3 Bugs can be reported on the help mailing list
4 sbcl-help@lists.sourceforge.net
5 or on the development mailing list
6 sbcl-devel@lists.sourceforge.net
8 Please include enough information in a bug report that someone reading
9 it can reproduce the problem, i.e. don't write
10 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
11 PRINT-OBJECT doesn't seem to work with *PRINT-LENGTH*. Is this a bug?
13 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
14 In sbcl-1.2.3 running under OpenBSD 4.5 on my Alpha box, when
15 I compile and load the file
16 (DEFSTRUCT (FOO (:PRINT-OBJECT (LAMBDA (X Y)
17 (LET ((*PRINT-LENGTH* 4))
20 then at the command line type
22 the program loops endlessly instead of printing the object.
27 There is also some information on bugs in the manual page and
28 in the TODO file. Eventually more such information may move here.
30 The gaps in the number sequence belong to old bug descriptions which
31 have gone away (typically because they were fixed, but sometimes for
32 other reasons, e.g. because they were moved elsewhere).
35 KNOWN BUGS OF NO SPECIAL CLASS:
38 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
39 instead of just punting when a contradictory structure definition
40 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
41 changes their layout, you probably have to rebuild your entire
42 program, even if you know or guess enough about the internals of
43 SBCL to wager that this (undefined in ANSI) operation would be safe.
46 ANSI specifies that a type mismatch in a structure slot
47 initialization value should not cause a warning.
49 This one might not be fixed for a while because while we're big
50 believers in ANSI compatibility and all, (1) there's no obvious
51 simple way to do it (short of disabling all warnings for type
52 mismatches everywhere), and (2) there's a good portable
53 workaround, and (3) by their own reasoning, it looks as though
54 ANSI may have gotten it wrong. ANSI justifies this specification
56 The restriction against issuing a warning for type mismatches
57 between a slot-initform and the corresponding slot's :TYPE
58 option is necessary because a slot-initform must be specified
59 in order to specify slot options; in some cases, no suitable
61 However, in SBCL (as in CMU CL or, for that matter, any compiler
62 which really understands Common Lisp types) a suitable default
63 does exist, in all cases, because the compiler understands the
64 concept of functions which never return (i.e. has return type NIL).
65 Thus, as a portable workaround, you can use a call to some
66 known-never-to-return function as the default. E.g.
68 (BAR (ERROR "missing :BAR argument")
69 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
71 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
72 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
73 (ERROR "missing required argument"))
75 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
76 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
77 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
78 Such code should compile without complaint and work correctly either
79 on SBCL or on any other completely compliant Common Lisp system.
82 bogus warnings about undefined functions for magic functions like
83 SB!C::%%DEFUN and SB!C::%DEFCONSTANT when cross-compiling files
84 like src/code/float.lisp. Fixing this will probably require
85 straightening out enough bootstrap consistency issues that
86 the cross-compiler can run with *TYPE-SYSTEM-INITIALIZED*.
87 Instead, the cross-compiler runs in a slightly flaky state
88 which is sane enough to compile SBCL itself, but which is
89 also unstable in several ways, including its inability
90 to really grok function declarations.
92 As of sbcl-0.7.5, sbcl's cross-compiler does run with
93 *TYPE-SYSTEM-INITIALIZED*; however, this bug remains.
96 The "compiling top-level form:" output ought to be condensed.
97 Perhaps any number of such consecutive lines ought to turn into a
98 single "compiling top-level forms:" line.
101 The way that the compiler munges types with arguments together
102 with types with no arguments (in e.g. TYPE-EXPAND) leads to
103 weirdness visible to the user:
104 (DEFTYPE FOO () 'FIXNUM)
106 (TYPEP 11 '(FOO)) => T, which seems weird
107 (TYPEP 11 'FIXNUM) => T
108 (TYPEP 11 '(FIXNUM)) signals an error, as it should
109 The situation is complicated by the presence of Common Lisp types
110 like UNSIGNED-BYTE (which can either be used in list form or alone)
111 so I'm not 100% sure that the behavior above is actually illegal.
112 But I'm 90+% sure, and the following related behavior,
114 treating the bare symbol AND as equivalent to '(AND), is specifically
115 forbidden (by the ANSI specification of the AND type).
118 It would be nice if the
120 (during macroexpansion)
121 said what macroexpansion was at fault, e.g.
123 (during macroexpansion of IN-PACKAGE,
124 during macroexpansion of DEFFOO)
127 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
128 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
129 (Also, when this is fixed, we can enable the code in PROCLAIM which
130 checks for incompatible FTYPE redeclarations.)
133 (I *think* this is a bug. It certainly seems like strange behavior. But
134 the ANSI spec is scary, dark, and deep.. -- WHN)
135 (FORMAT NIL "~,1G" 1.4) => "1. "
136 (FORMAT NIL "~3,1G" 1.4) => "1. "
139 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
141 (setf (find-class 'ccc1) (find-class 'ccc))
142 (defmethod zut ((c ccc1)) 123)
143 In sbcl-0.7.1.13, this gives an error,
144 There is no class named CCC1.
145 DTC's recommended workaround from the mailing list 3 Mar 2000:
146 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
149 Sometimes (SB-EXT:QUIT) fails with
150 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
151 Process inferior-lisp exited abnormally with code 1
152 I haven't noticed a repeatable case of this yet.
155 The printer doesn't report closures very well. This is true in
159 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
160 It would be nice to make closures have a settable name slot,
161 and make things like DEFSTRUCT and FLET, which create closures,
162 set helpful values into this slot.
165 And as long as we're wishing, it would be awfully nice if INSPECT could
166 also report on closures, telling about the values of the bound variables.
169 The compiler assumes that any time a function of declared FTYPE
170 doesn't signal an error, its arguments were of the declared type.
171 E.g. compiling and loading
172 (DECLAIM (OPTIMIZE (SAFETY 3)))
173 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
175 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
177 (COND ((> (FACTORIAL X) 1.0E6)
178 (FORMAT T "too big~%"))
180 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
182 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
185 will cause the INTEGERP case to be selected, giving bogus output a la
187 This violates the "declarations are assertions" principle.
188 According to the ANSI spec, in the section "System Class FUNCTION",
189 this is a case of "lying to the compiler", but the lying is done
190 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
191 not by the unexpectedly general definition of FACTORIAL. In any case,
192 "declarations are assertions" means that lying to the compiler should
193 cause an error to be signalled, and should not cause a bogus
194 result to be returned. Thus, the compiler should not assume
195 that arbitrary functions check their argument types. (It might
196 make sense to add another flag (CHECKED?) to DEFKNOWN to
197 identify functions which *do* check their argument types.)
