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. ANSI justifies this specification by saying
54 The restriction against issuing a warning for type mismatches
55 between a slot-initform and the corresponding slot's :TYPE
56 option is necessary because a slot-initform must be specified
57 in order to specify slot options; in some cases, no suitable
59 In SBCL, as in CMU CL (or, for that matter, any compiler which
60 really understands Common Lisp types) a suitable default does
61 exist, in all cases, because the compiler understands the concept
62 of functions which never return (i.e. has return type NIL, e.g.
63 ERROR). Thus, as a portable workaround, you can use a call to
64 some known-never-to-return function as the default. E.g.
66 (BAR (ERROR "missing :BAR argument")
67 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
69 (DECLAIM (FTYPE () NIL) MISSING-ARG)
70 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
71 (ERROR "missing required argument"))
73 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
74 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
75 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
76 Such code will compile without complaint and work correctly either
77 on SBCL or on a completely compliant Common Lisp system.
80 bogus warnings about undefined functions for magic functions like
81 SB!C::%%DEFUN and SB!C::%DEFCONSTANT when cross-compiling files
82 like src/code/float.lisp. Fixing this will probably require
83 straightening out enough bootstrap consistency issues that
84 the cross-compiler can run with *TYPE-SYSTEM-INITIALIZED*.
85 Instead, the cross-compiler runs in a slightly flaky state
86 which is sane enough to compile SBCL itself, but which is
87 also unstable in several ways, including its inability
88 to really grok function declarations.
91 The "byte compiling top-level form:" output ought to be condensed.
92 Perhaps any number of such consecutive lines ought to turn into a
93 single "byte compiling top-level forms:" line.
96 The way that the compiler munges types with arguments together
97 with types with no arguments (in e.g. TYPE-EXPAND) leads to
98 weirdness visible to the user:
99 (DEFTYPE FOO () 'FIXNUM)
101 (TYPEP 11 '(FOO)) => T, which seems weird
102 (TYPEP 11 'FIXNUM) => T
103 (TYPEP 11 '(FIXNUM)) signals an error, as it should
104 The situation is complicated by the presence of Common Lisp types
105 like UNSIGNED-BYTE (which can either be used in list form or alone)
106 so I'm not 100% sure that the behavior above is actually illegal.
107 But I'm 90+% sure, and the following related behavior,
109 treating the bare symbol AND as equivalent to '(AND), is specifically
110 forbidden (by the ANSI specification of the AND type).
113 It would be nice if the
115 (during macroexpansion)
116 said what macroexpansion was at fault, e.g.
118 (during macroexpansion of IN-PACKAGE,
119 during macroexpansion of DEFFOO)
122 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
123 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
124 (Also, when this is fixed, we can enable the code in PROCLAIM which
125 checks for incompatible FTYPE redeclarations.)
128 from DTC on the CMU CL mailing list 25 Feb 2000:
129 ;;; Compiler fails when this file is compiled.
131 ;;; Problem shows up in delete-block within ir1util.lisp. The assertion
132 ;;; (assert (member (functional-kind lambda) '(:let :mv-let :assignment)))
133 ;;; fails within bind node branch.
135 ;;; Note that if c::*check-consistency* is enabled then an un-reached
136 ;;; entry is also reported.
139 (declare (values nil))
156 (let ((ttt #'(lambda () (go cccc))))
157 (declare (special ttt))
158 (return-from bbbb nil))
161 (return-from bbbb nil))))))
164 (I *think* this is a bug. It certainly seems like strange behavior. But
165 the ANSI spec is scary, dark, and deep..)
166 (FORMAT NIL "~,1G" 1.4) => "1. "
167 (FORMAT NIL "~3,1G" 1.4) => "1. "
170 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
172 (setf (find-class 'ccc1) (find-class 'ccc))
173 (defmethod zut ((c ccc1)) 123)
174 DTC's recommended workaround from the mailing list 3 Mar 2000:
175 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
178 The ANSI spec, in section "22.3.5.2 Tilde Less-Than-Sign: Logical Block",
179 says that an error is signalled if ~W, ~_, ~<...~:>, ~I, or ~:T is used
180 inside "~<..~>" (without the colon modifier on the closing syntax).
