this bug is the main reason that 0.7.1 was released so
4203 ** DEFGENERIC is now choosier about the methods it redefines, so that
4204 reLOADing a previously-LOADed file containing DEFGENERICs does
4205 the right thing now. Thus, the Lispy edit/reLOAD-a-little/test
4206 cycle now works as it should. (thanks to Alexey Dejneka)
4207 ** Bug 106 (types (COMPLEX FOO) where FOO is an obscure type) was
4208 fixed by Christophe Rhodes. (He actually submitted this patch
4209 months ago, and I delayed until after 0.7.0.)
4210 ** Bug 111 (internal compiler confusion about runtime checks on
4211 FUNCTION types) was fixed by Alexey Dejneka.
4212 * Some internal cleanups (getting rid of variables which aren't
4213 needed now that the byte interpreter is gone) caused the fasl
4214 file format number to change again.
4216 changes in sbcl-0.7.0 relative to sbcl-0.6.13:
4217 * major incompatible change: The default fasl file extension, i.e. the
4218 default extension for files produced by COMPILE-FILE, has changed
4219 to ".fasl", for all architectures. (No longer ".x86f" and ".axpf".)
4221 ** There are many changes in the implementation of the compiler.
4222 SBCL is now essentially a compiler-only implementation of ANSI
4223 Common Lisp. EVAL still "interprets" a few special cases, but
4224 almost all the interesting cases are handled by creating
4225 a LAMBDA expression, calling COMPILE on it, then calling
4226 FUNCALL on the result.
4227 ** The EVAL-WHEN code has been rewritten to be ANSI-compliant, and
4228 various related bugs (IR1-1, IR1-2, IR1-3, IR1-3a) have gone away.
4229 Since the code is newer, there might still be some new bugs
4230 (though not as many as before Martin Atzmueller's fixes:-). But
4231 the new code is substantially simpler and clearer, and hopefully
4232 any remaining bugs will be simpler, less fundamental, and more
4233 fixable then the bugs in the old code.
4234 ** The revised compiler is still a little unsteady on its feet.
4236 *** The debugging information it produces (particularly the names
4237 of FUNCTION objects) is sometimes much less useful than what
4238 the old compiler produced.
4239 *** The support for inlining FOO when you (DECLAIM (INLINE FOO))
4240 then do (DEFUN FOO ..) in a non-null lexical environment (e.g.
4241 within a MACROLET) has been temporarily weakened.
4242 ** There are new compiler optimizations for various functions:
4243 *** the sequence functions FIND, POSITION, FIND-IF, POSITION-IF,
4244 FIND-IF-NOT, POSITION-IF-NOT, and FILL
4245 *** the math functions TRUNCATE, FLOOR, and CEILING
4246 *** the function-of-all-trades COERCE
4247 Mostly these should be transparent, but there's one
4248 potentially-annoying problem (bug 117): when the compiler
4249 inline-expands a function and does type analysis on the result,
4250 it can create control paths which have type mismatches, and
4251 when it can't prove that those control paths aren't taken,
4252 it will issue WARNINGs about the type mismatches. This is
4253 a particular problem in practice for the new sequence functions.
4254 It's not clear how this should be fixed, and for now, a
4255 workaround is given in the entry for 117 in the BUGS file.
4256 ** (Because of the interaction between the two previous items --
4257 occasional inlining problems and new inline expansions -- some
4258 of the new sequence function optimizations won't really kick in
4259 completely until debugging information, and then inlining, are
4260 straightened out in some future version.)
4261 * minor incompatible changes:
4262 ** As part of a bug fix by Christophe Rhodes to DIRECTORY behavior,
4263 DIRECTORY no longer implicitly promotes NIL slots of its
4264 pathname argument to :WILD. In particular, when you ask for the
4265 contents of a directory (which you used to be able to do without
4266 explicit wildcards, e.g. (DIRECTORY "/tmp/")) you now need to use
4267 explicit wildcards, e.g. (DIRECTORY "/tmp/*.*").
4268 ** changes in behavior that ANSI explicitly defines to be
4269 implementation dependent:
4270 *** The new compiler-only implementation still conforms with ANSI,
4271 but acts a little different than before. Besides the obvious
4272 changes in performance tradeoffs (that the cost per form passed
4273 to EVAL has gone up, and the cost per form executed by EVAL
4274 has gone down), the behavior of the system changes a little
4275 because there are no longer any interpreted function objects.
4276 COMPILED-FUNCTION-P is now synonymous with FUNCTIONP, and
4277 e.g. doing COMPILE on the output of interactive DEFUN is
4279 *** The value of INTERNAL-TIME-UNITS-PER-SECOND has been increased
4281 *** The default for the USE list in MAKE-PACKAGE and DEFPACKAGE
4282 has changed from (:CL) to NIL.
4283 *** The CHAR-NAME of unprintable ASCII characters which, unlike
4284 e.g. #\Newline and #\Tab, don't have names specified in the
4285 ANSI Common Lisp standard, is now based on their ASCII symbolic
4286 names (#\Nul, #\Soh, #\Stx, etc.) The old CMU-CL-style names
4287 (#\Null, #\^a, #\^b, etc.) are still accepted by NAME-CHAR, but
4288 are no longer used for output.
4289 ** changes in internal implementation constants:
4290 *** The default value of *BYTES-CONSED-BETWEEN-GCS* has doubled, to
4291 4 million. (If your application spends a lot of time GCing and
4292 you have a lot of RAM, you might want to experiment with
4293 increasing it even more.)
4294 ** The SB-C-CALL package has been merged into the SB-ALIEN package.
4295 However, almost all old code should still continue to work without
4296 immediate update, as SB-C-CALL is now a (deprecated) nickname
4298 ** Old operator names in the style DEF-FOO are now deprecated in
4299 favor of new corresponding names DEFINE-FOO, for consistency with
4300 the naming convention used in the ANSI standard (DEFSTRUCT, DEFVAR,
4301 DEFINE-CONDITION, DEFINE-MODIFY-MACRO..). This mostly affects
4302 internal symbols, but a few supported extensions like
4303 SB-ALIEN:DEF-ALIEN-FUNCTION are also affected. (So e.g.
4304 DEF-ALIEN-FUNCTION becomes DEFINE-ALIEN-FUNCTION.)
4305 ** The debugger prompt sequence now goes "5]", "5[2]", "5[3]",
4306 etc. as you get deeper into recursive calls to the debugger
4307 command loop, instead of the old "5]", "5]]", "5]]]"
4308 sequence. (I was motivated to do this when squabbles between
4309 ILISP and SBCL left me very deeply nested in the debugger. In the
4310 short term, this change will probably provoke more ILISP/SBCL
4311 squabbles, but hopefully it will be an improvement in the long run.)
4312 ** SB-ALIEN:DEFINE-ALIEN-FUNCTION (also known by the old deprecated
4313 name DEF-ALIEN-FUNCTION) now does DECLAIM FTYPE for the defined
4314 function, since declaiming return types involving aliens is
4315 (1) annoyingly messy to do by hand and (2) vital to efficient
4316 compilation of code which calls such functions.
4317 ** SB-ALIEN:LOAD-FOREIGN and SB-ALIEN:LOAD-1-FOREIGN are no
4318 longer reexported by the SB-EXT package. They're solely useful
4319 for alien code, so it seems more logical that you should get
4320 them from the SB-ALIEN package, not in SB-EXT.
4321 ** :SB-CONSTRAIN-FLOAT-TYPE, :SB-PROPAGATE-FLOAT-TYPE, and
4322 :SB-PROPAGATE-FUN-TYPE are no longer considered to be optional
4323 features. Instead, the code that they used to control is always
4324 built into the system.