198 (Also, verify that the compiler handles declared function
199 return types as assertions.)
202 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
203 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
205 (DECLARE (TYPE INDEXOID X))
206 (THE (VALUES INDEXOID)
208 where the implementation of the type check in function FOO
209 includes a full call to %TYPEP. There are also some fundamental problems
210 with the interpretation of VALUES types (inherited from CMU CL, and
211 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
212 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
215 The definitions of SIGCONTEXT-FLOAT-REGISTER and
216 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
217 supported on FreeBSD because the floating point state is not saved,
218 but at least as of FreeBSD 4.0, the floating point state *is* saved,
219 so they could be supported after all. Very likely
220 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
223 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
224 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
225 issue with 'union, 'and, 'or etc. So even though according to the
226 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
227 (and now SBCL) interpret them as legal types.
230 a slew of floating-point-related errors reported by Peter Van Eynde
232 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
233 should probably be 1.4012985e-45. In SBCL,
234 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
235 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
236 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
237 and LEAST-NEGATIVE-LONG-FLOAT.
238 c: Many expressions generate floating infinity on x86/Linux:
243 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
244 on x86/Linux generates the infinities instead. That might or
245 might not be conforming behavior, but it's also inconsistent,
246 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
247 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
248 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
250 d: (in section12.erg) various forms a la
251 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
252 don't give the right behavior.
255 type safety errors reported by Peter Van Eynde July 25, 2000:
256 a: (COERCE (QUOTE (A B C)) (QUOTE (VECTOR * 4)))
258 In general lengths of array type specifications aren't
259 checked by COERCE, so it fails when the spec is
260 (VECTOR 4), (STRING 2), (SIMPLE-BIT-VECTOR 3), or whatever.
261 b: CONCATENATE has the same problem of not checking the length
262 of specified output array types. MAKE-SEQUENCE and MAP and
263 MERGE also have the same problem.
264 c: (COERCE 'AND 'FUNCTION) returns something related to
265 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
266 h: (MAKE-CONCATENATED-STREAM (MAKE-STRING-OUTPUT-STREAM))
267 should signal TYPE-ERROR.
268 i: MAKE-TWO-WAY-STREAM doesn't check that its arguments can
269 be used for input and output as needed. It should fail with
270 TYPE-ERROR when handed e.g. the results of
271 MAKE-STRING-INPUT-STREAM or MAKE-STRING-OUTPUT-STREAM in
272 the inappropriate positions, but doesn't.
273 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
274 not a binary input stream, but instead cheerfully reads from
275 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
278 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
279 a: (DEFCLASS FOO () (A B A)) should signal a PROGRAM-ERROR, and
281 b: (DEFCLASS FOO () (A B A) (:DEFAULT-INITARGS X A X B)) should
282 signal a PROGRAM-ERROR, and doesn't.
283 c: (DEFCLASS FOO07 NIL ((A :ALLOCATION :CLASS :ALLOCATION :CLASS))),
284 and other DEFCLASS forms with duplicate specifications in their
285 slots, should signal a PROGRAM-ERROR, and doesn't.
286 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
287 causes a COMPILER-ERROR.
290 SYMBOL-MACROLET bugs reported by Peter Van Eynde July 25, 2000:
291 c: SYMBOL-MACROLET should signal PROGRAM-ERROR if something
292 it binds is declared SPECIAL inside.
295 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
297 (DEFGENERIC FOO02 (X))
298 (DEFMETHOD FOO02 ((X NUMBER)) T)
299 (LET ((M (FIND-METHOD (FUNCTION FOO02)
301 (LIST (FIND-CLASS (QUOTE NUMBER))))))
302 (REMOVE-METHOD (FUNCTION FOO02) M)
303 (DEFGENERIC FOO03 (X))
304 (ADD-METHOD (FUNCTION FOO03) M)))
305 should give an error, but SBCL allows it.
306 b: READ should probably return READER-ERROR, not the bare
307 arithmetic error, when input a la "1/0" or "1e1000" causes
311 It has been reported (e.g. by Peter Van Eynde) that there are
312 several metaobject protocol "errors". (In order to fix them, we might
313 need to document exactly what metaobject protocol specification
314 we're following -- the current code is just inherited from PCL.)
317 The implementation of #'+ returns its single argument without
318 type checking, e.g. (+ "illegal") => "illegal".
321 The debugger LIST-LOCATIONS command doesn't work properly.
324 Compiling and loading
325 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
327 then requesting a BACKTRACE at the debugger prompt gives no information
328 about where in the user program the problem occurred.
331 The compiler is supposed to do type inference well enough that
334 ((SIMPLE-ARRAY SINGLE-FLOAT)
336 (DECLARE (TYPE (SIMPLE-ARRAY SINGLE-FLOAT) X))
339 is redundant. However, as reported by Juan Jose Garcia Ripoll for
340 CMU CL, it sometimes doesn't. Adding declarations is a pretty good
341 workaround for the problem for now, but can't be done by the TYPECASE
342 macros themselves, since it's too hard for the macro to detect
343 assignments to the variable within the clause.
344 Note: The compiler *is* smart enough to do the type inference in
345 many cases. This case, derived from a couple of MACROEXPAND-1
346 calls on Ripoll's original test case,
348 (DECLARE (OPTIMIZE SPEED (SAFETY 0)))
349 (COND ((TYPEP A '(SIMPLE-ARRAY SINGLE-FLOAT)) NIL
350 (LET ((LENGTH (ARRAY-TOTAL-SIZE A)))
351 (LET ((I 0) (G2554 LENGTH))
352 (DECLARE (TYPE REAL G2554) (TYPE REAL I))
355 (WHEN (>= I G2554) (GO SB-LOOP::END-LOOP))
356 (SETF (ROW-MAJOR-AREF A I) (- (ROW-MAJOR-AREF A I)))
357 (GO SB-LOOP::NEXT-LOOP)
358 SB-LOOP::END-LOOP))))))
359 demonstrates the problem; but the problem goes away if the TAGBODY
360 and GO forms are removed (leaving the SETF in ordinary, non-looping
361 code), or if the TAGBODY and GO forms are retained, but the
362 assigned value becomes 0.0 instead of (- (ROW-MAJOR-AREF A I)).
