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.
93 The "compiling top-level form:" output ought to be condensed.
94 Perhaps any number of such consecutive lines ought to turn into a
95 single "compiling top-level forms:" line.
98 The way that the compiler munges types with arguments together
99 with types with no arguments (in e.g. TYPE-EXPAND) leads to
100 weirdness visible to the user:
101 (DEFTYPE FOO () 'FIXNUM)
103 (TYPEP 11 '(FOO)) => T, which seems weird
104 (TYPEP 11 'FIXNUM) => T
105 (TYPEP 11 '(FIXNUM)) signals an error, as it should
106 The situation is complicated by the presence of Common Lisp types
107 like UNSIGNED-BYTE (which can either be used in list form or alone)
108 so I'm not 100% sure that the behavior above is actually illegal.
109 But I'm 90+% sure, and the following related behavior,
111 treating the bare symbol AND as equivalent to '(AND), is specifically
112 forbidden (by the ANSI specification of the AND type).
115 It would be nice if the
117 (during macroexpansion)
118 said what macroexpansion was at fault, e.g.
120 (during macroexpansion of IN-PACKAGE,
121 during macroexpansion of DEFFOO)
124 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
125 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
126 (Also, when this is fixed, we can enable the code in PROCLAIM which
127 checks for incompatible FTYPE redeclarations.)
130 (I *think* this is a bug. It certainly seems like strange behavior. But
131 the ANSI spec is scary, dark, and deep.. -- WHN)
132 (FORMAT NIL "~,1G" 1.4) => "1. "
133 (FORMAT NIL "~3,1G" 1.4) => "1. "
136 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
138 (setf (find-class 'ccc1) (find-class 'ccc))
139 (defmethod zut ((c ccc1)) 123)
140 In sbcl-0.7.1.13, this gives an error,
141 There is no class named CCC1.
142 DTC's recommended workaround from the mailing list 3 Mar 2000:
143 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
146 Sometimes (SB-EXT:QUIT) fails with
147 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
148 Process inferior-lisp exited abnormally with code 1
149 I haven't noticed a repeatable case of this yet.
152 The printer doesn't report closures very well. This is true in
156 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
157 It would be nice to make closures have a settable name slot,
158 and make things like DEFSTRUCT and FLET, which create closures,
159 set helpful values into this slot.
162 And as long as we're wishing, it would be awfully nice if INSPECT could
163 also report on closures, telling about the values of the bound variables.
166 The compiler assumes that any time a function of declared FTYPE
167 doesn't signal an error, its arguments were of the declared type.
168 E.g. compiling and loading
169 (DECLAIM (OPTIMIZE (SAFETY 3)))
170 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
172 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
174 (COND ((> (FACTORIAL X) 1.0E6)
175 (FORMAT T "too big~%"))
177 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
179 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
182 will cause the INTEGERP case to be selected, giving bogus output a la
184 This violates the "declarations are assertions" principle.
185 According to the ANSI spec, in the section "System Class FUNCTION",
186 this is a case of "lying to the compiler", but the lying is done
187 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
188 not by the unexpectedly general definition of FACTORIAL. In any case,
189 "declarations are assertions" means that lying to the compiler should
190 cause an error to be signalled, and should not cause a bogus
191 result to be returned. Thus, the compiler should not assume
192 that arbitrary functions check their argument types. (It might
193 make sense to add another flag (CHECKED?) to DEFKNOWN to
194 identify functions which *do* check their argument types.)
195 (Also, verify that the compiler handles declared function
196 return types as assertions.)
199 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
200 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
202 (DECLARE (TYPE INDEXOID X))
203 (THE (VALUES INDEXOID)
205 where the implementation of the type check in function FOO
206 includes a full call to %TYPEP. There are also some fundamental problems
207 with the interpretation of VALUES types (inherited from CMU CL, and
208 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
209 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
212 The definitions of SIGCONTEXT-FLOAT-REGISTER and
213 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
214 supported on FreeBSD because the floating point state is not saved,
215 but at least as of FreeBSD 4.0, the floating point state *is* saved,
216 so they could be supported after all. Very likely
217 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
220 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
221 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
222 issue with 'union, 'and, 'or etc. So even though according to the
223 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
224 (and now SBCL) interpret them as legal types.
227 a slew of floating-point-related errors reported by Peter Van Eynde
229 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
230 should probably be 1.4012985e-45. In SBCL,
231 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
232 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
233 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
234 and LEAST-NEGATIVE-LONG-FLOAT.
235 c: Many expressions generate floating infinity on x86/Linux:
240 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
241 on x86/Linux generates the infinities instead. That might or
242 might not be conforming behavior, but it's also inconsistent,
243 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
244 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
245 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
247 d: (in section12.erg) various forms a la
248 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
249 don't give the right behavior.
252 type safety errors reported by Peter Van Eynde July 25, 2000:
253 a: (COERCE (QUOTE (A B C)) (QUOTE (VECTOR * 4)))
255 In general lengths of array type specifications aren't
256 checked by COERCE, so it fails when the spec is
257 (VECTOR 4), (STRING 2), (SIMPLE-BIT-VECTOR 3), or whatever.
258 b: CONCATENATE has the same problem of not checking the length
259 of specified output array types. MAKE-SEQUENCE and MAP and
260 MERGE also have the same problem.
261 c: (COERCE 'AND 'FUNCTION) returns something related to
262 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
263 h: (MAKE-CONCATENATED-STREAM (MAKE-STRING-OUTPUT-STREAM))
264 should signal TYPE-ERROR.
265 i: MAKE-TWO-WAY-STREAM doesn't check that its arguments can
266 be used for input and output as needed. It should fail with
267 TYPE-ERROR when handed e.g. the results of
268 MAKE-STRING-INPUT-STREAM or MAKE-STRING-OUTPUT-STREAM in
269 the inappropriate positions, but doesn't.
270 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
271 not a binary input stream, but instead cheerfully reads from
272 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
275 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
276 a: (DEFCLASS FOO () (A B A)) should signal a PROGRAM-ERROR, and
278 b: (DEFCLASS FOO () (A B A) (:DEFAULT-INITARGS X A X B)) should
279 signal a PROGRAM-ERROR, and doesn't.