181 However, SBCL doesn't do this:
182 * (FORMAT T "~<munge~wegnum~>" 12)
187 When too many files are opened, OPEN will fail with an
188 uninformative error message
189 error in function OPEN: error opening #P"/tmp/foo.lisp": NIL
190 instead of saying that too many files are open.
193 reported by Sam Steingold on the cmucl-imp mailing list 12 May 2000:
194 Also, there is another bug: `array-displacement' should return an
195 array or nil as first value (as per ANSI CL), while CMUCL declares
196 it as returning an array as first value always.
197 (Actually, I think the old CMU CL version in SBCL never returns NIL,
198 i.e. it's not just a declaration problem, but the definition doesn't
202 Sometimes (SB-EXT:QUIT) fails with
203 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
204 Process inferior-lisp exited abnormally with code 1
205 I haven't noticed a repeatable case of this yet.
208 some sort of bug in inlining and RETURN-FROM in sbcl-0.6.5: Compiling
211 (BLOCK USED-BY-SOME-Y?
214 (UNLESS (REJECTED? Y)
215 (RETURN-FROM USED-BY-SOME-Y? T)))))
216 (DECLARE (INLINE FROB))
221 error in function SB-KERNEL:ASSERT-ERROR:
222 The assertion (EQ (SB-C::CONTINUATION-KIND SB-C::CONT) :BLOCK-START) failed.
223 This is still present in sbcl-0.6.8.
226 In some cases the compiler believes type declarations on array
227 elements without checking them, e.g.
228 (DECLAIM (OPTIMIZE (SAFETY 3) (SPEED 1) (SPACE 1)))
231 (DECLARE (TYPE (SIMPLE-ARRAY CONS 1) X))
232 (WHEN (CONSP (AREF X 0))
234 (BAR (VECTOR (MAKE-FOO :A 11 :B 12)))
237 in SBCL 0.6.5 (and also in CMU CL 18b). This does not happen for
238 all cases, e.g. the type assumption *is* checked if the array
239 elements are declared to be of some structure type instead of CONS.
242 The printer doesn't report closures very well. This is true in
246 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
247 It would be nice to make closures have a settable name slot,
248 and make things like DEFSTRUCT and FLET, which create closures,
249 set helpful values into this slot.
252 And as long as we're wishing, it would be awfully nice if INSPECT could
253 also report on closures, telling about the values of the bound variables.
256 The compiler assumes that any time a function of declared FTYPE
257 doesn't signal an error, its arguments were of the declared type.
258 E.g. compiling and loading
259 (DECLAIM (OPTIMIZE (SAFETY 3)))
260 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
261 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE) FACTORIAL)))
263 (COND ((> (FACTORIAL X) 1.0E6)
264 (FORMAT T "too big~%"))
266 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
268 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
271 will cause the INTEGERP case to be selected, giving bogus output a la
273 This violates the "declarations are assertions" principle.
274 According to the ANSI spec, in the section "System Class FUNCTION",
275 this is a case of "lying to the compiler", but the lying is done
276 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
277 not by the unexpectedly general definition of FACTORIAL. In any case,
278 "declarations are assertions" means that lying to the compiler should
279 cause an error to be signalled, and should not cause a bogus
280 result to be returned. Thus, the compiler should not assume
281 that arbitrary functions check their argument types. (It might
282 make sense to add another flag (CHECKED?) to DEFKNOWN to
283 identify functions which *do* check their argument types.)
286 DEFMETHOD doesn't check the syntax of &REST argument lists properly,
287 accepting &REST even when it's not followed by an argument name:
288 (DEFMETHOD FOO ((X T) &REST) NIL)
291 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
292 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
294 (DECLARE (TYPE INDEXOID X))
295 (THE (VALUES INDEXOID)
297 where the implementation of the type check in function FOO
298 includes a full call to %TYPEP. There are also some fundamental problems
299 with the interpretation of VALUES types (inherited from CMU CL, and
300 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
301 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
304 The definitions of SIGCONTEXT-FLOAT-REGISTER and
305 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
306 supported on FreeBSD because the floating point state is not saved,
307 but at least as of FreeBSD 4.0, the floating point state *is* saved,
308 so they could be supported after all. Very likely
309 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
312 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
313 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
314 issue with 'union, 'and, 'or etc. So even though according to the
315 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
316 (and now SBCL) interpret them as legal types.