365 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
366 I am looking into this problem that showed up on the cmucl-help
367 list. It seems to me that the "implementation specific environment
368 hacking functions" found in pcl/walker.lisp are completely messed
369 up. The good thing is that they appear to be barely used within
370 PCL and the munged environment object is passed to cmucl only
371 in calls to macroexpand-1, which is probably why this case fails.
372 SBCL uses essentially the same code, so if the environment hacking
373 is screwed up, it affects us too.
376 Using the pretty-printer from the command prompt gives funny
377 results, apparently because the pretty-printer doesn't know
378 about user's command input, including the user's carriage return
379 that the user, and therefore the pretty-printer thinks that
380 the new output block should start indented 2 or more characters
381 rightward of the correct location.
384 ANSI specifies that the RESULT-TYPE argument of CONCATENATE must be
385 a subtype of SEQUENCE, but CONCATENATE doesn't check this properly:
386 (CONCATENATE 'SIMPLE-ARRAY #(1 2) '(3)) => #(1 2 3)
387 This also leads to funny behavior when derived type specifiers
388 are used, as originally reported by Milan Zamazal for CMU CL (on the
389 Debian bugs mailing list (?) 2000-02-27), then reported by Martin
390 Atzmueller for SBCL (2000-10-01 on sbcl-devel@lists.sourceforge.net):
391 (DEFTYPE FOO () 'SIMPLE-ARRAY)
392 (CONCATENATE 'FOO #(1 2) '(3))
393 => #<ARRAY-TYPE SIMPLE-ARRAY> is a bad type specifier for
395 The derived type specifier FOO should act the same way as the
396 built-in type SIMPLE-ARRAY here, but it doesn't. That problem
397 doesn't seem to exist for sequence types:
398 (DEFTYPE BAR () 'SIMPLE-VECTOR)
399 (CONCATENATE 'BAR #(1 2) '(3)) => #(1 2 3)
400 See also bug #46a./b., and discussion and patch sbcl-devel and
404 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
405 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
406 crashes SBCL. In general tracing anything which is used in the
407 implementation of TRACE is likely to have the same problem.
410 (DECLAIM (OPTIMIZE ..)) doesn't work properly inside LOCALLY forms.
413 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
414 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
415 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
416 WITH-OUTPUT-TO-STRING.
419 ANSI says in one place that type declarations can be abbreviated even
420 when the type name is not a symbol, e.g.
421 (DECLAIM ((VECTOR T) *FOOVECTOR*))
422 SBCL doesn't support this. But ANSI says in another place that this
423 isn't allowed. So it's not clear this is a bug after all. (See the
424 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
428 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
429 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
430 an easily guessable temporary filename in a way which might open
431 applications using LOAD-FOREIGN to hijacking by malicious users
432 on the same machine. Incantations for doing this safely are
433 floating around the net in various "how to write secure programs
434 despite Unix" documents, and it would be good to (1) fix this in
435 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
436 files and make it share the same new safe logic.
439 Functions are assigned names based on the context in which they're
440 defined. This is less than ideal for the functions which are
441 used to implement CLOS methods. E.g. the output of
442 (DESCRIBE 'PRINT-OBJECT) lists functions like
443 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
445 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
446 It would be better if these functions' names always identified
447 them as methods, and identified their generic functions and
451 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
452 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
453 the mean of the distribution can be systematically O(0.1%) wrong.
454 Just increasing R-I-E-B is probably not a good solution, since
455 it would decrease efficiency more than is probably necessary. Perhaps
456 using some sort of accept/reject method would be better.
459 Internally the compiler sometimes evaluates
460 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
461 (I stumbled across this when I added an
462 (assert (not (eq type1 *wild-type*)))
463 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
464 in a type context should probably be translated to T, and so it's
465 probably wrong to ask whether it's equal to the T type and then (using
466 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
467 (I haven't tried to investigate this bug enough to guess whether
468 there might be any user-level symptoms.)
471 Inconsistencies between derived and declared VALUES return types for
472 DEFUN aren't checked very well. E.g. the logic which successfully
473 catches problems like
474 (declaim (ftype (function (fixnum) float) foo))
476 (declare (type integer x))
477 (values x)) ; wrong return type, detected, gives warning, good!
479 (declaim (ftype (function (t) (values t t)) bar))
481 (values x)) ; wrong number of return values, no warning, bad!
482 The cause of this is seems to be that (1) the internal function
483 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
484 arguments symmetrically, and (2) when the type checking code was
485 written back when when SBCL's code was still CMU CL, the intent
487 (declaim (ftype (function (t) t) bar))
489 (values x x)) ; wrong number of return values; should give warning?
490 not be warned for, because a two-valued return value is considered
491 to be compatible with callers who expects a single value to be
492 returned. That intent is probably not appropriate for modern ANSI
493 Common Lisp, but fixing this might be complicated because of other
494 divergences between auld-style and new-style handling of
495 multiple-VALUES types. (Some issues related to this were discussed
496 on cmucl-imp at some length sometime in 2000.)
499 The facility for dumping a running Lisp image to disk gets confused
500 when run without the PURIFY option, and creates an unnecessarily large
501 core file (apparently representing memory usage up to the previous
502 high-water mark). Moreover, when the file is loaded, it confuses the
503 GC, so that thereafter memory usage can never be reduced below that
507 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
508 CL), out-of-line structure slot setters are horribly inefficient
509 whenever the type of the slot is declared, because out-of-line
510 structure slot setters are implemented as closures to save space,
511 so the compiler doesn't compile the type test into code, but
512 instead just saves the type in a lexical closure and interprets it
514 A proper solution involves deciding whether it's really worth
515 saving space by implementing structure slot accessors as closures.