4325 * many other bug fixes
4326 ** DEFSTRUCT and DEFCLASS have been substantially updated to take
4327 advantage of the new EVAL-WHEN stuff and to clean them up in
4328 general, and they are now more ANSI-compliant in a number of
4329 ways. Martin Atzmueller is responsible for a lot of this.
4330 ** Besides the cleanups discussed above, Martin Atzmueller fixed
4332 *** fixes in READ-SEQUENCE and WRITE-SEQUENCE
4333 *** correct ERROR type for various file operations
4334 *** some fixes for Lisp streams
4335 *** DEFMETHOD syntax checking
4336 *** changing old weird representation of debug information as
4337 strings (which, among their other deficiencies, don't transform
4338 correctly when you rename packages, and don't change their
4339 print representation when you change things like *PACKAGE*
4340 and *PRINT-LENGTH*) to symbols and lists of symbols
4341 He also made several improvements and fixed several bugs in DESCRIBE.
4342 ** Alexey Dejneka fixed many bugs, including classic bugs and bugs he
4344 *** misbehavior of WRITE-STRING/WRITE-LINE
4345 *** LOOP over keys of a hash table, LOOP bugs 49b and 81 and 103,
4346 and several other LOOP problems as well
4347 *** DIRECTORY when similar filenames are present
4348 *** DEFGENERIC with :METHOD options
4349 *** bug 126, in (MAKE-STRING N :INITIAL-ELEMENT #\SPACE))
4350 *** bug in the optimization of ARRAY-ELEMENT-TYPE
4351 *** argument ordering in FIND with :TEST option
4352 *** mishandled package designator argument in APROPOS-LIST
4353 *** various problems in the backquote readmacro
4354 *** a bug in APROPOS
4355 *** probably some others that I'm not describing very well here,
4356 since the CVS log documents them by reference to sbcl-devel
4357 messages, and the SourceForge archives aren't working well.:-(
4358 ** Dan Barlow improved the Alpha port (and is making progress on the
4359 PPC port, for those of you who think different).
4360 ** Besides the DIRECTORY fixes and changes mentioned elsewhere,
4361 Christophe Rhodes cleaned up the system self-test scripts (in tests/*),
4362 contributed the optimization of FIND-IF-NOT and POSITION-IF-NOT, and
4363 continues to work on the SPARC port (for those of you in a position
4364 to look down upon our little PC-compatible boxes from a great height).
4365 ** PPRINT-LOGICAL-BLOCK now copies the *PRINT-LINES* value on entry
4366 and uses that copy, rather than the current dynamic value, when
4367 it's trying to decide whether to truncate output. Thus e.g.
4368 (let ((*print-lines* 50))
4369 (pprint-logical-block (stream nil)
4371 (let ((*print-lines* 8))
4372 (print (aref possiblybigthings i) stream)))))
4373 should now truncate the logical block only at 50 lines, instead of
4374 often truncating it at 8 lines, as it did before.
4375 * The doc/cmucl/ directory, containing old CMU CL documentation
4376 from the time of the fork, is no longer part of the base system.
4377 SourceForge has shut down its anonymous FTP service, and with it
4378 my original plan for distributing the old CMU CL documentation
4379 there. For now, if you need these files you can download an old
4380 SBCL source release and extract them from it.
4381 * The fasl file version number changed again, for dozens of reasons,
4382 some of which are apparent above.
4384 changes in sbcl-0.6.13 relative to sbcl-0.6.12:
4385 * a port to the Compaq/DEC Alpha CPU, thanks to Dan Barlow
4386 * Martin Atzmueller ported Tim Moore's marvellous CMU CL DISASSEMBLE
4387 patch, so that DISASSEMBLE output is much nicer.
4388 * The code in the SB-PROFILE package now seems reasonably stable.
4389 I still haven't decided what the final interface should look like
4390 (I'd like PROFILE to interact cleanly with TRACE, since both
4391 facilities use function encapsulation) but if you have a need
4392 for profiling now, you can probably use it successfully with
4393 the current CMU-CL-style interface.
4394 * Pathnames and *DEFAULT-DIRECTORY-DEFAULTS* are much more
4395 ANSI-compliant, thanks to various fixes and tests from Dan Barlow.
4396 Also, at Dan Barlow's suggestion, TRUENAME on a dangling symbolic
4397 link now returns the dangling link itself, and for similar
4398 reasons, TRUENAME on a cyclic symbolic link returns the cyclic
4399 link itself. (In these cases the old code signalled an error and
4400 looped endlessly, respectively.) Thus, DIRECTORY now works even
4401 in the presence of dangling and cyclic symbolic links.
4402 * Compiler trace output (the :TRACE-FILE option to COMPILE-FILE)
4403 is now a supported extension again, since the consensus on
4404 sbcl-devel was that it can be useful for ordinary development
4405 work, not just for debugging SBCL itself.
4406 * The default for SB-EXT:*DERIVE-FUNCTION-TYPES* has changed to
4407 NIL, i.e. ANSI behavior, i.e. the compiler now recognizes
4408 that currently-defined functions might be redefined later with
4409 different return types.
4410 * Hash tables can be printed readably, as inspired by CMU CL code
4411 of Eric Marsden and SBCL code of Martin Atzmueller.
4412 * better error handling in CLOS method combination, thanks to
4413 Martin Atzmueller porting Pierre Mai's CMU CL patches
4414 * more overflow fixes for >16Mbyte I/O buffers
4415 * A bug in READ has been fixed, so that now a single Ctrl-D
4416 character suffices to cause end-of-file on character streams.
4417 In particular, now you only need one Ctrl-D at the command
4418 line (not two) to exit SBCL.
4419 * fixed bug 26: ARRAY-DISPLACEMENT now returns (VALUES NIL 0) for
4421 * fixed bug 107 (reported as a CMU CL bug by Erik Naggum on
4422 comp.lang.lisp 2001-06-11): (WRITE #*101 :RADIX T :BASE 36) now
4423 does the right thing.
4424 * The implementation of some type tests, especially for CONDITION
4425 types, is now tidier and maybe faster, due to CMU CL code
4426 originally by Douglas Crosher, ported by Martin Atzmueller.
4427 * Some math functions have been fixed, and there are new
4428 optimizers for deriving the types of COERCE and ARRAY-ELEMENT-TYPE,
4429 thanks to Raymond Toy's work on CMU CL, ported by Martin Atzmueller.
4430 * (There are also some new optimizers in contrib/*-extras.lisp. Those
4431 aren't built into sbcl-0.6.13, but are a sneak preview of what's
4432 likely to be built into sbcl-0.7.0.)
4433 * A bug in COPY-READTABLE was fixed. (Joao Cachopo's patch to CMU
4434 CL, ported to SBCL by Martin Atzmueller)
4435 * DESCRIBE now gives more information in some cases. (Pierre Mai's
4436 patch to CMU CL, ported to SBCL by Martin Atzmueller)
4437 * Martin Atzmueller and Bill Newman fixed some bugs in INSPECT.
4438 * There's a new slam.sh hack to shorten the edit/compile/debug
4439 cycle for low-level changes to SBCL itself, and a new
4440 :SB-AFTER-XC-CORE target feature to control the generation of
4441 the after-xc.core file needed by slam.sh.
4442 * minor incompatible change: The ENTRY-POINTS &KEY argument to
4443 COMPILE-FILE is no longer supported, so that now every function
4444 gets an entry point, so that block compilation looks a little
4445 more like the plain vanilla ANSI section 3.2.2.3 scheme.