319 ANSI specifies DEFINE-SYMBOL-MACRO, but it's not defined in SBCL.
320 CMU CL added it ca. Aug 13, 2000, after some discussion on the mailing
321 list, and it is probably possible to use substantially the same
322 patches to add it to SBCL.
325 a slew of floating-point-related errors reported by Peter Van Eynde
327 a: (fixed in sbcl-0.6.11.25)
328 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
329 should probably be 1.4012985e-45. In SBCL,
330 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
331 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
332 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
333 and LEAST-NEGATIVE-LONG-FLOAT.
334 c: Many expressions generate floating infinity:
339 PVE's regression tests want them to raise errors. SBCL
340 generates the infinities instead, which may or may not be
342 d: (in section12.erg) various forms a la
343 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
344 don't give the right behavior.
347 type safety errors reported by Peter Van Eynde July 25, 2000:
348 a: (COERCE (QUOTE (A B C)) (QUOTE (VECTOR * 4)))
350 In general lengths of array type specifications aren't
351 checked by COERCE, so it fails when the spec is
352 (VECTOR 4), (STRING 2), (SIMPLE-BIT-VECTOR 3), or whatever.
353 b: CONCATENATE has the same problem of not checking the length
354 of specified output array types. MAKE-SEQUENCE and MAP and
355 MERGE also have the same problem.
356 c: (COERCE 'AND 'FUNCTION) returns something related to
357 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
358 d: ELT signals SIMPLE-ERROR if its index argument
359 isn't a valid index for its sequence argument, but should
360 signal TYPE-ERROR instead.
361 e: FILE-LENGTH is supposed to signal a type error when its
362 argument is not a stream associated with a file, but doesn't.
363 f: (FLOAT-RADIX 2/3) should signal an error instead of
365 g: (LOAD "*.lsp") should signal FILE-ERROR.
366 h: (MAKE-CONCATENATED-STREAM (MAKE-STRING-OUTPUT-STREAM))
367 should signal TYPE-ERROR.
368 i: MAKE-TWO-WAY-STREAM doesn't check that its arguments can
369 be used for input and output as needed. It should fail with
370 TYPE-ERROR when handed e.g. the results of
371 MAKE-STRING-INPUT-STREAM or MAKE-STRING-OUTPUT-STREAM in
372 the inappropriate positions, but doesn't.
373 j: (PARSE-NAMESTRING (COERCE (LIST #\f #\o #\o (CODE-CHAR 0) #\4 #\8)
375 should probably signal an error instead of making a pathname with
377 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
378 not a binary input stream, but instead cheerfully reads from
379 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
382 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
383 a: (DEFCLASS FOO () (A B A)) should signal a PROGRAM-ERROR, and
385 b: (DEFCLASS FOO () (A B A) (:DEFAULT-INITARGS X A X B)) should
386 signal a PROGRAM-ERROR, and doesn't.
387 c: (DEFCLASS FOO07 NIL ((A :ALLOCATION :CLASS :ALLOCATION :CLASS))),
388 and other DEFCLASS forms with duplicate specifications in their
389 slots, should signal a PROGRAM-ERROR, and doesn't.
390 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
391 causes a COMPILER-ERROR.
394 SYMBOL-MACROLET bugs reported by Peter Van Eynde July 25, 2000:
395 a: (SYMBOL-MACROLET ((T TRUE)) ..) should probably signal
396 PROGRAM-ERROR, but SBCL accepts it instead.
397 b: SYMBOL-MACROLET should refuse to bind something which is
398 declared as a global variable, signalling PROGRAM-ERROR.
399 c: SYMBOL-MACROLET should signal PROGRAM-ERROR if something
400 it binds is declared SPECIAL inside.
403 LOOP bugs reported by Peter Van Eynde July 25, 2000:
404 a: (LOOP WITH (A B) DO (PRINT 1)) is a syntax error according to
405 the definition of WITH clauses given in the ANSI spec, but
406 compiles and runs happily in SBCL.