516 (If it's not worth it, the problem vanishes automatically. If it
517 is worth it, there are hacks we could use to force type tests to
518 be compiled anyway, and even shared. E.g. we could implement
519 an EQUAL hash table mapping from types to compiled type tests,
520 and save the appropriate compiled type test as part of each lexical
521 closure; or we could make the lexical closures be placeholders
522 which overwrite their old definition as a lexical closure with
523 a new compiled definition the first time that they're called.)
524 As a workaround for the problem, #'(SETF FOO) expressions can
525 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
526 (defmacro efficient-setf-function (place-function-name)
527 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
528 ;; a workaround for the problem, encouraging the
529 ;; inline expansion of the structure accessor, so
530 ;; that the compiler can optimize its type test
531 (let ((new-value (gensym "NEW-VALUE-"))
532 (structure-value (gensym "STRUCTURE-VALUE-")))
533 `(lambda (,new-value ,structure-value)
534 (setf (,place-function-name ,structure-value)
536 ;; no problem, can just use the ordinary expansion
537 `(function (setf ,place-function-name))))
540 There's apparently a bug in CEILING optimization which caused
541 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
542 applied the patches to SBCL and they didn't seem to cause problems
543 (as reported sbcl-devel 2001-05-04). However, since the patches
544 modify nontrivial code which was apparently written incorrectly
545 the first time around, until regression tests are written I'm not
546 comfortable merging the patches in the CVS version of SBCL.
549 (DESCRIBE 'SB-ALIEN:DEF-ALIEN-TYPE) reports the macro argument list
553 in #<PACKAGE "SB-ALIEN">.
554 Macro-function: #<FUNCTION "DEF!MACRO DEF-ALIEN-TYPE" {19F4A39}>
555 Macro arguments: (#:whole-470 #:environment-471)
556 On Sat, May 26, 2001 09:45:57 AM CDT it was compiled from:
557 /usr/stuff/sbcl/src/code/host-alieneval.lisp
558 Created: Monday, March 12, 2001 07:47:43 AM CST
561 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
562 a clean, just-started SBCL system. And it seems to be right:
563 (ROOM T) can bring a small computer to its knees for a *long*
564 time trying to GC afterwards. Surely there's some more economical
565 way to implement (ROOM T).
568 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
570 ;;; The compiler is flushing the argument type test, and the default
571 ;;; case in the cond, so that calling with say a fixnum 0 causes a
573 (declaim (optimize (safety 2) (speed 3)))
575 (declare (type (or string stream) x))
576 (cond ((typep x 'string) 'string)
577 ((typep x 'stream) 'stream)
580 The symptom in sbcl-0.6.12.42 on OpenBSD is actually (TST 0)=>STREAM
581 (not the SIGBUS reported in the comment) but that's broken too;
582 type declarations are supposed to be treated as assertions unless
583 SAFETY 0, so we should be getting a TYPE-ERROR.
586 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
588 (in-package :cl-user)
589 ;;; From: David Gadbois <gadbois@cyc.com>
591 ;;; Logical pathnames aren't externalizable.
593 (let ((tempfile "/tmp/test.lisp"))
594 (setf (logical-pathname-translations "XXX")
595 '(("XXX:**;*.*" "/tmp/**/*.*")))
596 (with-open-file (out tempfile :direction :output)
597 (write-string "(defvar *path* #P\"XXX:XXX;FOO.LISP\")" out))
598 (compile-file tempfile))
599 The error message in sbcl-0.6.12.42 is
601 ; (while making load form for #<SB-IMPL::LOGICAL-HOST "XXX">)
602 ; A logical host can't be dumped as a constant: #<SB-IMPL::LOGICAL-HOST "XXX">
605 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
607 (in-package :cl-user)
608 ;;; The following invokes a compiler error.
609 (declaim (optimize (speed 2) (debug 3)))
612 (unwind-protect nil)))
616 The error message in sbcl-0.6.12.42 is
617 internal error, failed AVER:
618 "(COMMON-LISP:EQ (SB!C::TN-ENVIRONMENT SB!C:TN) SB!C::TN-ENV)"
621 When the compiler inline expands functions, it may be that different
622 kinds of return values are generated from different code branches.
623 E.g. an inline expansion of POSITION generates integer results
624 from one branch, and NIL results from another. When that inline
625 expansion is used in a context where only one of those results
628 (aref *a1* (position x *a2*)))
629 and the compiler can't prove that the unacceptable branch is
630 never taken, then bogus type mismatch warnings can be generated.
631 If you need to suppress the type mismatch warnings, you can
632 suppress the inline expansion,
634 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
635 (aref *a1* (position x *a2*)))
636 or, sometimes, suppress them by declaring the result to be of an
639 (aref *a1* (the integer (position x *a2*))))
641 This is not a new compiler problem in 0.7.0, but the new compiler
642 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
643 more conspicuous. If you don't need performance from these functions,
644 and the bogus warnings are a nuisance for you, you can return to
645 your pre-0.7.0 state of grace with
646 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
649 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
650 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
651 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
652 when of course it should be NIL. (He says it only fails for X86,
653 not SPARC; dunno about Alpha.)
655 Also, "the same problem exists for LONG-FLOAT-EPSILON,
656 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
657 for the -negative- the + is replaced by a - in the test)."
659 Raymond Toy comments that this is tricky on the X86 since its FPU
660 uses 80-bit precision internally.
663 Even in sbcl-0.pre7.x, which is supposed to be free of the old
664 non-ANSI behavior of treating the function return type inferred
665 from the current function definition as a declaration of the
666 return type from any function of that name, the return type of NIL
667 is attached to FOO in 120a above, and used to optimize code which
671 There was some sort of screwup in handling of
672 (IF (NOT (IGNORE-ERRORS ..))). E.g.
674 (if (not (ignore-errors
675 (make-pathname :host "foo" :directory "!bla" :name "bar")))
677 (error "notunlessnot")))
678 The (NOT (IGNORE-ERRORS ..)) form evaluates to T, so this should be
679 printing "ok", but instead it's going to the ERROR. This problem
680 seems to've been introduced by MNA's HANDLER-CASE patch (sbcl-devel
681 2001-07-17) and as a workaround (put in sbcl-0.pre7.14.flaky4.12)
682 I reverted back to the old weird HANDLER-CASE code. However, I
683 think the problem looks like a compiler bug in handling RETURN-FROM,
684 so I left the MNA-patched code in HANDLER-CASE (suppressed with
685 #+NIL) and I'd like to go back to see whether this really is
686 a compiler bug before I delete this BUGS entry.