280 c: (DEFCLASS FOO07 NIL ((A :ALLOCATION :CLASS :ALLOCATION :CLASS))),
281 and other DEFCLASS forms with duplicate specifications in their
282 slots, should signal a PROGRAM-ERROR, and doesn't.
283 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
284 causes a COMPILER-ERROR.
287 SYMBOL-MACROLET bugs reported by Peter Van Eynde July 25, 2000:
288 a: (SYMBOL-MACROLET ((T TRUE)) ..) should probably signal
289 PROGRAM-ERROR, but SBCL accepts it instead.
290 b: SYMBOL-MACROLET should refuse to bind something which is
291 declared as a global variable, signalling PROGRAM-ERROR.
292 c: SYMBOL-MACROLET should signal PROGRAM-ERROR if something
293 it binds is declared SPECIAL inside.
296 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
298 (DEFGENERIC FOO02 (X))
299 (DEFMETHOD FOO02 ((X NUMBER)) T)
300 (LET ((M (FIND-METHOD (FUNCTION FOO02)
302 (LIST (FIND-CLASS (QUOTE NUMBER))))))
303 (REMOVE-METHOD (FUNCTION FOO02) M)
304 (DEFGENERIC FOO03 (X))
305 (ADD-METHOD (FUNCTION FOO03) M)))
306 should give an error, but SBCL allows it.
307 b: READ should probably return READER-ERROR, not the bare
308 arithmetic error, when input a la "1/0" or "1e1000" causes
312 It has been reported (e.g. by Peter Van Eynde) that there are
313 several metaobject protocol "errors". (In order to fix them, we might
314 need to document exactly what metaobject protocol specification
315 we're following -- the current code is just inherited from PCL.)
318 The implementation of #'+ returns its single argument without
319 type checking, e.g. (+ "illegal") => "illegal".
322 The debugger LIST-LOCATIONS command doesn't work properly.
325 Compiling and loading
326 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
328 then requesting a BACKTRACE at the debugger prompt gives no information
329 about where in the user program the problem occurred.
332 The compiler is supposed to do type inference well enough that
335 ((SIMPLE-ARRAY SINGLE-FLOAT)
337 (DECLARE (TYPE (SIMPLE-ARRAY SINGLE-FLOAT) X))
340 is redundant. However, as reported by Juan Jose Garcia Ripoll for
341 CMU CL, it sometimes doesn't. Adding declarations is a pretty good
342 workaround for the problem for now, but can't be done by the TYPECASE
343 macros themselves, since it's too hard for the macro to detect
344 assignments to the variable within the clause.
345 Note: The compiler *is* smart enough to do the type inference in
346 many cases. This case, derived from a couple of MACROEXPAND-1
347 calls on Ripoll's original test case,
349 (DECLARE (OPTIMIZE SPEED (SAFETY 0)))
350 (COND ((TYPEP A '(SIMPLE-ARRAY SINGLE-FLOAT)) NIL
351 (LET ((LENGTH (ARRAY-TOTAL-SIZE A)))
352 (LET ((I 0) (G2554 LENGTH))
353 (DECLARE (TYPE REAL G2554) (TYPE REAL I))
356 (WHEN (>= I G2554) (GO SB-LOOP::END-LOOP))
357 (SETF (ROW-MAJOR-AREF A I) (- (ROW-MAJOR-AREF A I)))
358 (GO SB-LOOP::NEXT-LOOP)
359 SB-LOOP::END-LOOP))))))
360 demonstrates the problem; but the problem goes away if the TAGBODY
361 and GO forms are removed (leaving the SETF in ordinary, non-looping
362 code), or if the TAGBODY and GO forms are retained, but the
363 assigned value becomes 0.0 instead of (- (ROW-MAJOR-AREF A I)).
366 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
367 I am looking into this problem that showed up on the cmucl-help
368 list. It seems to me that the "implementation specific environment
369 hacking functions" found in pcl/walker.lisp are completely messed
370 up. The good thing is that they appear to be barely used within
371 PCL and the munged environment object is passed to cmucl only
372 in calls to macroexpand-1, which is probably why this case fails.
373 SBCL uses essentially the same code, so if the environment hacking
374 is screwed up, it affects us too.
377 Using the pretty-printer from the command prompt gives funny
378 results, apparently because the pretty-printer doesn't know
379 about user's command input, including the user's carriage return
380 that the user, and therefore the pretty-printer thinks that
381 the new output block should start indented 2 or more characters
382 rightward of the correct location.
385 (probably related to bug #70; maybe related to bug #109)
386 As reported by Carl Witty on submit@bugs.debian.org 1999-05-08,
388 (in-package "CL-USER")
389 (defun equal-terms (termx termy)
391 ((alpha-equal-bound-term-lists (listx listy)
392 (or (and (null listx) (null listy))
394 (let ((bindings-x (bindings-of-bound-term (car listx)))
395 (bindings-y (bindings-of-bound-term (car listy))))
396 (if (and (null bindings-x) (null bindings-y))
397 (alpha-equal-terms (term-of-bound-term (car listx))
398 (term-of-bound-term (car listy)))
399 (and (= (length bindings-x) (length bindings-y))
401 (enter-binding-pairs (bindings-of-bound-term (car listx))
402 (bindings-of-bound-term (car listy)))
403 (alpha-equal-terms (term-of-bound-term (car listx))
404 (term-of-bound-term (car listy)))
405 (exit-binding-pairs (bindings-of-bound-term (car listx))
406 (bindings-of-bound-term (car listy)))))))
407 (alpha-equal-bound-term-lists (cdr listx) (cdr listy)))))
409 (alpha-equal-terms (termx termy)
410 (if (and (variable-p termx)
412 (equal-bindings (id-of-variable-term termx)
413 (id-of-variable-term termy))
414 (and (equal-operators-p (operator-of-term termx) (operator-of-term termy))
415 (alpha-equal-bound-term-lists (bound-terms-of-term termx)
416 (bound-terms-of-term termy))))))
420 (with-variable-invocation (alpha-equal-terms termx termy))))))
421 causes an assertion failure
422 The assertion (EQ (C::LAMBDA-TAIL-SET C::CALLER)
423 (C::LAMBDA-TAIL-SET (C::LAMBDA-HOME C::CALLEE))) failed.