407 b: a messy one involving package iteration:
408 interpreted Form: (LET ((PACKAGE (MAKE-PACKAGE "LOOP-TEST"))) (INTERN "blah" PACKAGE) (LET ((BLAH2 (INTERN "blah2" PACKAGE))) (EXPORT BLAH2 PACKAGE)) (LIST (SORT (LOOP FOR SYM BEING EACH PRESENT-SYMBOL OF PACKAGE FOR SYM-NAME = (SYMBOL-NAME SYM) COLLECT SYM-NAME) (FUNCTION STRING<)) (SORT (LOOP FOR SYM BEING EACH EXTERNAL-SYMBOL OF PACKAGE FOR SYM-NAME = (SYMBOL-NAME SYM) COLLECT SYM-NAME) (FUNCTION STRING<))))
409 Should be: (("blah" "blah2") ("blah2"))
410 SBCL: (("blah") ("blah2"))
411 * (LET ((X 1)) (LOOP FOR I BY (INCF X) FROM X TO 10 COLLECT I))
412 doesn't work -- SBCL's LOOP says BY isn't allowed in a FOR clause.
415 type system errors reported by Peter Van Eynde July 25, 2000:
416 a: (SUBTYPEP 'BIGNUM 'INTEGER) => NIL, NIL
417 but should be (VALUES T T) instead.
418 b: (SUBTYPEP 'EXTENDED-CHAR 'CHARACTER) => NIL, NIL
419 but should be (VALUES T T) instead.
420 c: (SUBTYPEP '(INTEGER (0) (0)) 'NIL) dies with nested errors.
421 d: In general, the system doesn't like '(INTEGER (0) (0)) -- it
422 blows up at the level of SPECIFIER-TYPE with
423 "Lower bound (0) is greater than upper bound (0)." Probably
424 SPECIFIER-TYPE should return NIL instead.
425 e: (TYPEP 0 '(COMPLEX (EQL 0)) fails with
426 "Component type for Complex is not numeric: (EQL 0)."
427 This might be easy to fix; the type system already knows
428 that (SUBTYPEP '(EQL 0) 'NUMBER) is true.
429 f: The type system doesn't know about the condition system,
430 so that e.g. (TYPEP 'SIMPLE-ERROR 'ERROR)=>NIL.
431 g: The type system isn't all that smart about relationships
432 between hairy types, as shown in the type.erg test results,
433 e.g. (SUBTYPEP 'CONS '(NOT ATOM)) => NIL, NIL.
436 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
438 (DEFGENERIC FOO02 (X))
439 (DEFMETHOD FOO02 ((X NUMBER)) T)
440 (LET ((M (FIND-METHOD (FUNCTION FOO02)
442 (LIST (FIND-CLASS (QUOTE NUMBER))))))
443 (REMOVE-METHOD (FUNCTION FOO02) M)
444 (DEFGENERIC FOO03 (X))
445 (ADD-METHOD (FUNCTION FOO03) M)))
446 should give an error, but SBCL allows it.
447 b: READ should probably return READER-ERROR, not the bare
448 arithmetic error, when input a la "1/0" or "1e1000" causes
452 It has been reported (e.g. by Peter Van Eynde) that there are
453 several metaobject protocol "errors". (In order to fix them, we might
454 need to document exactly what metaobject protocol specification
455 we're following -- the current code is just inherited from PCL.)
458 another error from Peter Van Eynde 5 September 2000:
459 (FORMAT NIL "~F" "FOO") should work, but instead reports an error.
460 PVE submitted a patch to deal with this bug, but it exposes other
461 comparably serious bugs, so I didn't apply it. It looks as though
462 the FORMAT code needs a fair amount of rewriting in order to comply
463 with the various details of the ANSI spec.
466 The implementation of #'+ returns its single argument without
467 type checking, e.g. (+ "illegal") => "illegal".
470 In sbcl-0.6.7, there is no doc string for CL:PUSH, probably
471 because it's defined with the DEFMACRO-MUNDANELY macro and something
472 is wrong with doc string setting in that macro.