689 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
690 the entire lexical environment at the point of MACROLET available
691 in the bodies of the macroexpander functions. In particular, it
692 allows the function bodies (which run at compile time) to try to
693 access lexical variables (which are only defined at runtime).
694 It doesn't even issue a warning, which is bad.
696 The SBCL behavior arguably conforms to the ANSI spec (since the
697 spec says that the behavior is undefined, ergo anything conforms).
698 However, it would be better to issue a compile-time error.
699 Unfortunately I (WHN) don't see any simple way to detect this
700 condition in order to issue such an error, so for the meantime
701 SBCL just does this weird broken "conforming" thing.
703 The ANSI standard says, in the definition of the special operator
705 The macro-expansion functions defined by MACROLET are defined
706 in the lexical environment in which the MACROLET form appears.
707 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
708 the local macro definitions in a MACROLET, but the consequences
709 are undefined if the local macro definitions reference any
710 local variable or function bindings that are visible in that
712 Then it seems to contradict itself by giving the example
714 (macrolet ((fudge (z)
715 ;The parameters x and flag are not accessible
716 ; at this point; a reference to flag would be to
717 ; the global variable of that name.
718 ` (if flag (* ,z ,z) ,z)))
719 ;The parameters x and flag are accessible here.
723 The comment "a reference to flag would be to the global variable
724 of the same name" sounds like good behavior for the system to have.
725 but actual specification quoted above says that the actual behavior
729 (as reported by Gabe Garza on cmucl-help 2001-09-21)
731 (defun test-pred (x y)
735 (func (lambda () x)))
736 (print (eq func func))
737 (print (test-pred func func))
738 (delete func (list func))))
739 Now calling (TEST-CASE) gives output
742 (#<FUNCTION {500A9EF9}>)
743 Evidently Python thinks of the lambda as a code transformation so
744 much that it forgets that it's also an object.
747 The DEFSTRUCT section of the ANSI spec, in the :CONC-NAME section,
748 specifies a precedence rule for name collisions between slot accessors of
749 structure classes related by inheritance. As of 0.7.0, SBCL still
753 insufficient syntax checking in MACROLET:
755 (macrolet ((defmacro bar (z) `(+ z z)))
757 shouldn't compile without error (because of the extra DEFMACRO symbol).
760 As of sbcl-0.pre7.86.flaky7.3, the cross-compiler, and probably
761 the CL:COMPILE function (which is based on the same %COMPILE
762 mechanism) get confused by
764 (labels ((sxhash-number (x)
766 (fixnum (sxhash x)) ; through DEFTRANSFORM
767 (integer (sb!bignum:sxhash-bignum x))
768 (single-float (sxhash x)) ; through DEFTRANSFORM
769 (double-float (sxhash x)) ; through DEFTRANSFORM
770 #!+long-float (long-float (error "stub: no LONG-FLOAT"))
771 (ratio (let ((result 127810327))
772 (declare (type fixnum result))
773 (mixf result (sxhash-number (numerator x)))
774 (mixf result (sxhash-number (denominator x)))
776 (complex (let ((result 535698211))
777 (declare (type fixnum result))
778 (mixf result (sxhash-number (realpart x)))
779 (mixf result (sxhash-number (imagpart x)))
781 (sxhash-recurse (x &optional (depthoid +max-hash-depthoid+))
782 (declare (type index depthoid))
786 (mix (sxhash-recurse (car x) (1- depthoid))
787 (sxhash-recurse (cdr x) (1- depthoid)))
790 (if (typep x 'structure-object)
792 (sxhash ; through DEFTRANSFORM
793 (class-name (layout-class (%instance-layout x)))))
795 (symbol (sxhash x)) ; through DEFTRANSFORM
796 (number (sxhash-number x))
799 (simple-string (sxhash x)) ; through DEFTRANSFORM
800 (string (%sxhash-substring x))
801 (bit-vector (let ((result 410823708))
802 (declare (type fixnum result))
803 (dotimes (i (min depthoid (length x)))
804 (mixf result (aref x i)))
806 (t (logxor 191020317 (sxhash (array-rank x))))))
809 (sxhash (char-code x)))) ; through DEFTRANSFORM
812 complaining "function called with two arguments, but wants exactly
813 one" about SXHASH-RECURSE. (This might not be strictly a new bug,
814 since IIRC post-fork CMU CL has also had problems with &OPTIONAL
815 arguments in FLET/LABELS: it might be an old Python bug which is
816 only exercised by the new arrangement of the SBCL compiler.)
819 Ideally, uninterning a symbol would allow it, and its associated
820 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
821 at least as of sbcl-0.7.0, this isn't the case. Information about
822 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
823 essentially in ordinary (non-weak) hash tables keyed by symbols.
824 Thus, once a system has an entry in this system, it tends to live
825 forever, even when it is uninterned and all other references to it
829 (reported by Arnaud Rouanet on cmucl-imp 2001-12-18)
830 (defmethod foo ((x integer))
832 (defmethod foo :around ((x integer))
835 Now (FOO 3) should return 3, but instead it returns 4.
838 (reported by Alexey Dejneka sbcl-devel 2002-01-03)
840 SUBTYPEP does not work well with redefined classes:
849 * (defclass b (a) ())
862 This bug was fixed in sbcl-0.7.4.1 by invalidating the PCL wrapper
863 class upon redefinition. Unfortunately, doing so causes bug #176 to
864 appear. Pending further investication, one or other of these bugs
865 might be present at any given time.
868 Pretty-printing nested backquotes doesn't work right, as
869 reported by Alexey Dejneka sbcl-devel 2002-01-13:
871 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
872 * (lisp-implementation-version)
876 (as reported by Lynn Quam on cmucl-imp ca. 2002-01-16)
877 %NATURALIZE-C-STRING conses a lot, like 16 bytes per byte
878 of the naturalized string. We could probably port the patches
879 from the cmucl-imp mailing list.
882 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
883 prominent SourceForge web/db bug tracking system, which is
884 unfortunately not a reliable way to get a timely response from
885 the SBCL maintainers)
886 In the course of trying to build a test case for an
887 application error, I encountered this behavior:
888 If you start up sbcl, and then lay on CTRL-C for a
889 minute or two, the lisp process will eventually say:
890 %PRIMITIVE HALT called; the party is over.