425 Bob Rogers reports (1999-07-28 on cmucl-imp@cons.org) a smaller test
426 case with the same problem:
427 (defun parse-fssp-alignment ()
428 ;; Given an FSSP alignment file named by the argument . . .
429 (labels ((get-fssp-char ()
433 ;; Stub body, enough to tickle the bug.
434 (list (read-fssp-char)
438 ANSI specifies that the RESULT-TYPE argument of CONCATENATE must be
439 a subtype of SEQUENCE, but CONCATENATE doesn't check this properly:
440 (CONCATENATE 'SIMPLE-ARRAY #(1 2) '(3)) => #(1 2 3)
441 This also leads to funny behavior when derived type specifiers
442 are used, as originally reported by Milan Zamazal for CMU CL (on the
443 Debian bugs mailing list (?) 2000-02-27), then reported by Martin
444 Atzmueller for SBCL (2000-10-01 on sbcl-devel@lists.sourceforge.net):
445 (DEFTYPE FOO () 'SIMPLE-ARRAY)
446 (CONCATENATE 'FOO #(1 2) '(3))
447 => #<ARRAY-TYPE SIMPLE-ARRAY> is a bad type specifier for
449 The derived type specifier FOO should act the same way as the
450 built-in type SIMPLE-ARRAY here, but it doesn't. That problem
451 doesn't seem to exist for sequence types:
452 (DEFTYPE BAR () 'SIMPLE-VECTOR)
453 (CONCATENATE 'BAR #(1 2) '(3)) => #(1 2 3)
456 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
457 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
458 crashes SBCL. In general tracing anything which is used in the
459 implementation of TRACE is likely to have the same problem.
462 (probably related to bug #65; maybe related to bug #109)
463 The compiler doesn't like &OPTIONAL arguments in LABELS and FLET
465 (DEFUN FIND-BEFORE (ITEM SEQUENCE &KEY (TEST #'EQL))
466 (LABELS ((FIND-ITEM (OBJ SEQ TEST &OPTIONAL (VAL NIL))
467 (LET ((ITEM (FIRST SEQ)))
470 ((FUNCALL TEST OBJ ITEM)
473 (FIND-ITEM OBJ (REST SEQ) TEST (NCONC VAL `(,ITEM))))))))
474 (FIND-ITEM ITEM SEQUENCE TEST)))
475 from David Young's bug report on cmucl-help@cons.org 30 Nov 2000
476 causes sbcl-0.6.9 to fail with
477 error in function SB-KERNEL:ASSERT-ERROR:
478 The assertion (EQ (SB-C::LAMBDA-TAIL-SET SB-C::CALLER)
479 (SB-C::LAMBDA-TAIL-SET
480 (SB-C::LAMBDA-HOME SB-C::CALLEE))) failed.
483 (DECLAIM (OPTIMIZE ..)) doesn't work properly inside LOCALLY forms.
486 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
487 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
488 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
489 WITH-OUTPUT-TO-STRING.
492 ANSI says in one place that type declarations can be abbreviated even
493 when the type name is not a symbol, e.g.
494 (DECLAIM ((VECTOR T) *FOOVECTOR*))
495 SBCL doesn't support this. But ANSI says in another place that this
496 isn't allowed. So it's not clear this is a bug after all. (See the
497 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
501 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
502 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
503 an easily guessable temporary filename in a way which might open
504 applications using LOAD-FOREIGN to hijacking by malicious users
505 on the same machine. Incantations for doing this safely are
506 floating around the net in various "how to write secure programs
507 despite Unix" documents, and it would be good to (1) fix this in
508 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
509 files and make it share the same new safe logic.
512 Functions are assigned names based on the context in which they're
513 defined. This is less than ideal for the functions which are
514 used to implement CLOS methods. E.g. the output of
515 (DESCRIBE 'PRINT-OBJECT) lists functions like
516 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
518 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
519 It would be better if these functions' names always identified
520 them as methods, and identified their generic functions and
524 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
525 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
526 the mean of the distribution can be systematically O(0.1%) wrong.
527 Just increasing R-I-E-B is probably not a good solution, since
528 it would decrease efficiency more than is probably necessary. Perhaps
529 using some sort of accept/reject method would be better.
532 Internally the compiler sometimes evaluates
533 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
534 (I stumbled across this when I added an
535 (assert (not (eq type1 *wild-type*)))
536 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
537 in a type context should probably be translated to T, and so it's
538 probably wrong to ask whether it's equal to the T type and then (using
539 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
540 (I haven't tried to investigate this bug enough to guess whether
541 there might be any user-level symptoms.)
544 a latent cross-compilation/bootstrapping bug: The cross-compilation
545 host's CL:CHAR-CODE-LIMIT is used in target code in readtable.lisp
546 and possibly elsewhere. Instead, we should use the target system's
547 CHAR-CODE-LIMIT. This will probably cause problems if we try to
548 bootstrap on a system which uses a different value of CHAR-CODE-LIMIT
552 Inconsistencies between derived and declared VALUES return types for
553 DEFUN aren't checked very well. E.g. the logic which successfully
554 catches problems like
555 (declaim (ftype (function (fixnum) float) foo))
557 (declare (type integer x))
558 (values x)) ; wrong return type, detected, gives warning, good!
560 (declaim (ftype (function (t) (values t t)) bar))
562 (values x)) ; wrong number of return values, no warning, bad!
563 The cause of this is seems to be that (1) the internal function
564 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
565 arguments symmetrically, and (2) when the type checking code was
566 written back when when SBCL's code was still CMU CL, the intent
568 (declaim (ftype (function (t) t) bar))
570 (values x x)) ; wrong number of return values; should give warning?
571 not be warned for, because a two-valued return value is considered
572 to be compatible with callers who expects a single value to be
573 returned. That intent is probably not appropriate for modern ANSI
574 Common Lisp, but fixing this might be complicated because of other
575 divergences between auld-style and new-style handling of
576 multiple-VALUES types. (Some issues related to this were discussed
577 on cmucl-imp at some length sometime in 2000.)