4446 * minor incompatible change: SB-EXT:GET-BYTES-CONSED now
4447 returns the number of bytes consed since the system started,
4448 rather than the number consed since the first time the function
4449 was called. (The new definition parallels ANSI functions like
4450 CL:GET-INTERNAL-RUN-TIME.)
4451 * minor incompatible change: The old CMU-CL-style DIRECTORY options,
4452 i.e. :ALL, :FOLLOW-LINKS, and :CHECK-FOR-SUBDIRS, are no longer
4453 supported. Now DIRECTORY always does the abstract Common-Lisp-y
4454 thing, i.e. :ALL T :FOLLOW-LINKS T :CHECK-FOR-SUBDIRS T.
4455 * Fasl file version numbers are now independent of the target CPU,
4456 since historically most system changes which required version
4457 number changes have affected all CPUs equally. Similarly,
4458 the byte fasl file version is now equal to the ordinary
4461 changes in sbcl-0.6.12 relative to sbcl-0.6.11:
4462 * incompatible change: The old SB-EXT:OPTIMIZE-INTERFACE declaration
4463 is no longer recognized. I apologize for this, because it was
4464 listed in SB-EXT as a supported extension, but I found that
4465 its existing behavior was poorly specified, as well as incorrectly
4466 specified, and it looked like too much of a mess to straighten it
4467 out. I have enough on my hands trying to get ANSI stuff to work..
4468 * many patches ported from CMU CL by Martin Atzmueller, with
4469 half a dozen bug fixes in pretty-printing and the debugger, and
4470 half a dozen others elsewhere
4471 * fixed bug 13: Floating point infinities are now supported again.
4472 They might still be a little bit flaky, but thanks to bug reports
4473 from Nathan Froyd and CMU CL patches from Raymond Toy they're not
4474 as flaky as they were.
4475 * The --noprogrammer command line option is now supported. (Its
4476 behavior is slightly different in detail from what the old man
4477 page claimed it would do, but it's still appropriate under the
4478 same circumstances that the man page talks about.)
4479 * The :SB-PROPAGATE-FLOAT-TYPE and :SB-PROPAGATE-FUN-TYPE features
4480 are now supported, and enabled by default. Thus, the compiler can
4481 handle many floating point and complex operations much less
4482 inefficiently. (Thus e.g. you can implement a complex FFT
4484 * The compiler now detects type mismatches between DECLAIM FTYPE
4485 and DEFUN better, and implements CHECK-TYPE more correctly, and
4486 SBCL builds under CMU CL again despite its non-ANSI EVAL-WHEN,
4487 thanks to patches from Martin Atzmueller.
4488 * various fixes to make the cross-compiler more portable to
4489 ANSI-conforming-but-different cross-compilation hosts (notably
4490 Lispworks for Windows, following bug reports from Arthur Lemmens)
4491 * A bug in READ-SEQUENCE for CONCATENATED-STREAM, and a gross
4492 ANSI noncompliance in DEFMACRO &KEY argument parsing, have been
4493 fixed thanks to Pierre Mai's CMU CL patches.
4494 * fixes to keep the system from overflowing internal counters when
4495 it tries to use i/o buffers larger than 16M bytes
4496 * fixed bug 45a: Various internal functions required to support
4497 complex special functions have been merged from CMU CL sources.
4498 (When I was first setting up SBCL, I misunderstood a compile-time
4499 conditional #-OLD-SPECFUN, and so accidentally deleted them.)
4500 * improved support for type intersection and union, fixing bug 12
4501 (e.g., now (SUBTYPEP 'KEYWORD 'SYMBOL)=>T,T) and some other
4502 more obscure bugs as well
4503 * some steps toward byte-compiling non-performance-critical
4504 parts of the system, courtesy of patches from Martin Atzmueller
4505 * Christophe Rhodes has made some debian packages of sbcl at
4506 <http://www-jcsu.jesus.cam.ac.uk/ftp/pub/debian/lisp>.
4507 From his sbcl-devel e-mail of 2001-04-08 they're not completely
4508 stable, but are nonetheless usable. When he's ready, I'd be happy
4509 to add them to the SourceForge "File Releases" section. (And if
4510 anyone wants to do RPMs or *BSD packages, they'd be welcome too.)
4511 * new fasl file format version number (because of changes in
4512 internal representation of (OR ..) types to accommodate the new
4513 support for (AND ..) types, among other things)
4515 changes in sbcl-0.6.11 relative to sbcl-0.6.10:
4516 * Martin Atzmueller pointed out that bugs #9 and #25 are gone in
4518 * bug 34 fixed by Martin Atzmueller: dumping/loading instances works
4520 * fixed bug 40: TYPEP, SUBTYPEP, UPGRADED-ARRAY-ELEMENT-TYPE,
4521 and UPGRADED-COMPLEX-PART-TYPE now work better with of compound
4522 types built from undefined types, e.g. '(VECTOR SOME-UNDEF-TYPE).
4523 * DESCRIBE now works on structure objects again.
4524 * Most function call argument type mismatches are now handled as
4525 STYLE-WARNINGs instead of full WARNINGs, since the compiler doesn't
4526 know whether the function will be redefined before the call is
4527 executed. (The compiler could flag local calls with full WARNINGs,
4528 as per the ANSI spec "3.2.2.3 Semantic Constraints", but right now
4529 it doesn't keep track of enough information to know whether calls
4530 are local in this sense.)
4531 * Compiler output is now more verbose, with messages truncated
4532 later than before. (There should be some supported way for users
4533 to override the default verbosity, but I haven't decided how to
4534 provide it yet, so this behavior is still controlled by the internal
4535 SB-C::*COMPILER-ERROR-PRINT-FOO* variables in
4536 src/compiler/ir1util.lisp.)
4537 * Fasl file format version numbers have increased again, because
4538 support for the Gray streams extension changes the layout of the
4539 system's STREAM objects.
4540 * The Gray subclassable streams extension now works, thanks to a
4541 patch from Martin Atzmueller.
4542 * The full LOAD-FOREIGN extension (not just the primitive
4543 LOAD-FOREIGN-1) now works, thanks to a patch from Martin Atzmueller.
4544 * The default behavior of RUN-PROGRAM has changed. Now, unlike CMU CL
4545 but like most other programs, it defaults to copying the Unix
4546 environment from the original process instead of starting the
4547 new process in an empty environment.
4548 * Extensions which manipulate the Unix environment now support
4549 an :ENVIRONMENT keyword option which doesn't smash case or
4550 do other bad things. The CMU-CL-style :ENV option is retained
4551 for porting convenience.
4552 * LOAD-FOREIGN (and LOAD-1-FOREIGN) now support logical pathnames,
4553 as per Daniel Barlow's suggestion and Martin Atzmueller's patch
4555 changes in sbcl-0.6.10 relative to sbcl-0.6.9:
4557 * A patch from Martin Atzmueller seems to have solved the SIGINT
4558 problem, and as far as we know, signal-handling now works cleanly.
4559 (If you find any new bugs, please report them!)
4560 * The system no longer defaults Lisp source file names to types
4561 ".l", ".cl", or ".lsp", but only to ".lisp".
4562 * The compiler no longer uses special default file extensions for
4563 byte-compiled code. (The ANSI definition of COMPILE-FILE-PATHNAME
4564 seems to expect a single default extension for all compiled code,
4565 and there's no compelling reason to try to stretch the standard
4566 to allow two different extensions.) Instead, byte-compiled files
4567 default to the same extension as native-compiled files.