891 and throw you into the monitor. If I start up lisp,
892 attach to the process with strace, and then do the same
893 (abusive) thing, I get instead:
894 access failure in heap page not marked as write-protected
895 and the monitor again. I don't know enough to have the
896 faintest idea of what is going on here.
897 This is with sbcl 6.12, uname -a reports:
898 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
899 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
900 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
901 it took more than two minutes (but less than five) for me.
904 (This was once known as IR1-4, but it lived on even after the
905 IR1 interpreter went to the big bit bucket in the sky.)
906 The system accepts DECLAIM in most places where DECLARE would be
907 accepted, without even issuing a warning. ANSI allows this, but since
908 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
909 meaning is rather different, and it's unlikely that the user
910 has a good reason for doing DECLAIM not at top level, it would be
911 good to issue a STYLE-WARNING when this happens. A possible
912 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
913 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
914 [This is considered an IR1-interpreter-related bug because until
915 EVAL-WHEN is rewritten, which won't happen until after the IR1
916 interpreter is gone, the system's notion of what's a top-level form
917 and what's not will remain too confused to fix this problem.]
920 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
921 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
922 COMPLEX implementation didn't deal with this, and hasn't been
923 upgraded to do so. (This doesn't seem to be a high priority
924 conformance problem, since seems hard to construct useful code
928 Floating point errors are reported poorly. E.g. on x86 OpenBSD
931 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
932 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
933 No traps are enabled? How can this be?
934 It should be possible to be much more specific (overflow, division
935 by zero, etc.) and of course the "How can this be?" should be fixable.
937 See also bugs #45.c and #183
940 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
941 (in-package :cl-user)
944 (defstruct foo bar bletch)
946 (labels ((kidify1 (kid)
954 (declare (inline kid-frob))
957 (the simple-vector (foo-bar perd)))))
959 debugger invoked on condition of type TYPE-ERROR:
960 The value NIL is not of type SB-C::NODE.
961 The location of this failure has moved around as various related
962 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
963 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
966 (essentially the same problem as a CMU CL bug reported by Martin
967 Cracauer on cmucl-imp 2002-02-19)
968 There is a hole in structure slot type checking. Compiling and LOADing
969 (declaim (optimize safety))
971 (bla 0 :type fixnum))
973 (let ((foo (make-foo)))
974 (setf (foo-bla foo) '(1 . 1))
975 (format t "Is ~a of type ~a a cons? => ~a~%"
977 (type-of (foo-bla foo))
978 (consp (foo-bla foo)))))
980 should signal an error, but in sbcl-0.7.1.21 instead gives the output
981 Is (1 . 1) of type CONS a cons? => NIL
982 without signalling an error.
985 Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
986 UPGRADED-COMPLEX-PART-TYPE should have an optional environment argument.
987 (reported by Alexey Dejneka sbcl-devel 2002-04-12)
990 (reported by Robert E. Brown 2002-04-16)
991 When a function is called with too few arguments, causing the
992 debugger to be entered, the uninitialized slots in the bad call frame
993 seem to cause GCish problems, being interpreted as tagged data even
994 though they're not. In particular, executing ROOM in the
995 debugger at that point causes AVER failures:
998 * (lisp-implementation-version)
1004 failed AVER: "(SAP= CURRENT END)"
1005 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
1006 isn't too surprising since there are many differences in stack
1007 implementation and GC conservatism between the X86 and other ports.)
1011 (in-package :cl-user)
1013 (defmethod permanentize ((uustk uustk))
1014 (flet ((frob (hash-table test-for-deletion)
1016 (obj-entry.stale? (oe)
1017 (destructuring-bind (key . datum) oe
1018 (declare (type simple-vector key))
1019 (deny0 (void? datum))
1020 (some #'stale? key))))
1021 (declare (inline frob obj-entry.stale?))
1022 (frob (uustk.args-hash->obj-alist uustk)
1024 (frob (uustk.hash->memoized-objs-list uustk)
1027 in sbcl-0.7.3.11 causes an assertion failure,
1030 (AND (NULL (BLOCK-SUCC B))
1031 (NOT (BLOCK-DELETE-P B))
1032 (NOT (EQ B (COMPONENT-HEAD #)))))"
1035 In sbcl-0.7.3.11, compiling the (illegal) code
1036 (in-package :cl-user)
1037 (defmethod prove ((uustk uustk))
1038 (zap ((frob () nil))
1040 gives the (not terribly clear) error message
1042 ; (during macroexpansion of (DEFMETHOD PROVE ...))
1043 ; can't get template for (FROB NIL NIL)
1044 The problem seems to be that the code walker used by the DEFMETHOD
1045 macro is unhappy with the illegal syntax in the method body, and
1046 is giving an unclear error message.
1049 (reported by Dan Barlow on sbcl-devel 2002-05-10)
1050 In sbcl-0.7.3.12, doing
1051 (defstruct foo bar baz)
1052 (compile nil (lambda (x) (or x (foo-baz x))))
1054 debugger invoked on condition of type SB-INT:BUG:
1055 full call to SB-KERNEL:%INSTANCE-REF
1056 This is probably a bug in SBCL itself. [...]
1057 Since this is a reasonable user error, it shouldn't be reported as
1061 (reported by Pierre Mai while investigating bug 47):
1062 (DEFCLASS FOO () ((A :SILLY T)))
1063 signals a SIMPLE-ERROR, not a PROGRAM-ERROR.
1066 sbcl's treatment of at least macro lambda lists is too permissive;
1067 e.g., in sbcl-0.7.3.7:
1068 (defmacro foo (&rest rest bar) `(,bar ,rest))
1069 (macroexpand '(foo quux zot)) -> (QUUX (QUUX ZOT))
1070 whereas section 3.4.4 of the CLHS doesn't allow required parameters
1071 to come after the rest argument.