580 The facility for dumping a running Lisp image to disk gets confused
581 when run without the PURIFY option, and creates an unnecessarily large
582 core file (apparently representing memory usage up to the previous
583 high-water mark). Moreover, when the file is loaded, it confuses the
584 GC, so that thereafter memory usage can never be reduced below that
588 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
589 CL), out-of-line structure slot setters are horribly inefficient
590 whenever the type of the slot is declared, because out-of-line
591 structure slot setters are implemented as closures to save space,
592 so the compiler doesn't compile the type test into code, but
593 instead just saves the type in a lexical closure and interprets it
595 A proper solution involves deciding whether it's really worth
596 saving space by implementing structure slot accessors as closures.
597 (If it's not worth it, the problem vanishes automatically. If it
598 is worth it, there are hacks we could use to force type tests to
599 be compiled anyway, and even shared. E.g. we could implement
600 an EQUAL hash table mapping from types to compiled type tests,
601 and save the appropriate compiled type test as part of each lexical
602 closure; or we could make the lexical closures be placeholders
603 which overwrite their old definition as a lexical closure with
604 a new compiled definition the first time that they're called.)
605 As a workaround for the problem, #'(SETF FOO) expressions can
606 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
607 (defmacro efficient-setf-function (place-function-name)
608 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
609 ;; a workaround for the problem, encouraging the
610 ;; inline expansion of the structure accessor, so
611 ;; that the compiler can optimize its type test
612 (let ((new-value (gensym "NEW-VALUE-"))
613 (structure-value (gensym "STRUCTURE-VALUE-")))
614 `(lambda (,new-value ,structure-value)
615 (setf (,place-function-name ,structure-value)
617 ;; no problem, can just use the ordinary expansion
618 `(function (setf ,place-function-name))))
621 There's apparently a bug in CEILING optimization which caused
622 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
623 applied the patches to SBCL and they didn't seem to cause problems
624 (as reported sbcl-devel 2001-05-04). However, since the patches
625 modify nontrivial code which was apparently written incorrectly
626 the first time around, until regression tests are written I'm not
627 comfortable merging the patches in the CVS version of SBCL.
630 As reported by Arthur Lemmens sbcl-devel 2001-05-05, ANSI
631 requires that SYMBOL-MACROLET refuse to rebind special variables,
632 but SBCL doesn't do this. (Also as reported by AL in the same
633 message, SBCL depended on this nonconforming behavior to build
634 itself, because of the way that **CURRENT-SEGMENT** was implemented.
635 As of sbcl-0.7.3.x, this dependence on the nonconforming behavior
636 has been fixed, but the nonconforming behavior remains.)
639 (DESCRIBE 'SB-ALIEN:DEF-ALIEN-TYPE) reports the macro argument list
643 in #<PACKAGE "SB-ALIEN">.
644 Macro-function: #<FUNCTION "DEF!MACRO DEF-ALIEN-TYPE" {19F4A39}>
645 Macro arguments: (#:whole-470 #:environment-471)
646 On Sat, May 26, 2001 09:45:57 AM CDT it was compiled from:
647 /usr/stuff/sbcl/src/code/host-alieneval.lisp
648 Created: Monday, March 12, 2001 07:47:43 AM CST
651 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
652 a clean, just-started SBCL system. And it seems to be right:
653 (ROOM T) can bring a small computer to its knees for a *long*
654 time trying to GC afterwards. Surely there's some more economical
655 way to implement (ROOM T).
658 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
660 ;;; This file fails to compile.
661 ;;; Maybe this bug is related to bugs #65, #70 in the BUGS file.
662 (in-package :cl-user)
668 ;; Uncomment and it works
671 In SBCL 0.6.12.42, the problem is
672 internal error, failed AVER:
673 "(COMMON-LISP:EQ (SB!C::LAMBDA-TAIL-SET SB!C::CALLER)
674 (SB!C::LAMBDA-TAIL-SET (SB!C::LAMBDA-HOME SB!C::CALLEE)))"
677 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
679 ;;; The compiler is flushing the argument type test, and the default
680 ;;; case in the cond, so that calling with say a fixnum 0 causes a
682 (declaim (optimize (safety 2) (speed 3)))
684 (declare (type (or string stream) x))
685 (cond ((typep x 'string) 'string)
686 ((typep x 'stream) 'stream)
689 The symptom in sbcl-0.6.12.42 on OpenBSD is actually (TST 0)=>STREAM
690 (not the SIGBUS reported in the comment) but that's broken too;
691 type declarations are supposed to be treated as assertions unless
692 SAFETY 0, so we should be getting a TYPE-ERROR.
695 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
697 (in-package :cl-user)
698 ;;; From: David Gadbois <gadbois@cyc.com>
700 ;;; Logical pathnames aren't externalizable.
702 (let ((tempfile "/tmp/test.lisp"))
703 (setf (logical-pathname-translations "XXX")
704 '(("XXX:**;*.*" "/tmp/**/*.*")))
705 (with-open-file (out tempfile :direction :output)
706 (write-string "(defvar *path* #P\"XXX:XXX;FOO.LISP\")" out))
707 (compile-file tempfile))
708 The error message in sbcl-0.6.12.42 is
710 ; (while making load form for #<SB-IMPL::LOGICAL-HOST "XXX">)
711 ; A logical host can't be dumped as a constant: #<SB-IMPL::LOGICAL-HOST "XXX">
714 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
716 (in-package :cl-user)
717 ;;; The following invokes a compiler error.
718 (declaim (optimize (speed 2) (debug 3)))
721 (unwind-protect nil)))
725 The error message in sbcl-0.6.12.42 is
726 internal error, failed AVER:
727 "(COMMON-LISP:EQ (SB!C::TN-ENVIRONMENT SB!C:TN) SB!C::TN-ENV)"
730 When the compiler inline expands functions, it may be that different
731 kinds of return values are generated from different code branches.
732 E.g. an inline expansion of POSITION generates integer results
733 from one branch, and NIL results from another. When that inline
734 expansion is used in a context where only one of those results
737 (aref *a1* (position x *a2*)))
738 and the compiler can't prove that the unacceptable branch is
739 never taken, then bogus type mismatch warnings can be generated.