475 Attempting to use COMPILE on something defined by DEFMACRO fails:
476 (DEFMACRO FOO (X) (CONS X X))
478 Error in function C::GET-LAMBDA-TO-COMPILE:
479 #<Closure Over Function "DEFUN (SETF MACRO-FUNCTION)" {480E21B1}> was defined in a non-null environment.
482 (SUBTYPEP '(AND ZILCH INTEGER) 'ZILCH) => NIL, NIL
483 Note: I looked into fixing this in 0.6.11.15, but gave up. The
484 problem seems to be that there are two relevant type methods for
485 the subtypep operation, HAIRY :COMPLEX-SUBTYPEP-ARG2 and
486 INTERSECTION :COMPLEX-SUBTYPEP-ARG1, and only the first is
487 called. This could be fixed, but type dispatch is messy and
488 confusing enough already, I don't want to complicate it further.
489 Perhaps someday we can make CLOS cross-compiled (instead of compiled
490 after bootstrapping) so that we don't need to have the type system
491 available before CLOS, and then we can rewrite the type methods to
492 CLOS methods, and then expressing the solutions to stuff like this
493 should become much more straightforward. -- WHN 2001-03-14
496 CL:*DEFAULT-PATHNAME-DEFAULTS* doesn't behave as ANSI suggests (reflecting
497 current working directory). And there's no supported way to update
498 or query the current working directory (a la Unix "chdir" and "pwd"),
499 which is functionality that ILISP needs (and currently gets with low-level
501 When this is fixed, probably the more-or-less-parallel Unix-level
504 %SET-DEFAULT-DIRECTORY
506 should go away. Also we need to figure out what's the proper way to
507 deal with the interaction of users assigning new values to
508 *DEFAULT-PATHNAME-DEFAULTS* and cores being saved and restored.
509 (Perhaps just make restoring from a save always overwrite the old
510 value with the new Unix-level default directory?)
513 The debugger LIST-LOCATIONS command doesn't work properly.
516 Compiling and loading
517 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
519 then requesting a BACKTRACE at the debugger prompt gives no information
520 about where in the user program the problem occurred.
523 The compiler is supposed to do type inference well enough that
526 ((SIMPLE-ARRAY SINGLE-FLOAT)
528 (DECLARE (TYPE (SIMPLE-ARRAY SINGLE-FLOAT) X))
531 is redundant. However, as reported by Juan Jose Garcia Ripoll for
532 CMU CL, it sometimes doesn't. Adding declarations is a pretty good
533 workaround for the problem for now, but can't be done by the TYPECASE
534 macros themselves, since it's too hard for the macro to detect
535 assignments to the variable within the clause.
536 Note: The compiler *is* smart enough to do the type inference in
537 many cases. This case, derived from a couple of MACROEXPAND-1
538 calls on Ripoll's original test case,
540 (DECLARE (OPTIMIZE SPEED (SAFETY 0)))
541 (COND ((TYPEP A '(SIMPLE-ARRAY SINGLE-FLOAT)) NIL
542 (LET ((LENGTH (ARRAY-TOTAL-SIZE A)))
543 (LET ((I 0) (G2554 LENGTH))
544 (DECLARE (TYPE REAL G2554) (TYPE REAL I))
547 (WHEN (>= I G2554) (GO SB-LOOP::END-LOOP))
548 (SETF (ROW-MAJOR-AREF A I) (- (ROW-MAJOR-AREF A I)))
549 (GO SB-LOOP::NEXT-LOOP)
550 SB-LOOP::END-LOOP))))))
551 demonstrates the problem; but the problem goes away if the TAGBODY
552 and GO forms are removed (leaving the SETF in ordinary, non-looping
553 code), or if the TAGBODY and GO forms are retained, but the
554 assigned value becomes 0.0 instead of (- (ROW-MAJOR-AREF A I)).
557 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
558 I am looking into this problem that showed up on the cmucl-help
559 list. It seems to me that the "implementation specific environment
560 hacking functions" found in pcl/walker.lisp are completely messed
561 up. The good thing is that they appear to be barely used within
562 PCL and the munged environment object is passed to cmucl only
563 in calls to macroexpand-1, which is probably why this case fails.