1074 The compiler sometimes tries to constant-fold expressions before
1075 it checks to see whether they can be reached. This can lead to
1076 bogus warnings about errors in the constant folding, e.g. in code
1079 (WRITE-STRING (> X 0) "+" "0"))
1080 compiled in a context where the compiler can prove that X is NIL,
1081 and the compiler complains that (> X 0) causes a type error because
1082 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
1083 caused a full WARNING, which made the bug really annoying because then
1084 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
1085 code. Since then the warning has been downgraded to STYLE-WARNING,
1086 so it's still a bug but at least it's a little less annoying.
1089 The error message from attempting to use a #\Return format
1091 (format nil "~^M") ; replace "^M" with a literal #\Return
1092 debugger invoked on condition of type SB-FORMAT::FORMAT-ERROR:
1093 error in format: unknown format directive
1096 is not terribly helpful; this is more noticeable than parallel cases
1097 with e.g. #\Backspace because of the differing newline conventions
1098 on various operating systems. (reported by Harald Hanche-Olsen on
1099 cmucl-help 2002-05-31)
1102 reported by Alexey Dejneka 08 Jun 2002 in sbcl-devel:
1103 Playing with McCLIM, I've received an error "Unbound variable WRAPPER
1104 in SB-PCL::CHECK-WRAPPER-VALIDITY".
1105 (defun check-wrapper-validity (instance)
1106 (let* ((owrapper (wrapper-of instance)))
1107 (if (not (invalid-wrapper-p owrapper))
1109 (let* ((state (wrapper-state wrapper)) ; !!!
1111 I've tried to replace it with OWRAPPER, but now OBSOLETE-INSTANCE-TRAP
1112 breaks with "NIL is not of type SB-KERNEL:LAYOUT".
1114 partial fix: The undefined variable WRAPPER resulted from an error
1115 in recent refactoring, as can be seen by comparing to the code in e.g.
1116 sbcl-0.7.2. Replacing WRAPPER with OWRAPPER (done by WHN in sbcl-0.7.4.22)
1117 should bring the code back to its behavior as of sbcl-0.7.2, but
1118 that still leaves the OBSOLETE-INSTANCE-TRAP bug. An example of
1119 input which triggers that bug is
1121 (let ((lastname (intern (format nil "C~D" (1- i))))
1122 (name (intern (format nil "C~D" i))))
1123 (eval `(defclass ,name
1124 (,@(if (= i 0) nil (list lastname)))
1126 (eval `(defmethod initialize-instance :after ((x ,name) &rest any)
1127 (declare (ignore any))))))
1129 (defclass c0 (b) ())
1130 (make-instance 'c19)
1134 178: "AVER failure compiling confused THEs in FUNCALL"
1135 In sbcl-0.7.4.24, compiling
1137 (funcall (the function (the standard-object x))))
1140 "(AND (EQ (IR2-CONTINUATION-PRIMITIVE-TYPE 2CONT) FUNCTION-PTYPE) (EQ CHECK T))"
1141 This variant compiles OK, though:
1142 (defun bug178alternative (x)
1143 (funcall (the nil x)))
1145 181: "bad type specifier drops compiler into debugger"
1147 (in-package :cl-user)
1149 (declare (type 0 x))
1152 bad thing to be a type specifier: 0
1153 which seems fine, but also enters the debugger (instead of having
1154 the compiler handle the error, convert it into a COMPILER-ERROR, and
1155 continue compiling) which seems wrong.
1157 183: "IEEE floating point issues"
1158 Even where floating point handling is being dealt with relatively
1159 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
1160 accrued-exceptions and current-exceptions part of the fp control
1161 word don't seem to bear much relation to reality. E.g. on
1165 debugger invoked on condition of type DIVISION-BY-ZERO:
1166 arithmetic error DIVISION-BY-ZERO signalled
1167 0] (sb-vm::get-floating-point-modes)
1169 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1170 :ROUNDING-MODE :NEAREST
1171 :CURRENT-EXCEPTIONS NIL
1172 :ACCRUED-EXCEPTIONS (:INEXACT)
1175 * (sb-vm::get-floating-point-modes)
1176 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1177 :ROUNDING-MODE :NEAREST
1178 :CURRENT-EXCEPTIONS (:INEXACT)
1179 :ACCRUED-EXCEPTIONS (:INEXACT)
1182 185: "top-level forms at the REPL"
1183 * (locally (defstruct foo (a 0 :type fixnum)))
1186 ; (in macroexpansion of (SB-KERNEL::%DELAYED-GET-COMPILER-LAYOUT BAR))
1187 however, compiling and loading the same expression in a file works
1190 187: "type inference confusion around DEFTRANSFORM time"
1191 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
1192 bug in DEFTRANSFORM")
1193 After the file below is compiled and loaded in sbcl-0.7.5, executing
1194 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
1195 at the REPL returns an adjustable vector, which is wrong. Presumably
1196 somehow the DERIVE-TYPE information for the output values of %WAD is
1197 being mispropagated as a type constraint on the input values of %WAD,
1198 and so causing the type test to be optimized away. It's unclear how
1199 hand-expanding the DEFTRANSFORM would change this, but it suggests
1200 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
1201 invoked at a particular phase) is involved.
1202 (cl:in-package :sb-c)
1203 (eval-when (:compile-toplevel)
1204 ;;; standin for %DATA-VECTOR-AND-INDEX
1205 (defknown %dvai (array index)
1207 (foldable flushable))
1208 (deftransform %dvai ((array index)
1212 (let* ((atype (continuation-type array))
1213 (eltype (array-type-specialized-element-type atype)))
1214 (when (eq eltype *wild-type*)
1215 (give-up-ir1-transform
1216 "specialized array element type not known at compile-time"))
1217 (when (not (array-type-complexp atype))
1218 (give-up-ir1-transform "SIMPLE array!"))
1219 `(if (array-header-p array)
1220 (%wad array index nil)
1221 (values array index))))
1222 ;;; standin for %WITH-ARRAY-DATA
1223 (defknown %wad (array index (or index null))
1224 (values (simple-array * (*)) index index index)
1225 (foldable flushable))
1226 ;;; (Commenting out this optimizer causes the bug to go away.)