740 If you need to suppress the type mismatch warnings, you can
741 suppress the inline expansion,
743 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
744 (aref *a1* (position x *a2*)))
745 or, sometimes, suppress them by declaring the result to be of an
748 (aref *a1* (the integer (position x *a2*))))
750 This is not a new compiler problem in 0.7.0, but the new compiler
751 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
752 more conspicuous. If you don't need performance from these functions,
753 and the bogus warnings are a nuisance for you, you can return to
754 your pre-0.7.0 state of grace with
755 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
758 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
759 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
760 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
761 when of course it should be NIL. (He says it only fails for X86,
762 not SPARC; dunno about Alpha.)
764 Also, "the same problem exists for LONG-FLOAT-EPSILON,
765 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
766 for the -negative- the + is replaced by a - in the test)."
768 Raymond Toy comments that this is tricky on the X86 since its FPU
769 uses 80-bit precision internally.
772 The compiler incorrectly figures the return type of
773 (DEFUN FOO (FRAME UP-FRAME)
780 This problem exists in CMU CL 18c too. When I reported it on
781 cmucl-imp@cons.org, Raymond Toy replied 23 Aug 2001 with
782 a partial explanation, but no fix has been found yet.
785 Even in sbcl-0.pre7.x, which is supposed to be free of the old
786 non-ANSI behavior of treating the function return type inferred
787 from the current function definition as a declaration of the
788 return type from any function of that name, the return type of NIL
789 is attached to FOO in 120a above, and used to optimize code which
793 There was some sort of screwup in handling of
794 (IF (NOT (IGNORE-ERRORS ..))). E.g.
796 (if (not (ignore-errors
797 (make-pathname :host "foo" :directory "!bla" :name "bar")))
799 (error "notunlessnot")))
800 The (NOT (IGNORE-ERRORS ..)) form evaluates to T, so this should be
801 printing "ok", but instead it's going to the ERROR. This problem
802 seems to've been introduced by MNA's HANDLER-CASE patch (sbcl-devel
803 2001-07-17) and as a workaround (put in sbcl-0.pre7.14.flaky4.12)
804 I reverted back to the old weird HANDLER-CASE code. However, I
805 think the problem looks like a compiler bug in handling RETURN-FROM,
806 so I left the MNA-patched code in HANDLER-CASE (suppressed with
807 #+NIL) and I'd like to go back to see whether this really is
808 a compiler bug before I delete this BUGS entry.
811 The *USE-IMPLEMENTATION-TYPES* hack causes bugs, particularly
812 (IN-PACKAGE :SB-KERNEL)
813 (TYPE= (SPECIFIER-TYPE '(VECTOR T))
814 (SPECIFIER-TYPE '(VECTOR UNDEFTYPE)))
815 Then because of this, the compiler bogusly optimizes
816 (TYPEP #(11) '(SIMPLE-ARRAY UNDEF-TYPE 1))
817 to T. Unfortunately, just setting *USE-IMPLEMENTATION-TYPES* to
818 NIL around sbcl-0.pre7.14.flaky4.12 didn't work: the compiler complained
819 about type mismatches (probably harmlessly, another instance of bug 117);
820 and then cold init died with a segmentation fault.
823 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
824 the entire lexical environment at the point of MACROLET available
825 in the bodies of the macroexpander functions. In particular, it
826 allows the function bodies (which run at compile time) to try to
827 access lexical variables (which are only defined at runtime).
828 It doesn't even issue a warning, which is bad.
830 The SBCL behavior arguably conforms to the ANSI spec (since the
831 spec says that the behavior is undefined, ergo anything conforms).
832 However, it would be better to issue a compile-time error.
833 Unfortunately I (WHN) don't see any simple way to detect this
834 condition in order to issue such an error, so for the meantime
835 SBCL just does this weird broken "conforming" thing.
837 The ANSI standard says, in the definition of the special operator
839 The macro-expansion functions defined by MACROLET are defined
840 in the lexical environment in which the MACROLET form appears.
841 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
842 the local macro definitions in a MACROLET, but the consequences
843 are undefined if the local macro definitions reference any
844 local variable or function bindings that are visible in that
846 Then it seems to contradict itself by giving the example
848 (macrolet ((fudge (z)
849 ;The parameters x and flag are not accessible
850 ; at this point; a reference to flag would be to
851 ; the global variable of that name.
852 ` (if flag (* ,z ,z) ,z)))
853 ;The parameters x and flag are accessible here.
857 The comment "a reference to flag would be to the global variable
858 of the same name" sounds like good behavior for the system to have.
859 but actual specification quoted above says that the actual behavior
863 (as reported by Gabe Garza on cmucl-help 2001-09-21)
865 (defun test-pred (x y)
869 (func (lambda () x)))
870 (print (eq func func))
871 (print (test-pred func func))
872 (delete func (list func))))
873 Now calling (TEST-CASE) gives output
876 (#<FUNCTION {500A9EF9}>)
877 Evidently Python thinks of the lambda as a code transformation so
878 much that it forgets that it's also an object.
881 The DEFSTRUCT section of the ANSI spec, in the :CONC-NAME section,
882 specifies a precedence rule for name collisions between slot accessors of
883 structure classes related by inheritance. As of 0.7.0, SBCL still
887 insufficient syntax checking in MACROLET:
889 (macrolet ((defmacro bar (z) `(+ z z)))
891 shouldn't compile without error (because of the extra DEFMACRO symbol).