4568 * Fasl file format version numbers have increased again, because
4569 a rearrangement of internal implementation packages made some
4570 dumped symbols in old fasl files unreadable in new cores.
4571 * DECLARE/DECLAIM/PROCLAIM logic is more nearly ANSI in general, with
4572 many fewer weird special cases.
4573 * Bug #17 (differing COMPILE-FILE behavior between logical and
4574 physical pathnames) has been fixed, and some related misbehavior too,
4575 thanks to a patch from Martin Atzmueller.
4576 * Bug #30 (reader problems) is gone, thanks to a CMU CL patch
4577 by Tim Moore, ported to SBCL by Martin Atzmueller.
4578 * Martin Atzmueller fixed several filesystem-related problems,
4579 including bug #36, in part by porting CMU CL patches, which were
4580 written in part by Paul Werkowski.
4581 * More compiler warnings in src/runtime/ are gone, thanks to
4582 more patches from Martin Atzmueller.
4583 * Martin Atzmueller pointed out that bug 37 was fixed by his patches
4586 changes in sbcl-0.6.9 relative to sbcl-0.6.8:
4588 * DESCRIBE now works on CONDITION objects.
4589 * The debugger now handles errors which arise when trying to print
4590 *DEBUG-CONDITION*, so that it's less likely to fall into infinite
4592 * The build system now uses an additional file, customize-target-features.lisp,
4593 to allow local modifications to the target *FEATURES* list. (The point of
4594 this is that now I can set up a custom configuration, e.g. with :SB-SHOW
4595 debugging features enabled, without having to worry about propagating it
4596 into everyone's system when I do a "cvs update".) When no
4597 customize-target-features.lisp file exists, the target *FEATURES* list
4598 should be constructed the same way as before.
4599 * fixed bugs in DEFCONSTANT ANSI-compatibility:
4600 ** DEFCONSTANT now tests reassignments using EQL, not EQUAL, in order to
4601 warn about behavior which is undefined under the ANSI spec. Note: This
4602 is specified by ANSI, but it's not very popular with programmers.
4603 If it causes you problems, take a look at the new SB-INT:DEFCONSTANT-EQX
4604 macro in the SBCL sources for an example of a workaround which you
4605 might use to make portable ANSI-standard code which does what you want.
4606 ** DEFCONSTANT's implementation is now based on EVAL-WHEN instead of on
4607 pre-ANSI IR1 translation magic, so it does the ANSI-specified thing
4608 when it's used as a non-toplevel form. (This is required in order
4609 to implement the DEFCONSTANT-EQX macro.)
4610 ** (DEFCONSTANT X 1) (DEFVAR X) (SETF X 2) no longer "works".
4611 ** Unfortunately, non-toplevel DEFCONSTANT forms can still do some
4612 funny things, due to bugs in the implementation of EVAL-WHEN
4613 (bug #IR1-3). This probably won't be fixed until 0.7.x. (Fortunately,
4614 non-toplevel DEFCONSTANTs are uncommon.)
4615 * The core file version number and fasl file version number have been
4616 incremented, because the old noncompliant DEFCONSTANT behavior involved
4617 calling functions which no longer exist, and because I also took the
4618 opportunity to chop an unsupported slot out of the DEBUG-SOURCE structure.
4619 * fixed bug 1 (error handling before read-eval-print loop starts), and
4620 redid debugger restarts and related debugger commands somewhat while
4622 ** The QUIT debugger command is gone, since it did something
4623 rather different than the SB-EXT:QUIT command, and since it never
4624 worked properly outside the main toplevel read/eval/print loop.
4625 Invoking the new TOPLEVEL restart provides the same functionality.
4626 ** The GO debugger command is also gone, since you can just invoke
4627 the CONTINUE restart directly instead.
4628 ** The TOP debugger command is also gone, since it's redundant with the
4629 FRAME 0 command, and since it interfered with abbreviations for the
4631 * The system now recovers better from non-PACKAGE values of the *PACKAGE*
4633 * The system now understands compound CONS types (e.g. (CONS FIXNUM T))
4634 as required by ANSI. (thanks to Douglas Crosher's CMU CL patches, with
4635 some porting work by Martin Atzmueller)
4636 * Martin Atzmueller reviewed the CMU CL mailing lists and came back
4637 with a boatload of patches which he ported to SBCL. Now that those
4639 ** The system tries to make sure that its low-priority messages
4640 are prefixed by semicolons, to help people who like to use
4641 syntax highlighting in their ILISP buffer. (This patch
4642 was originally due to Raymond Toy.)
4643 ** The system now optimizes INTEGER-LENGTH better, thanks to more
4644 patches originally written by Raymond Toy.
4645 ** The compiler understands coercion between single-value and
4646 multiple-VALUES type expressions better, getting rid of some very
4647 weird behavior, thanks to patches originally by Robert MacLachlan
4648 and Douglas Crosher.
4649 ** The system understands ANSI-style non-KEYWORD &KEY arguments in
4650 lambda lists, thanks to a patch originally by Pierre Mai.
4651 ** The system no longer bogusly warns about "abbreviated type
4653 ** The compiler gets less confused by inlining and RETURN-FROM,
4654 thanks to some patches originally by Tim Moore.
4655 ** The system no longer hangs when dumping circular lists to fasl
4656 files, thanks to a patch originally from Douglas Crosher.
4657 * Martin Atzmueller also fixed ROOM, so that it no longer fails with an
4658 undefined function error.
4659 * gave up on fixing bug 3 (forbidden-by-ANSI warning for type mismatch
4660 in structure slot initforms) for now, documented workaround instead:-|
4661 * fixed bug 4 (no WARNING for DECLAIM FTYPE of slot accessor function)
4662 * fixed bug 5: added stubs for various Gray stream functions called
4663 in the not-a-CL:STREAM case, so that even when Gray streams aren't
4664 installed, at least appropriate type errors are generated
4665 * fixed bug 8: better reporting of various PROGRAM-ERRORs
4666 * fixed bug 9: IGNORE and IGNORABLE now work reasonably and more
4667 consistently in DEFMETHOD forms.
4668 * removed bug 21 from BUGS, since Martin Atzmueller points out that
4669 it doesn't seem to affect SBCL after all
4670 * The C runtime system now builds with better optimization and many
4671 fewer warnings, thanks to lots of cleanups by Martin Atzmueller.
4673 changes in sbcl-0.6.8 relative to sbcl-0.6.7:
4675 * The system is now under CVS at SourceForge (instead of the
4676 CVS repository on my home machine).
4677 * The new signal handling code has been tweaked to treat register
4678 contents as (UNSIGNED-BYTE 32), as the old CMU CL code did,
4679 instead of (SIGNED-BYTE 32), as the C header files have it. (Code
4680 downstream, e.g. in debug-int.lisp, has implicit dependencies
4681 on the unsignedness of integer representation of machine words,
4682 and that caused the system to bomb out with infinite regress
4683 when trying to recover from type errors involving signed values,
4684 e.g. (BUTLAST '(1 2 3) -1).)
4685 * (BUTLAST NIL) and (NBUTLAST NIL) now return NIL as they should.
4686 (This was one of the bugs Peter Van Eynde reported back in July.)