564 SBCL uses essentially the same code, so if the environment hacking
565 is screwed up, it affects us too.
568 Using the pretty-printer from the command prompt gives funny
569 results, apparently because the pretty-printer doesn't know
570 about user's command input, including the user's carriage return
571 that the user, and therefore the pretty-printer thinks that
572 the new output block should start indented 2 or more characters
573 rightward of the correct location.
576 (probably related to bug #70)
577 As reported by Carl Witty on submit@bugs.debian.org 1999-05-08,
579 (in-package "CL-USER")
580 (defun equal-terms (termx termy)
582 ((alpha-equal-bound-term-lists (listx listy)
583 (or (and (null listx) (null listy))
585 (let ((bindings-x (bindings-of-bound-term (car listx)))
586 (bindings-y (bindings-of-bound-term (car listy))))
587 (if (and (null bindings-x) (null bindings-y))
588 (alpha-equal-terms (term-of-bound-term (car listx))
589 (term-of-bound-term (car listy)))
590 (and (= (length bindings-x) (length bindings-y))
592 (enter-binding-pairs (bindings-of-bound-term (car listx))
593 (bindings-of-bound-term (car listy)))
594 (alpha-equal-terms (term-of-bound-term (car listx))
595 (term-of-bound-term (car listy)))
596 (exit-binding-pairs (bindings-of-bound-term (car listx))
597 (bindings-of-bound-term (car listy)))))))
598 (alpha-equal-bound-term-lists (cdr listx) (cdr listy)))))
600 (alpha-equal-terms (termx termy)
601 (if (and (variable-p termx)
603 (equal-bindings (id-of-variable-term termx)
604 (id-of-variable-term termy))
605 (and (equal-operators-p (operator-of-term termx) (operator-of-term termy))
606 (alpha-equal-bound-term-lists (bound-terms-of-term termx)
607 (bound-terms-of-term termy))))))
611 (with-variable-invocation (alpha-equal-terms termx termy))))))
612 causes an assertion failure
613 The assertion (EQ (C::LAMBDA-TAIL-SET C::CALLER)
614 (C::LAMBDA-TAIL-SET (C::LAMBDA-HOME C::CALLEE))) failed.
616 Bob Rogers reports (1999-07-28 on cmucl-imp@cons.org) a smaller test
617 case with the same problem:
618 (defun parse-fssp-alignment ()
619 ;; Given an FSSP alignment file named by the argument . . .
620 (labels ((get-fssp-char ()
624 ;; Stub body, enough to tickle the bug.
625 (list (read-fssp-char)
629 ANSI specifies that the RESULT-TYPE argument of CONCATENATE must be
630 a subtype of SEQUENCE, but CONCATENATE doesn't check this properly:
631 (CONCATENATE 'SIMPLE-ARRAY #(1 2) '(3)) => #(1 2 3)
632 This also leads to funny behavior when derived type specifiers
633 are used, as originally reported by Milan Zamazal for CMU CL (on the
634 Debian bugs mailing list (?) 2000-02-27), then reported by Martin
635 Atzmueller for SBCL (2000-10-01 on sbcl-devel@lists.sourceforge.net):
636 (DEFTYPE FOO () 'SIMPLE-ARRAY)
637 (CONCATENATE 'FOO #(1 2) '(3))
638 => #<ARRAY-TYPE SIMPLE-ARRAY> is a bad type specifier for
640 The derived type specifier FOO should act the same way as the
641 built-in type SIMPLE-ARRAY here, but it doesn't. That problem
642 doesn't seem to exist for sequence types:
643 (DEFTYPE BAR () 'SIMPLE-VECTOR)
644 (CONCATENATE 'BAR #(1 2) '(3)) => #(1 2 3)
647 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
648 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
649 crashes SBCL. In general tracing anything which is used in the
650 implementation of TRACE is likely to have the same problem.
653 As reported by Daniel Solaz on cmucl-help@cons.org 2000-11-23,
654 SXHASH returns the same value for all non-STRUCTURE-OBJECT instances,
655 notably including all PCL instances. There's a limit to how much
656 SXHASH can do to return unique values for instances, but at least
657 it should probably look at the class name, the way that it does
658 for STRUCTURE-OBJECTs.