894 As of sbcl-0.pre7.86.flaky7.3, the cross-compiler, and probably
895 the CL:COMPILE function (which is based on the same %COMPILE
896 mechanism) get confused by
898 (labels ((sxhash-number (x)
900 (fixnum (sxhash x)) ; through DEFTRANSFORM
901 (integer (sb!bignum:sxhash-bignum x))
902 (single-float (sxhash x)) ; through DEFTRANSFORM
903 (double-float (sxhash x)) ; through DEFTRANSFORM
904 #!+long-float (long-float (error "stub: no LONG-FLOAT"))
905 (ratio (let ((result 127810327))
906 (declare (type fixnum result))
907 (mixf result (sxhash-number (numerator x)))
908 (mixf result (sxhash-number (denominator x)))
910 (complex (let ((result 535698211))
911 (declare (type fixnum result))
912 (mixf result (sxhash-number (realpart x)))
913 (mixf result (sxhash-number (imagpart x)))
915 (sxhash-recurse (x &optional (depthoid +max-hash-depthoid+))
916 (declare (type index depthoid))
920 (mix (sxhash-recurse (car x) (1- depthoid))
921 (sxhash-recurse (cdr x) (1- depthoid)))
924 (if (typep x 'structure-object)
926 (sxhash ; through DEFTRANSFORM
927 (class-name (layout-class (%instance-layout x)))))
929 (symbol (sxhash x)) ; through DEFTRANSFORM
930 (number (sxhash-number x))
933 (simple-string (sxhash x)) ; through DEFTRANSFORM
934 (string (%sxhash-substring x))
935 (bit-vector (let ((result 410823708))
936 (declare (type fixnum result))
937 (dotimes (i (min depthoid (length x)))
938 (mixf result (aref x i)))
940 (t (logxor 191020317 (sxhash (array-rank x))))))
943 (sxhash (char-code x)))) ; through DEFTRANSFORM
946 complaining "function called with two arguments, but wants exactly
947 one" about SXHASH-RECURSE. (This might not be strictly a new bug,
948 since IIRC post-fork CMU CL has also had problems with &OPTIONAL
949 arguments in FLET/LABELS: it might be an old Python bug which is
950 only exercised by the new arrangement of the SBCL compiler.)
953 Ideally, uninterning a symbol would allow it, and its associated
954 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
955 at least as of sbcl-0.7.0, this isn't the case. Information about
956 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
957 essentially in ordinary (non-weak) hash tables keyed by symbols.
958 Thus, once a system has an entry in this system, it tends to live
959 forever, even when it is uninterned and all other references to it
963 (reported by Arnaud Rouanet on cmucl-imp 2001-12-18)
964 (defmethod foo ((x integer))
966 (defmethod foo :around ((x integer))
969 Now (FOO 3) should return 3, but instead it returns 4.
972 (reported by Alexey Dejneka sbcl-devel 2002-01-03)
974 SUBTYPEP does not work well with redefined classes:
983 * (defclass b (a) ())
996 This bug was fixed in sbcl-0.7.4.1 by invalidating the PCL wrapper
997 class upon redefinition. Unfortunately, doing so causes bug #176 to
998 appear. Pending further investication, one or other of these bugs
999 might be present at any given time.
1002 Pretty-printing nested backquotes doesn't work right, as
1003 reported by Alexey Dejneka sbcl-devel 2002-01-13:
1005 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
1006 * (lisp-implementation-version)
1010 (as reported by Lynn Quam on cmucl-imp ca. 2002-01-16)
1011 %NATURALIZE-C-STRING conses a lot, like 16 bytes per byte
1012 of the naturalized string. We could probably port the patches
1013 from the cmucl-imp mailing list.
1016 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
1017 prominent SourceForge web/db bug tracking system, which is
1018 unfortunately not a reliable way to get a timely response from
1019 the SBCL maintainers)
1020 In the course of trying to build a test case for an
1021 application error, I encountered this behavior:
1022 If you start up sbcl, and then lay on CTRL-C for a
1023 minute or two, the lisp process will eventually say:
1024 %PRIMITIVE HALT called; the party is over.
1025 and throw you into the monitor. If I start up lisp,
1026 attach to the process with strace, and then do the same
1027 (abusive) thing, I get instead:
1028 access failure in heap page not marked as write-protected
1029 and the monitor again. I don't know enough to have the
1030 faintest idea of what is going on here.
1031 This is with sbcl 6.12, uname -a reports:
1032 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
1033 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
1034 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
1035 it took more than two minutes (but less than five) for me.
1038 (This was once known as IR1-4, but it lived on even after the
1039 IR1 interpreter went to the big bit bucket in the sky.)
1040 The system accepts DECLAIM in most places where DECLARE would be
1041 accepted, without even issuing a warning. ANSI allows this, but since
1042 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
1043 meaning is rather different, and it's unlikely that the user
1044 has a good reason for doing DECLAIM not at top level, it would be
1045 good to issue a STYLE-WARNING when this happens. A possible
1046 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
1047 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
1048 [This is considered an IR1-interpreter-related bug because until
1049 EVAL-WHEN is rewritten, which won't happen until after the IR1
1050 interpreter is gone, the system's notion of what's a top-level form
1051 and what's not will remain too confused to fix this problem.]
1054 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
1055 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
1056 COMPLEX implementation didn't deal with this, and hasn't been
1057 upgraded to do so. (This doesn't seem to be a high priority
1058 conformance problem, since seems hard to construct useful code
1062 Floating point errors are reported poorly. E.g. on x86 OpenBSD
1065 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
1066 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
1067 No traps are enabled? How can this be?
1068 It should be possible to be much more specific (overflow, division
1069 by zero, etc.) and of course the "How can this be?" should be fixable.
1072 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
1073 (in-package :cl-user)
1076 (defstruct foo bar bletch)
1078 (labels ((kidify1 (kid)
1083 (m+ (frobnicate kid)
1086 (declare (inline kid-frob))
1089 (the simple-vector (foo-bar perd)))))
1091 debugger invoked on condition of type TYPE-ERROR:
1092 The value NIL is not of type SB-C::NODE.
1093 The location of this failure has moved around as various related
1094 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
1095 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
1098 (essentially the same problem as a CMU CL bug reported by Martin
1099 Cracauer on cmucl-imp 2002-02-19)
1100 There is a hole in structure slot type checking. Compiling and LOADing
1101 (declaim (optimize safety))
1103 (bla 0 :type fixnum))
1105 (let ((foo (make-foo)))
1106 (setf (foo-bla foo) '(1 . 1))
1107 (format t "Is ~a of type ~a a cons? => ~a~%"
1109 (type-of (foo-bla foo))
1110 (consp (foo-bla foo)))))
1112 should signal an error, but in sbcl-0.7.1.21 instead gives the output
1113 Is (1 . 1) of type CONS a cons? => NIL
1114 without signalling an error.
1117 There's some sort of problem with aborting back out of the debugger
1118 after a %DETECT-STACK-EXHAUSTION error in sbcl-0.7.1.38. In some cases
1119 telling the debugger to ABORT doesn't get you back to the main REPL,
1120 but instead just gives you another stack exhaustion error. The problem
1121 doesn't occur in the trivial case
1122 * (defun frob () (frob) (frob))
1125 but it has happened in more complicated cases (which I haven't
1126 figured out how to reproduce).