4687 * The system now uses code inspired by Colin Walters' O(N)
4688 implementation of MAP (from the cmucl-imp@cons.org mailing
4689 list, 2 September 2000) when it can't use a DEFTRANSFORM to
4690 inline the MAP operation, and there is more than one
4691 sequence argument to the MAP call (so that it can't just
4692 do ETYPECASE once and for all based on the type of the
4693 single sequence argument). (The old non-inline implementation
4694 of the general M-argument sequence-of-length-N case required
4695 O(M*N*N) time when any of the sequence arguments were LISTs.)
4696 * The QUIT :UNIX-CODE keyword argument has been renamed to
4697 QUIT :UNIX-STATUS. (The old name still works, but is deprecated.)
4698 * Raymond Wiker's patches to port RUN-PROGRAM from CMU CL to SBCL
4700 * Raymond Wiker's patches to port dynamic loading from Linux to
4701 FreeBSD have been added.
4702 * The BUGS file is now more nearly up to date, thanks in large part
4703 to Martin Atzmueller's review of it.
4704 * The debugger now flushes standard output streams before it begins
4705 its output ("debugger invoked" and so forth).
4706 * The core version number and fasl file version number have both
4707 been incremented, because of incompatible changes in the layout
4709 * FINISH-OUTPUT is now called more consistently on QUIT. (It
4710 used to not be called for a saved Lisp image.)
4711 * Martin Atzmueller's version of a patch to fix a compiler crash,
4712 as posted on sbcl-devel 13 September 2000, has been installed.
4713 * Instead of installing Martin Atzmueller's patch for the
4714 compiler transform for SUBSEQ, I deleted the compiler transform,
4715 and transforms for some similar consing operations.
4716 * A bug in signal handling which kept TRACE from working on OpenBSD
4718 * added enough DEFTRANSFORMs to allow (SXHASH 'FOO) to be optimized
4719 away by constant folding
4720 * The system now defines its address space constants in one place
4721 (in the Lisp sources), and propagates them automatically elsewhere
4722 (through GENESIS and the sbcl.h file). Therefore, patching the
4723 address map is less unnecessarily tedious and error-prone. The
4724 Lisp names of address space constants have also been systematized.
4725 * CVS tags like dollar-Header-dollar have been removed from
4726 the sources, because they have never saved me trouble and
4727 they've been source of trouble working with patches and other
4728 diff-related operations.
4729 * fixed the PROG1-vs.-PROGN bug in HANDLER-BIND (reported by
4730 ole.rohne@cern.ch on cmucl-help@cons.org 2000-10-25)
4732 changes in sbcl-0.6.7 relative to sbcl-0.6.6:
4734 * The system has been ported to OpenBSD.
4735 * The system now compiles with a simple "sh make.sh" on the systems
4736 that it's supported on. I.e., now you no longer need to tweak
4737 text in the target-features.lisp-expr and symlinks in src/runtime/
4738 by hand, the make.sh takes care of it for you.
4739 * The system is no longer so grossly inefficient when compiling code
4740 involving vectors implemented as general (not simple) vectors (VECTOR T),
4741 so code which dares to use VECTOR-PUSH-EXTEND and FILL-POINTER, or
4742 which dares to use the various sequence functions on non-simple
4743 vectors, takes less of a performance hit.
4744 * There is now a primitive type predicate VECTOR-T-P
4745 to test for the (VECTOR T) type, so that e.g.
4746 (DEFUN FOO (V) (DECLARE (TYPE (VECTOR T) V)) (AREF V 3))
4747 can now be compiled with some semblance of efficiency. (The old code
4748 turned the type declaration into a full call to %TYPEP at runtime!)
4749 * AREF on (VECTOR T) is still not fast, since it's still compiled
4750 as a full call to SB-KERNEL:DATA-VECTOR-REF, but at least the
4751 ETYPECASE used in DATA-VECTOR-REF is now compiled reasonably
4752 efficiently. (The old version made full calls to SUBTYPEP at runtime!)
4753 * (MAKE-ARRAY 12 :FILL-POINTER T) is now executed less inefficiently,
4754 without making full calls to SUBTYPEP at runtime.
4755 (Some analogous efficiency issues for non-simple vectors specialized to
4756 element types other than T, or for non-simple multidimensional arrays,
4757 have not been addressed. They could almost certainly be handled the
4758 same way if anyone is motivated to do so.)
4759 * The changes in array handling break binary compatibility, so
4760 *BACKEND-FASL-FILE-VERSION* has been bumped to 4.
4761 * (TYPEP (MAKE-ARRAY 12 :FILL-POINTER 4) 'VECTOR) now returns (VALUES T)
4762 instead of (VALUES T T).
4763 * By following the instructions that Dan Barlow posted to sbcl-devel
4764 on 2 July 2000, I was able to enable primitive dynamic object
4765 file loading code for Linux. The full-blown CMU CL LOAD-FOREIGN
4766 functionality is not implemented (since it calls ld to resolve
4767 library references automatically, requiring RUN-PROGRAM for its
4768 implementation), but a simpler SB-EXT:LOAD-1-FOREIGN (which doesn't
4769 try to resolve library references) is now supported.
4770 * The system now flushes the standard output streams when it terminates,
4771 unless QUIT is used with the RECKLESSLY-P option set. It also flushes
4772 them at several other probably-convenient times, e.g. in each pass of
4773 the toplevel read-eval-print loop, and after evaluating a form given
4774 as an "--eval" command-line option. (These changes were motivated by a
4775 discussion of stream flushing issues on cmucl-imp in August 2000.)
4776 * The source transform for TYPEP of array types no longer assumes
4777 that an array whose element type is a not-yet-defined type
4778 is implemented as an array of T, but instead punts, so that the
4779 type will be interpreted at runtime.
4780 * There is now some support for cross-compiling in make.sh: each of
4781 the phases of make.sh has its own script. (This should be transparent
4782 to people doing ordinary, non-cross-compile builds.)
4783 * Since my laptop doesn't have hundreds of megabytes of memory like
4784 my desktop machine, I became more motivated to do some items on
4785 my to-do list in order to reduce the size of the system a little:
4786 ** Arrange for various needed-only-at-cold-init things to be
4787 uninterned after cold init. To support this, those things have
4788 been renamed from FOO and *FOO* to !FOO and *!FOO* (i.e., all
4789 symbols with such names are now uninterned after cold init).
4790 ** Bind SB!C::*TOP-LEVEL-LAMBDA-MAX* to a nonzero value when building
4791 fasl files for cold load.
4792 ** Remove the old compiler structure pooling code (which used to
4793 be conditional on the target feature :SB-ALLOC) completely.
4794 ** Redo the representation of some data in cold init to be more compact.
4795 (I also looked into supporting byte compiled code at bootstrap time,
4796 which would probably reduce the size of the system a lot, but that
4797 looked too complicated, so I punted for now.)
4798 * The maximum signal nesting depth in the src/runtime/ support code has
4799 been reduced from 4096 to 256. (I don't know any reason for the very
4800 large old value. If the new smaller value turns out to break something,
4801 I'll probably just bump it back up.)
4802 * PPRINT-LOGICAL-BLOCK is now pickier about the types of its arguments,
4804 * Many, many bugs reported by Peter Van Eynde have been added to
4805 the BUGS list; some have even been fixed.
4806 * While enabling dynamic object file loading, I tried to make the
4807 code easier to understand, renaming various functions and variables
4808 with less ambiguous names, and changing some function calling
4809 conventions to be Lispier (e.g. returning NIL instead of 0 for failure).
4810 * While trying to figure out how to do the OpenBSD port, I tried to
4811 clean up some of the code in src/runtime/. In particular, I dropped
4812 support for non-POSIX signal handling, added various comments,
4813 tweaked the code to reduce the number of compilation warnings, and
4814 renamed some files to increase consistency.
4815 * To support the new automatic configuration functionality in make.sh,
4816 the source file target-features.lisp-expr has been replaced with the
4817 source file base-target-features.lisp-expr and the machine-generated
4818 file local-target-features.lisp-expr.