661 As reported by Martin Atzmueller on the sbcl-devel list 2000-11-22,
662 > There remains one issue, that is a bug in SBCL:
663 > According to my interpretation of the spec, the ":" and "@" modifiers
664 > should appear _after_ the comma-seperated arguments.
665 > Well, SBCL (and CMUCL for that matter) accept
666 > (ASSERT (STRING= (FORMAT NIL "~:8D" 1) " 1"))
667 > where the correct way (IMHO) should be
668 > (ASSERT (STRING= (FORMAT NIL "~8:D" 1) " 1"))
669 Probably SBCL should stop accepting the "~:8D"-style format arguments,
670 or at least issue a warning.
673 (probably related to bug #65)
674 The compiler doesn't like &OPTIONAL arguments in LABELS and FLET
676 (DEFUN FIND-BEFORE (ITEM SEQUENCE &KEY (TEST #'EQL))
677 (LABELS ((FIND-ITEM (OBJ SEQ TEST &OPTIONAL (VAL NIL))
678 (LET ((ITEM (FIRST SEQ)))
681 ((FUNCALL TEST OBJ ITEM)
684 (FIND-ITEM OBJ (REST SEQ) TEST (NCONC VAL `(,ITEM))))))))
685 (FIND-ITEM ITEM SEQUENCE TEST)))
686 from David Young's bug report on cmucl-help@cons.org 30 Nov 2000
687 causes sbcl-0.6.9 to fail with
688 error in function SB-KERNEL:ASSERT-ERROR:
689 The assertion (EQ (SB-C::LAMBDA-TAIL-SET SB-C::CALLER)
690 (SB-C::LAMBDA-TAIL-SET
691 (SB-C::LAMBDA-HOME SB-C::CALLEE))) failed.
694 (DECLAIM (OPTIMIZE ..)) doesn't work. E.g. even after
695 (DECLAIM (OPTIMIZE (SPEED 3))), things are still optimized with
696 the previous SPEED policy. This bug will probably get fixed in
697 0.6.9.x in a general cleanup of optimization policy.
700 (DECLAIM (OPTIMIZE ..)) doesn't work properly inside LOCALLY forms.
703 As noted in the ANSI specification for COERCE, (COERCE 3 'COMPLEX)
704 gives a result which isn't COMPLEX. The result type optimizer
705 for COERCE doesn't know this, perhaps because it was written before
706 ANSI threw this curveball: the optimizer thinks that COERCE always
707 returns a result of the specified type. Thus while the interpreted
709 (DEFUN TRICKY (X) (TYPEP (COERCE X 'COMPLEX) 'COMPLEX))
710 returns the correct result,
712 the compiled function
718 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
719 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
720 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
721 WITH-OUTPUT-TO-STRING.
724 ANSI says in one place that type declarations can be abbreviated even
725 when the type name is not a symbol, e.g.
726 (DECLAIM ((VECTOR T) *FOOVECTOR*))
727 SBCL doesn't support this. But ANSI says in another place that this
728 isn't allowed. So it's not clear this is a bug after all. (See the
729 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
733 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
734 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
735 an easily guessable temporary filename in a way which might open
736 applications using LOAD-FOREIGN to hijacking by malicious users
737 on the same machine. Incantations for doing this safely are
738 floating around the net in various "how to write secure programs
739 despite Unix" documents, and it would be good to (1) fix this in
740 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
741 files and make it share the same new safe logic.
744 (fixed early Feb 2001 by MNA)
747 As reported by wbuss@TELDA.NET (Wolfhard Buss) on cmucl-help
750 (loop with (a . b) of-type float = '(0.0 . 1.0)
751 and (c . d) of-type float = '(2.0 . 3.0)
752 return (list a b c d))
753 should evaluate to (0.0 1.0 2.0 3.0). cmucl-18c disagrees and
754 invokes the debugger: "B is not of type list".
755 SBCL does the same thing.
758 Functions are assigned names based on the context in which they're
759 defined. This is less than ideal for the functions which are
760 used to implement CLOS methods. E.g. the output of
761 (DESCRIBE 'PRINT-OBJECT) lists functions like
762 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
764 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
765 It would be better if these functions' names always identified
766 them as methods, and identified their generic functions and
770 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
771 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
772 the mean of the distribution can be systematically O(0.1%) wrong.