1129 Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
1130 UPGRADED-COMPLEX-PART-TYPE should have an optional environment argument.
1131 (reported by Alexey Dejneka sbcl-devel 2002-04-12)
1134 (reported by Robert E. Brown 2002-04-16)
1135 When a function is called with too few arguments, causing the
1136 debugger to be entered, the uninitialized slots in the bad call frame
1137 seem to cause GCish problems, being interpreted as tagged data even
1138 though they're not. In particular, executing ROOM in the
1139 debugger at that point causes AVER failures:
1142 * (lisp-implementation-version)
1148 failed AVER: "(SAP= CURRENT END)"
1149 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
1150 isn't too surprising since there are many differences in stack
1151 implementation and GC conservatism between the X86 and other ports.)
1154 Array types with element-types of some unknown type are falsely being
1155 assumed to be of type (ARRAY T) by the compiler in some cases. The
1156 following code demonstrates the problem:
1159 (declare (type (vector bar) x))
1161 (deftype bar () 'single-float)
1162 (foo (make-array 3 :element-type 'bar))
1163 -> TYPE-ERROR "The value #(0.0 0.0 0.0) is not of type (VECTOR BAR)."
1164 (typep (make-array 3 :element-type 'bar) '(vector bar))
1167 The easy solution is to make the functions which depend on knowing
1168 the upgraded-array-element-type (in compiler/array-tran and
1169 compiler/generic/vm-tran as of sbcl-0.7.3.x) be slightly smarter about
1170 unknown types; an alternative is to have the
1171 specialized-element-type slot in the ARRAY-TYPE structure be
1172 *WILD-TYPE* for UNKNOWN-TYPE element types.
1176 (in-package :cl-user)
1178 (defmethod permanentize ((uustk uustk))
1179 (flet ((frob (hash-table test-for-deletion)
1181 (obj-entry.stale? (oe)
1182 (destructuring-bind (key . datum) oe
1183 (declare (type simple-vector key))
1184 (deny0 (void? datum))
1185 (some #'stale? key))))
1186 (declare (inline frob obj-entry.stale?))
1187 (frob (uustk.args-hash->obj-alist uustk)
1189 (frob (uustk.hash->memoized-objs-list uustk)
1192 in sbcl-0.7.3.11 causes an assertion failure,
1195 (AND (NULL (BLOCK-SUCC B))
1196 (NOT (BLOCK-DELETE-P B))
1197 (NOT (EQ B (COMPONENT-HEAD #)))))"
1200 In sbcl-0.7.3.11, compiling the (illegal) code
1201 (in-package :cl-user)
1202 (defmethod prove ((uustk uustk))
1203 (zap ((frob () nil))
1205 gives the (not terribly clear) error message
1207 ; (during macroexpansion of (DEFMETHOD PROVE ...))
1208 ; can't get template for (FROB NIL NIL)
1209 The problem seems to be that the code walker used by the DEFMETHOD
1210 macro is unhappy with the illegal syntax in the method body, and
1211 is giving an unclear error message.
1214 (reported by Dan Barlow on sbcl-devel 2002-05-10)
1215 In sbcl-0.7.3.12, doing
1216 (defstruct foo bar baz)
1217 (compile nil (lambda (x) (or x (foo-baz x))))
1219 debugger invoked on condition of type SB-INT:BUG:
1220 full call to SB-KERNEL:%INSTANCE-REF
1221 This is probably a bug in SBCL itself. [...]
1222 Since this is a reasonable user error, it shouldn't be reported as
1226 (reported by Pierre Mai while investigating bug 47):
1227 (DEFCLASS FOO () ((A :SILLY T)))
1228 signals a SIMPLE-ERROR, not a PROGRAM-ERROR.
1231 sbcl's treatment of at least macro lambda lists is too permissive;
1232 e.g., in sbcl-0.7.3.7:
1233 (defmacro foo (&rest rest bar) `(,bar ,rest))
1234 (macroexpand '(foo quux zot)) -> (QUUX (QUUX ZOT))
1235 whereas section 3.4.4 of the CLHS doesn't allow required parameters
1236 to come after the rest argument.
1239 The compiler sometimes tries to constant-fold expressions before
1240 it checks to see whether they can be reached. This can lead to
1241 bogus warnings about errors in the constant folding, e.g. in code
1244 (WRITE-STRING (> X 0) "+" "0"))
1245 compiled in a context where the compiler can prove that X is NIL,
1246 and the compiler complains that (> X 0) causes a type error because
1247 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
1248 caused a full WARNING, which made the bug really annoying because then
1249 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
1250 code. Since then the warning has been downgraded to STYLE-WARNING,
1251 so it's still a bug but at least it's a little less annoying.
1254 The error message from attempting to use a #\Return format
1256 (format nil "~^M") ; replace "^M" with a literal #\Return
1257 debugger invoked on condition of type SB-FORMAT::FORMAT-ERROR:
1258 error in format: unknown format directive
1261 is not terribly helpful; this is more noticeable than parallel cases
1262 with e.g. #\Backspace because of the differing newline conventions
1263 on various operating systems. (reported by Harald Hanche-Olsen on
1264 cmucl-help 2002-05-31)
1267 reported by Alexey Dejneka 08 Jun 2002 in sbcl-devel:
1268 Playing with McCLIM, I've received an error "Unbound variable WRAPPER
1269 in SB-PCL::CHECK-WRAPPER-VALIDITY".
1270 (defun check-wrapper-validity (instance)
1271 (let* ((owrapper (wrapper-of instance)))
1272 (if (not (invalid-wrapper-p owrapper))
1274 (let* ((state (wrapper-state wrapper)) ; !!!
1276 I've tried to replace it with OWRAPPER, but now OBSOLETE-INSTANCE-TRAP
1277 breaks with "NIL is not of type SB-KERNEL:LAYOUT".
1279 partial fix: The undefined variable WRAPPER resulted from an error
1280 in recent refactoring, as can be seen by comparing to the code in e.g.