4819 * fixed a stupid quoting error in make.sh so that using CMU CL
4820 "lisp -batch" as cross-compilation host works again
4822 changes in sbcl-0.6.6 relative to sbcl-0.6.5:
4824 * DESCRIBE no longer tries to call itself recursively to describe
4825 bound/fbound values, so that it no longer fails on symbols which are
4826 bound to themselves (like keywords, T, and NIL).
4827 * DESCRIBE now works on generic functions.
4828 * The printer now prints less-screwed-up representations of closures
4829 (not naively trying to bogusly use the %FUNCTION-NAME accessor on them).
4830 * A private symbol is used instead of the :EMPTY keyword previously
4831 used to mark empty slots in hash tables. Thus
4832 (DEFVAR *HT* (MAKE-HASH-TABLE))
4833 (SETF (GETHASH :EMPTY *HT*) :EMPTY)
4834 (MAPHASH (LAMBDA (K V) (FORMAT T "~&~S ~S~%" K V)))
4835 now does what ANSI says that it should. (You can still get
4836 similar noncompliant behavior if bang on the hash table
4837 implementation with all the symbols you get back from
4838 DO-ALL-SYMBOLS, but at least that's a little harder to do.)
4839 This breaks binary compatibility, since tests for equality to
4840 :EMPTY are wired into things like the macroexpansion of
4841 WITH-HASH-TABLE-ITERATOR in FASL files produced by earlier
4843 * There's now a minimal placeholder implementation for CL:STEP,
4844 as required by ANSI.
4845 * An obscure bug in the interaction of the normal compiler, the byte
4846 compiler, inlining, and structure predicates has been patched
4847 by setting the flags for the DEFTRANSFORM of %INSTANCE-TYPEP as
4848 :WHEN :BOTH (as per Raymond Toy's suggestion on the cmucl-imp@cons.org
4850 * Missing ordinary arguments in a macro call are now detected even
4851 when the macro lambda list contains &KEY or &REST.
4852 * The debugger no longer complains about encountering the top of the
4853 stack when you type "FRAME 0" to explicitly instruct it to go to
4854 the top of the stack. And it now prints the frame you request even
4855 if it's the current frame (instead of saying "You are here.").
4856 * As specified by ANSI, the system now always prints keywords
4857 as #\: followed by SYMBOL-NAME, even when *PACKAGE* is the
4859 * The default initial SIZE of HASH-TABLEs is now smaller.
4860 * Type information from CLOS class dispatch is now propagated
4861 into DEFMETHOD bodies, so that e.g.
4862 (DEFMETHOD FOO ((X SINGLE-FLOAT))
4864 is now basically equivalent to
4865 (DEFMETHOD FOO ((X SINGLE-FLOAT))
4866 (DECLARE (TYPE SINGLE-FLOAT X))
4868 and the compiler can compile (+ X 123.0) as a SINGLE-FLOAT-only
4869 operation, without having to do run-time type dispatch.
4870 * The macroexpansion of DEFMETHOD has been tweaked so that it has
4871 reasonable behavior when arguments are declared IGNORE or IGNORABLE.
4872 * Since I don't seem to be making big file reorganizations very often
4873 any more (and since my archive of sbcl-x.y.zv.tar.bz2 snapshots
4874 is overflowing my ability to conveniently back them up), I've finally
4875 checked the system into CVS. (The CVS repository is on my home system,
4876 not at SourceForge -- putting it on SourceForge might come later.)
4877 * SB-EXT:*GC-NOTIFY-STREAM* has been added, to control where the
4878 high-level GC-NOTIFY-FOO functions send their output. (There's
4879 still very little control of where low-level verbose GC functions
4880 send their output.) The SB-EXT:*GC-VERBOSE* variable now controls
4881 less than it used to -- the GC-NOTIFY-FOO functions are now under
4882 the control of *GC-NOTIFY-STREAM*, not *GC-VERBOSE*.
4883 * The system now stores the version string (LISP-IMPLEMENTATION-VERSION)
4884 in only one place in the source code, and propagates it automatically
4885 everywhere that it's needed. Thus e.g. when I bump the version from
4886 0.6.6 to 0.6.7, I'll only need to modify the sources in one place.
4887 * The C source files now include boilerplate legalese and documentation
4888 at the head of each file (just as the Lisp source files already did).
4889 * At Dan Barlow's suggestion, the hyperlink from the SBCL website
4890 to his page will be replaced with a link to his new CLiki service.
4892 changes in sbcl-0.6.5 relative to sbcl-0.6.4:
4894 * Raymond Wiker's patches to port the system to FreeBSD have been merged.
4895 * The build process now looks for GNU make under the default name "gmake",
4896 instead of "make" as it used to. If GNU make is not available as "gmake"
4897 on your system, you can change this default behavior by setting the
4898 GNUMAKE environment variable.
4899 * Replace #+SB-DOC with #!+SB-DOC in seq.lisp so that the system
4900 can build without error under CMU CL.
4902 changes in sbcl-0.6.4 relative to sbcl-0.6.3:
4904 * There is now a partial SBCL user manual (with some new text and some
4905 text cribbed from the CMU CL manual).
4906 * The beginnings of a profiler have been added (starting with the
4907 CMU CL profiler and simplifying and cleaning up). Eventually the
4908 main interface should be through the TRACE macro, but for now,
4909 it's still accessed through vaguely CMU-CL-style functions and macros
4910 exported from the package SB-PROFILE.
4911 * Some problems left over from porting CMU CL to the new
4912 cross-compilation bootstrap process have been cleaned up:
4913 ** DISASSEMBLE now works. (There was a problem in using DEFMACRO
4914 instead of SB!XC:DEFMACRO, compounded by an oversight on my
4915 part when getting rid of the compiler *BACKEND* stuff.)
4916 ** The value of *NULL-TYPE* was screwed up, because it was
4917 being initialized before the type system knew the final
4918 definition of the 'NULL type. This screwed up several key
4919 optimizations in the compiler, causing inefficiency in all sorts
4920 of places. (I found it because I wanted to understand why
4921 GET-INTERNAL-RUN-TIME was consing.)
4922 * fixed a bug in DEFGENERIC which was causing it to overwrite preexisting
4923 PROCLAIM FTYPE information. Unfortunately this broke binary
4924 compatibility again, since now the forms output by DEFGENERIC
4925 to refer to functions which didn't exist in 0.6.3.
4926 * added declarations so that SB-PCL::USE-CACHING-DFUN-P
4927 can use the new (as of 0.6.3) transform for SOME into MAP into
4929 * changed (MOD 1000000) type declarations for Linux timeval.tv_usec slot
4930 values to (INTEGER 0 1000000), so that the time code will no longer
4931 occasionally get blown up by Linux returning 1000000 microseconds
4932 * PRINT-UNREADABLE-OBJECT has been tweaked to make the spacing of
4933 its output conform to the ANSI spec. (Alas, this makes its output
4934 uglier in the :TYPE T :IDENTITY NIL case, but them's the breaks.)
4935 * A full call to MAP NIL with a single sequence argument no longer conses.
4936 * fixes to problems pointed out by Martin Atzmueller:
4937 * The manual page no longer talks about multiprocessing as though
4938 it were currently supported.
4939 * The ILISP support patches have been removed from the distribution,
4940 because as of version 5.10.1, ILISP now supports SBCL without us
4941 having to maintain patches.