1227 (defoptimizer (%wad derive-type) ((array start end))
1228 (let ((atype (continuation-type array)))
1229 (when (array-type-p atype)
1230 (values-specifier-type
1231 `(values (simple-array ,(type-specifier
1232 (array-type-specialized-element-type atype))
1234 index index index)))))
1236 (defun %wad (array start end)
1237 (format t "~&in %WAD~%")
1238 (%with-array-data array start end))
1239 (cl:in-package :cl-user)
1241 (declare (type (vector t) v))
1242 (declare (notinline sb-kernel::%with-array-data))
1243 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
1247 188: "compiler performance fiasco involving type inference and UNION-TYPE"
1248 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
1249 bug could be fixed properly, so you won't see the bug unless you restore
1250 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
1254 (declare (optimize (safety 3)))
1255 (declare (optimize (compilation-speed 2)))
1256 (declare (optimize (speed 1) (debug 1) (space 1)))
1257 (let ((fn "if-this-file-exists-the-universe-is-strange"))
1258 (load fn :if-does-not-exist nil)
1259 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
1260 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
1261 (load (concatenate 'string fn ".misc-garbage")
1262 :if-does-not-exist nil)))))
1264 134.552 seconds of real time
1265 133.35156 seconds of user run time
1266 0.03125 seconds of system run time
1267 [Run times include 2.787 seconds GC run time.]
1269 246883368 bytes consed.
1270 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
1271 thinking about type relationships involving types like
1273 (OR (INTEGER 576 576)
1284 190: "PPC/Linux pipe? buffer? bug"
1285 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
1286 on the PPC/Linux platform, waiting for a zombie env process. This
1287 is a classic symptom of buffer filling and deadlock, but it seems
1288 only sporadically reproducible.
1290 191: "Miscellaneous PCL deficiencies"
1291 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
1292 a. DEFCLASS does not inform the compiler about generated
1293 functions. Compiling a file with
1294 (DEFCLASS A-CLASS ()
1296 (DEFUN A-CLASS-X (A)
1297 (WITH-SLOTS (A-CLASS-X) A
1299 results in a STYLE-WARNING:
1301 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
1303 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
1304 Mai points out that the declamation of functions is in fact
1305 incorrect in some cases (most notably for structure
1306 classes). This means that at present erroneous attempts to use
1307 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
1308 won't get the corresponding STYLE-WARNING.
1309 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
1310 lists and &KEY arguments) do not signal errors when they should.
1312 192: "Python treats free type declarations as promises."
1313 a. original report by Alexey Dejneka (on sbcl-devel 2002-08-26):
1314 (declaim (optimize (speed 0) (safety 3)))
1318 (locally (declare (single-float x))
1320 Now (F NIL) correctly gives a type error, but (F 100) gives
1321 a segmentation violation.
1322 b. same fundamental problem in a different way, easy to stumble
1323 across if you mistype and declare the wrong index in
1324 (DOTIMES (I ...) (DOTIMES (J ...) (DECLARE ...) ...)):
1325 (declaim (optimize (speed 1) (safety 3)))
1326 (defun trust-assertion (i)
1328 (declare (type (mod 4) i)) ; when commented out, behavior changes!
1331 (trust-assertion 6) ; prints nothing unless DECLARE is commented out
1333 193: "unhelpful CLOS error reporting when the primary method is missing"
1335 (defmethod foo :before ((x t)) (print x))
1336 is the only method defined on FOO, the error reporting when e.g.
1338 is relatively unhelpful:
1339 There is no primary method for the generic function
1340 #<STANDARD-GENERIC-FUNCTION FOO (1)>.
1341 with the offending argument nowhere visible in the backtrace. This
1342 continues even if there *are* primary methods, just not for the
1343 specified arg type, e.g.
1344 (defmethod foo ((x character)) (print x))
1345 (defmethod foo ((x string)) (print x))
1346 (defmethod foo ((x pathname)) ...)
1347 In that case it could be very helpful to know what argument value is
1348 falling through the cracks of the defined primary methods, but the
1349 error message stays the same (even BACKTRACE doesn't tell you what the
1350 bad argument value is).
1352 194: "no error from (THE REAL '(1 2 3)) in some cases"
1354 (multiple-value-prog1 (progn (the real '(1 2 3))))
1355 returns (1 2 3) instead of signalling an error. Also in sbcl-0.7.7.9,
1356 a more complicated instance of this bug kept
1357 (IGNORE-ERRORS (MIN '(1 2 3))) from returning NIL as it should when
1358 the MIN source transform expanded to (THE REAL '(1 2 3)), because
1359 (IGNORE-ERRORS (THE REAL '(1 2 3))) returns (1 2 3).
1360 Alexey Dejneka pointed out that
1361 (IGNORE-ERRORS (IDENTITY (THE REAL '(1 2 3)))) works as it should.
1362 (IGNORE-ERRORS (VALUES (THE REAL '(1 2 3)))) also works as it should.
1363 Perhaps this is another case of VALUES type intersections behaving
1365 When I (WHN) tried to use the VALUES trick to work around this bug
1366 in the MIN source transform, it didn't work for
1367 (assert (null (ignore-errors (min 1 #(1 2 3)))))
1368 Hand-expanding the source transform, I get
1369 (assert (null (ignore-errors
1371 (arg2 (identity (the real #(1 2 3)))))
1372 (if (< arg1 arg2) arg1 arg2)))))
1373 which fails (i.e. the assertion fails, because the IGNORE-ERRORS
1374 doesn't report MIN signalling a type error). At the REPL
1375 (null (ignore-errors
1377 (arg2 (identity (the real #(1 2 3)))))
1378 (if (< arg1 arg2) arg1 arg2))))
1380 but when this expression is used as the body of (DEFUN FOO () ...)
1383 195: "confusing reporting of not-a-REAL TYPE-ERRORs from THE REAL"
1384 In sbcl-0.7.7.10, (THE REAL #(1 2 3)) signals a type error which
1385 prints as "This is not a (OR SINGLE-FLOAT DOUBLE-FLOAT RATIONAL)".
1386 The (OR SINGLE-FLOAT DOUBLE-FLOAT RATIONAL) representation of
1387 REAL is unnecessarily confusing, especially since it relies on
1388 internal implementation knowledge that even with SHORT-FLOAT
1389 and LONG-FLOAT left out of the union, this type is equal to REAL.
1390 So it'd be better just to say "This is not a REAL".
1392 DEFUNCT CATEGORIES OF BUGS
1394 These labels were used for bugs related to the old IR1 interpreter.
1395 The # values reached 6 before the category was closed down.