1281 sbcl-0.7.2. Replacing WRAPPER with OWRAPPER (done by WHN in sbcl-0.7.4.22)
1282 should bring the code back to its behavior as of sbcl-0.7.2, but
1283 that still leaves the OBSOLETE-INSTANCE-TRAP bug. An example of
1284 input which triggers that bug is
1286 (let ((lastname (intern (format nil "C~D" (1- i))))
1287 (name (intern (format nil "C~D" i))))
1288 (eval `(defclass ,name
1289 (,@(if (= i 0) nil (list lastname)))
1291 (eval `(defmethod initialize-instance :after ((x ,name) &rest any)
1292 (declare (ignore any))))))
1294 (defclass c0 (b) ())
1295 (make-instance 'c19)
1299 178: "AVER failure compiling confused THEs in FUNCALL"
1300 In sbcl-0.7.4.24, compiling
1302 (funcall (the function (the standard-object x))))
1305 "(AND (EQ (IR2-CONTINUATION-PRIMITIVE-TYPE 2CONT) FUNCTION-PTYPE) (EQ CHECK T))"
1306 This variant compiles OK, though:
1307 (defun bug178alternative (x)
1308 (funcall (the nil x)))
1310 181: "bad type specifier drops compiler into debugger"
1312 (in-package :cl-user)
1314 (declare (type 0 x))
1317 bad thing to be a type specifier: 0
1318 which seems fine, but also enters the debugger (instead of having
1319 the compiler handle the error, convert it into a COMPILER-ERROR, and
1320 continue compiling) which seems wrong.
1322 182: "SPARC/Linux floating point"
1323 Evaluating (/ 1.0 0.0) at the prompt causes a Bus error and complete
1324 death of the environment. Other floating point operations sometimes
1325 return infinities when they should raise traps (e.g. the test case
1326 for bug #146, (expt 2.0 12777)).
1328 183: "IEEE floating point issues"
1329 Even where floating point handling is being dealt with relatively
1330 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
1331 accrued-exceptions and current-exceptions part of the fp control
1332 word don't seem to bear much relation to reality. E.g. on
1336 debugger invoked on condition of type DIVISION-BY-ZERO:
1337 arithmetic error DIVISION-BY-ZERO signalled
1338 0] (sb-vm::get-floating-point-modes)
1340 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1341 :ROUNDING-MODE :NEAREST
1342 :CURRENT-EXCEPTIONS NIL
1343 :ACCRUED-EXCEPTIONS (:INEXACT)
1346 * (sb-vm::get-floating-point-modes)
1347 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1348 :ROUNDING-MODE :NEAREST
1349 :CURRENT-EXCEPTIONS (:INEXACT)
1350 :ACCRUED-EXCEPTIONS (:INEXACT)
1353 184: "division by zero becomes frozen into RATIO"
1354 (reported by Wolfhard Buss on cmucl-imp 18 Jun 2002, fails on
1359 185: "top-level forms at the REPL"
1360 * (locally (defstruct foo (a 0 :type fixnum)))
1363 ; (in macroexpansion of (SB-KERNEL::%DELAYED-GET-COMPILER-LAYOUT BAR))
1364 however, compiling and loading the same expression in a file works
1367 187: "type inference confusion around DEFTRANSFORM time"
1368 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
1369 bug in DEFTRANSFORM")
1370 After the file below is compiled and loaded in sbcl-0.7.5, executing
1371 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
1372 at the REPL returns an adjustable vector, which is wrong. Presumably
1373 somehow the DERIVE-TYPE information for the output values of %WAD is
1374 being mispropagated as a type constraint on the input values of %WAD,
1375 and so causing the type test to be optimized away. It's unclear how
1376 hand-expanding the DEFTRANSFORM would change this, but it suggests
1377 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
1378 invoked at a particular phase) is involved.
1379 (cl:in-package :sb-c)
1380 (eval-when (:compile-toplevel)
1381 ;;; standin for %DATA-VECTOR-AND-INDEX
1382 (defknown %dvai (array index)
1384 (foldable flushable))
1385 (deftransform %dvai ((array index)
1389 (let* ((atype (continuation-type array))
1390 (eltype (array-type-specialized-element-type atype)))
1391 (when (eq eltype *wild-type*)
1392 (give-up-ir1-transform
1393 "specialized array element type not known at compile-time"))
1394 (when (not (array-type-complexp atype))
1395 (give-up-ir1-transform "SIMPLE array!"))
1396 `(if (array-header-p array)
1397 (%wad array index nil)
1398 (values array index))))
1399 ;;; standin for %WITH-ARRAY-DATA
1400 (defknown %wad (array index (or index null))
1401 (values (simple-array * (*)) index index index)
1402 (foldable flushable))
1403 ;;; (Commenting out this optimizer causes the bug to go away.)
1404 (defoptimizer (%wad derive-type) ((array start end))
1405 (let ((atype (continuation-type array)))
1406 (when (array-type-p atype)
1407 (values-specifier-type
1408 `(values (simple-array ,(type-specifier
1409 (array-type-specialized-element-type atype))
1411 index index index)))))
1413 (defun %wad (array start end)
1414 (format t "~&in %WAD~%")
1415 (%with-array-data array start end))
1416 (cl:in-package :cl-user)
1418 (declare (type (vector t) v))
1419 (declare (notinline sb-kernel::%with-array-data))
1420 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
1424 188: "compiler performance fiasco involving type inference and UNION-TYPE"
1425 In sbcl-0.7.5.11 on a 700 MHz Pentium III,
1429 (declare (optimize (safety 3)))
1430 (declare (optimize (compilation-speed 2)))
1431 (declare (optimize (speed 1) (debug 1) (space 1)))
1432 (let ((fn "if-this-file-exists-the-universe-is-strange"))
1433 (load fn :if-does-not-exist nil)
1434 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
1435 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
1436 (load (concatenate 'string fn ".misc-garbage")
1437 :if-does-not-exist nil)))))
1439 134.552 seconds of real time
1440 133.35156 seconds of user run time
1441 0.03125 seconds of system run time
1442 [Run times include 2.787 seconds GC run time.]
1444 246883368 bytes consed.
1445 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
1446 thinking about type relationships involving types like
1448 (OR (INTEGER 576 576)
1459 DEFUNCT CATEGORIES OF BUGS
1461 These labels were used for bugs related to the old IR1 interpreter.
1462 The # values reached 6 before the category was closed down.