4942 * added a modified version of Raymond Toy's recent CMU CL patch for
4943 EQUALP comparison of HASH-TABLE
4945 changes in sbcl-0.6.3 relative to sbcl-0.6.2:
4947 * The system still can't cross-compile itself with
4948 *TYPE-SYSTEM-INITIALIZED* (and all the consistency checks that
4949 entails), but at least it can compile more of itself that way
4950 than it used to be able to, and various buglets which were uncovered
4951 by trying to cross-compile itself that way have now been fixed.
4952 * This release breaks binary compatibility again. This time
4953 at least I've incremented the FASL file format version to 2, so that the
4954 problem can be detected reliably instead of just causing weird errors.
4955 * various new style warnings:
4956 ** using DEFUN, DEFMETHOD, or DEFGENERIC to overwrite an old definition
4957 ** using the deprecated EVAL/LOAD/COMPILE situation names in EVAL-WHEN
4958 ** using the lexical binding of a variable named in the *FOO* style
4959 * DESCRIBE has been substantially rewritten. It now calls DESCRIBE-OBJECT
4960 as specified by ANSI.
4961 * *RANDOM-STATE* is no longer automatically initialized from
4962 (GET-UNIVERSAL-TIME), but instead from a constant seed. Thus, the
4963 default behavior of the system is to repeat its behavior every time
4964 it's run. If you'd like to change this behavior, you can always
4965 explicitly set the seed from (GET-UNIVERSAL-TIME); whereas under the
4966 old convention there was no comparably easy way to get the system to
4967 repeat its behavior every time it was run.
4968 * Support for the pre-CLTL2 interpretation of FUNCTION declarations as
4969 FTYPE declarations has been removed, in favor of their ANSI
4970 interpretation as TYPE FUNCTION declarations. (See p. 228 of CLTL2.)
4971 * The quantifiers SOME, EVERY, NOTANY, and NOTEVERY no longer cons when
4972 the types of their sequence arguments can be determined at compile time.
4973 This is done through a new open code expansion for MAP which eliminates
4974 consing for (MAP NIL ..), and reduces consing otherwise, when sequence
4975 argument types can be determined at compile time.
4976 * The optimizer now transforms COERCE into an identity operation when it
4977 can prove that the coerced object is already of the correct type. (This
4978 can be a win for machine generated code, including the output of other
4979 optimization transforms, such as the MAP transform above.)
4980 * Credit information has been moved from source file headers into CREDITS.
4981 * Source file headers have been made more standard.
4982 * The CASE macro now compiles without complaining even when it has
4985 changes in sbcl-0.6.2 relative to sbcl-0.6.1:
4987 * (Note that the way that the PCL macroexpansions were rewritten
4988 to accommodate the change in DEFGENERIC below breaks binary
4989 compatibility. That is, fasl files compiled under sbcl-0.6.1 may
4990 not run under sbcl-0.6.2. Once we get out of alpha releases,
4991 i.e. hit release 1.0.0, we'll probably try to maintain binary
4992 compatibility between maintenance releases, e.g. between sbcl-1.4.3
4993 and sbcl-1.4.4. Until then, however, it might be fairly common
4994 for maintenance releases to break binary compatibility.)
4995 * A bug in the calculation of WARNINGS-P and FAILURE-P in COMPILE-FILE
4997 * The reporting of unhandled signals has been changed to print some
4998 explanatory text as well as the report form. (Previously only
4999 the report form was printed.)
5000 * The macroexpansion for DEFGENERIC now DECLAIMs the function that
5001 it defines, so that the compiler no longer issues undefined function
5002 warnings for compiled-but-not-yet-loaded generic functions.
5003 * The CLTL-style "LISP" and "USER" nicknames for the "COMMON-LISP"
5004 and "COMMON-LISP-USER" packages have been removed. Now only the "CL"
5005 and "CL-USER" standard nicknames from the "11.1.2 Standardized Packages"
5006 section of the ANSI spec are supported.
5007 * The "" nickname for the "KEYWORD" package has been removed.
5008 The reader still handles symbol tokens which begin with a package marker
5009 as keywords, but it doesn't expose its mechanism for doing so in the
5010 (PACKAGE-NICKNAMES (FIND-PACKAGE "KEYWORD")) list.
5011 * The system now issues STYLE-WARNINGs for contradictory TYPE
5012 proclamations. (Warnings for contradictory FTYPE proclamations would
5013 be nice too, but those can't be done usefully unless the type system
5014 is made smarter about FUNCTION types.)
5015 * The names of source files "*host-*.lisp" and "*target-*.lisp" have been
5016 systematized, so that "*target-*.lisp is supposed to exist only on the
5017 target and imply that there's a related file which exists on the
5018 host, and *host-*.lisp is supposed to exist only on the host and imply
5019 that there's a related file which exists on the target. This involves a
5020 lot of renaming. Hopefully the acute confusion caused by the renaming
5021 will be justified by the reduction in chronic confusion..
5022 ** runtime-type.lisp -> early-target-type.lisp
5023 ** target-type.lisp -> late-target-type.lisp
5024 ** early-host-format.lisp -> early-format.lisp
5025 ** late-host-format.lisp -> late-format.lisp
5026 ** host-error.lisp -> misc-error.lisp
5027 ** early-error.lisp -> early-target-error.lisp
5028 ** late-error.lisp -> late-target-error.lisp
5029 ** host-defboot.lisp -> early-defboot.lisp
5030 ** code/misc.lisp -> code/target-misc.lisp
5031 ** code/host-misc.lisp -> code/misc.lisp
5032 ** code/numbers.lisp -> code/target-numbers.lisp
5033 ** code/early-numbers.lisp -> numbers.lisp
5034 ** early-host-type.lisp -> early-type.lisp
5035 ** late-host-type.lisp -> late-type.lisp
5036 ** host-typep.lisp -> typep.lisp
5037 ** load.lisp -> target-load.lisp
5038 ** host-load.lisp -> load.lisp
5039 ** host-disassem.lisp -> disassem.lisp
5040 ** host-insts.lisp -> insts.lisp
5041 ** byte-comp.lisp -> target-byte-comp.lisp
5042 ** host-byte-comp.lisp -> byte-comp.lisp
5043 ** host-signal.lisp -> signal.lisp
5044 ** host-defstruct.lisp -> defstruct.lisp
5045 ** late-target-type.lisp -> deftypes-for-target.lisp
5046 Furthermore, several other previously target-only files foo.lisp (e.g.
5047 hash-table.lisp and random.lisp) have been split into a target-and-host
5048 foo.lisp file and a target-only target-foo.lisp file, with their key type
5049 definitions in the target-and-host part, so that the cross-compiler will
5050 know more about target types.
5051 * DEFSTRUCT BACKEND, and the BACKEND-valued *BACKEND* variable, have
5052 gone away. In their place are various *BACKEND-FOO* variables
5053 corresponding to the slots of the old structure.
5054 * A bug which caused the SB-COLD bootstrap-time package to be propagated
5055 into the target SBCL has been fixed.
5056 * The chill.lisp system for loading cold code into a running SBCL
5058 * Support for the CMU CL "scavenger hook" extension has been removed.
5059 (It was undocumented and unused in the CMU CL sources that SBCL was
5060 derived from, and stale in sbcl-0.6.1.)
5061 * Various errors in the cross-compiler type system were detected
5062 by running the cross-compiler with *TYPE-SYSTEM-INITIALIZED*
5063 (enabling various consistency checks). Many of them were fixed,
5064 but some hard problems remain, so the compiler is back to
5065 running without *TYPE-SYSTEM-INITIALIZED* for now.
5066 * As part of the cross-compiler type system cleanup, I implemented
5067 DEF!TYPE and got rid of early-ugly-duplicates.lisp.
5068 * I have started adding UNCROSS calls throughout the type system
5069 and the INFO database. (Thus perhaps eventually the blanket UNCROSS
5070 on cross-compiler input files will be able to go away, and various
5072 * CONSTANTP now returns true for quoted forms (as explicitly required
5075 changes in sbcl-0.6.1 relative to sbcl-0.6.0:
5077 * changed build optimization from (SAFETY 1) to (SAFETY 3) as a short-term
5078 fix for various type-unsafety bugs, e.g. failures with (LENGTH 123) and
5079 (MAKE-LIST -1). In the longer term, it ought to become true
5080 that declarations are assertions even at SAFETY 1. For now, it's not
5081 quite true even at SAFETY 3, but it's at least more nearly true..
5082 (Note that this change seems to increases the size of the system by
5083 O(5%) and to decrease the speed of the compiler by 20% or more.)
5084 * changed ALIEN printing to be much more abbreviated, as a short-term fix
5085 for the problem of printing dozens of lines of distracting information
5086 about low-level system machinery as part of the top stack frame
5087 on entry to the debugger when an undefined function was called.
5088 * tweaked the debugger's use of WITH-STANDARD-IO-SYNTAX so that *PACKAGE*
5089 is not reset to COMMON-LISP-USER.
5090 * Compilation of stuff related to dyncount.lisp has been made conditional
5091 on the :SB-DYNCOUNT target feature, so that the ordinary core system is
5092 smaller. The various dyncount-related symbols have been moved into
5093 a new "SB-DYNCOUNT" package.
5094 * tty-inspect.lisp has been renamed to inspect.lisp.
5095 * unix-glibc2.lisp has been renamed to unix.lisp, and the :GLIBC2
5096 feature has gone away. (When we eventually port to other flavors of
5097 libc and/or Unix, we'll try to make the differences between flavors
5098 invisible at the user level.)
5099 * Various other *FEATURES* tags, and/or their associated conditionals,
5100 have been removed if obsolescent, or given better documentation, or
5101 sometimes given more-mnemonic names.
5103 changes in sbcl-0.6.0 relative to sbcl-0.5.0:
5105 * tidied up "make.sh" script
5106 * tidied up system directory structure
5107 * better "clean.sh" behavior
5108 * added doc/FOR-CMUCL-DEVELOPERS
5109 * many many small tweaks to output format, e.g. removing possibly-confusing
5110 trailing #\. character in DESCRIBE-INSTANCE
5111 * (EQUALP #\A 'A) no longer signals an error.
5112 * new hashing code, including EQUALP hashing
5113 * tidied up Lisp initialization and toplevel
5114 * initialization files (e.g. /etc/sbclrc and $HOME/.sbclrc)
5115 * command line argument processing
5116 * added POSIX-GETENV function to deal with Unix-ish environment variables
5117 * more-Unixy handling of *STANDARD-INPUT* and other Lisp streams, e.g.
5118 terminating SBCL on EOF
5119 * non-verbose GC by default
5120 * There is no more "sbcl" shell script; the sbcl file is now the C
5121 runtime executable (just like CMU CL).
5122 * removed some unused fops, e.g. FOP-UNIFORM-VECTOR, FOP-CHARACTER, and
5124 * tweaked debug-info.lisp and debug-int.lisp to make the debugger store
5125 symbol and package information as Lisp native symbol and package objects
5126 instead of strings naming symbols and strings naming packages. This way,
5127 whenever packages are renamed (as in warm init), debug information is
5128 transformed along with everything else.
5129 * tweaked the optimization policy declarations which control the building
5130 of SBCL itself. Now, among other things, the system no longer saves
5131 source location debugging information. (This helps two problems at once
5132 by reducing SBCL size and by keeping SBCL from trying to look for its
5133 sources -- which may not exist -- when reporting errors.)
5134 * added src/cold/chill.lisp, to let SBCL read its own cold sources for
5135 debugging and testing purposes
5136 * cleaned up printing, making the printer call PRINT-OBJECT for
5137 instances, and using PRINT-UNREADABLE-OBJECT for most PRINT-OBJECT
5138 methods, giving nearly-ANSI behavior
5139 * converted almost all special variables to use *FOO* naming convention
5140 * deleted PARSE-TIME functionality, since it can be done portably
5141 * moved some files out of cold init into warm init
5142 * deleted DEFUN UNDEFINED-VALUE, replaced (UNDEFINED-VALUE) forms
5144 * regularized formatting of source files
5145 * added an install.sh script
5146 * fixed ridiculous memory usage of cross-compiler by making
5147 compiler/alloc.lisp not try to do pooling unless it can hook
5148 itself into the GC of the cross-compilation host. Now the system
5149 builds nicely on my old laptop.
5150 * added :SB-ALLOC in target-features.lisp-expr
5151 * deleted mention of :ANSI-DOC from target-features.lisp-expr (since it
5152 was not implemented)
5153 * re-did condition handling and note reporting in the compiler. Notes
5154 are no longer handled by signalling conditions. Style warnings
5155 and warnings are handled more correctly and reported in such a way
5156 that it's easy to find one or the other in your output (so that you
5157 can e.g. figure out which of many problems caused COMPILE-FILE to
5159 * changed the severity of several compiler warnings from full WARNING
5160 to STYLE-WARNING in order to conform with the ANSI spec; also changed
5161 compiler note reporting so that it doesn't use the condition system
5162 at all (and hence affects neither FAILURE-P nor WARNINGS-P in the
5163 COMPILE-FILE command)
5164 * made PROCLAIM and DECLAIM conform to ANSI. PROCLAIM is now an ordinary
5165 function. As a consequence, START-BLOCK and END-BLOCK declarations are
5166 no longer supported, since their implementation was deeply intertwingled
5167 with the magical, non-ANSI treatment that PROCLAIM received in CMU CL.
5168 * removed bogus "support" for compiler macros named (SETF FOO), and
5169 removed the compiler macro for SETF INFO (but only after making a fool
5170 of myself on the cmucl-imp mailing list by posting a bogus patch for
5171 DEFINE-COMPILER-MACRO..)
5172 * Compiled files containing forms which have side effects on the Lisp
5173 reader (such as DEFPACKAGE forms) are now handled more correctly.
5174 (Compiler queuing of top level lambdas has been suppressed by setting
5175 *TOP-LEVEL-LAMBDA-MAX* to 0. )
5176 * deleted various currently-unused source files, e.g. gengc.lisp. They
5177 may be added back at some point e.g. when porting to other architectures,
5178 but until they are it's distracting to distribute them and to try to
5180 * deleted "UNCROSS couldn't recurse through.." style warnings, since
5181 there were so many of them they're just distractions, and UNCROSS is
5182 known to be able to handle the current sources
5183 * moved PROFILE functionality into TRACE, so that it will be clear
5184 how the wrapping and unwrapping of functions when you profile them
5185 interacts with the wrapping and unwrapping of functions when you
5186 trace them. (Actually, the functionality isn't there yet, but at least
5187 the interface specification is there. Hopefully, the functionality will
5188 arrive with some future maintenance release.)
5189 * removed host-oops.lisp
5190 * changed signature of QUIT function to allow UNIX-CODE argument
5191 * fixed READ-SEQUENCE bug
5192 * tweaked verbose GC output so that it looks more like the progress
5193 output that ANSI specifies for functions like LOAD
5194 * set up the system on sourceforge.com, with home pages, mailing lists, etc.
5195 * added <http://sbcl.sourceforge.com> to the banner information printed by