773 Just increasing R-I-E-B is probably not a good solution, since
774 it would decrease efficiency more than is probably necessary. Perhaps
775 using some sort of accept/reject method would be better.
778 (SUBTYPEP '(SATISFIES SOME-UNDEFINED-FUN) NIL)=>NIL,T (should be NIL,NIL)
781 Internally the compiler sometimes evaluates
782 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
783 (I stumbled across this when I added an
784 (assert (not (eq type1 *wild-type*)))
785 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
786 in a type context should probably be translated to T, and so it's
787 probably to ask whether it's equal to the T type and then (using the
788 EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
789 (I haven't tried to investigate this bug enough to guess whether
790 there might be any user-level symptoms.)
793 a latent cross-compilation/bootstrapping bug: The cross-compilation
794 host's CL:CHAR-CODE-LIMIT is used in target code in readtable.lisp
795 and possibly elsewhere. Instead, we should use the target system's
796 CHAR-CODE-LIMIT. This will probably cause problems if we try to
797 bootstrap on a system which uses a different value of CHAR-CODE-LIMIT
801 (subtypep '(or (integer -1 1)
805 (integer -1 1))) => NIL,T
806 An analogous problem with SINGLE-FLOAT and REAL types was fixed in
807 sbcl-0.6.11.22, but some peculiarites of the RATIO type make it
808 awkward to generalize the fix to INTEGER and RATIONAL. It's not
809 clear what's the best fix. (See the "bug in type handling" discussion
810 on cmucl-imp ca. 2001-03-22 and ca. 2001-02-12.)
813 In sbcl-0.6.11.26, (COMPILE 'IN-HOST-COMPILATION-MODE) in
814 src/cold/shared.lisp doesn't correctly translate the
816 (defun in-host-compilation-mode (fn)
817 (let ((*features* (cons :sb-xc-host *features*))
818 ;; the CROSS-FLOAT-INFINITY-KLUDGE, as documented in
819 ;; base-target-features.lisp-expr:
820 (*shebang-features* (set-difference *shebang-features*
821 '(:sb-propagate-float-type
822 :sb-propagate-fun-type))))
823 (with-additional-nickname ("SB-XC" "SB!XC")
825 No error is reported by the compiler, but when the function is executed,
827 TYPE-ERROR in SB-KERNEL::OBJECT-NOT-TYPE-ERROR-HANDLER:
828 (:LINUX :X86 :IEEE-FLOATING-POINT :SB-CONSTRAIN-FLOAT-TYPE :SB-TEST
829 :SB-INTERPRETER :SB-DOC :UNIX ...) is not of type SYMBOL.
832 Inconsistencies between derived and declared VALUES return types for
833 DEFUN aren't checked very well. E.g. the logic which successfully
834 catches problems like
835 (declaim (ftype (function (fixnum) float) foo))
837 (declare (type integer x))
838 (values x)) ; wrong return type, detected, gives warning, good!
840 (declaim (ftype (function (t) (values t t)) bar))
842 (values x)) ; wrong number of return values, no warning, bad!
843 The cause of this is seems to be that (1) the internal function
844 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
845 arguments symmetrically, and (2) when the type checking code was
846 written back when when SBCL's code was still CMU CL, the intent
848 (declaim (ftype (function (t) t) bar))
850 (values x x)) ; wrong number of return values; should give warning?
851 not be warned for, because a two-valued return value is considered
852 to be compatible with callers who expects a single value to be
853 returned. That intent is probably not appropriate for modern ANSI
854 Common Lisp, but fixing this might be complicated because of other
855 divergences between auld-style and new-style handling of
856 multiple-VALUES types. (Some issues related to this were discussed
857 on cmucl-imp at some length sometime in 2000.)
860 The facility for dumping a running Lisp image to disk gets confused
861 when run without the PURIFY option, and creates an unnecessarily large
862 core file (apparently representing memory usage up to the previous
863 high-water mark). Moreover, when the file is loaded, it confuses the
864 GC, so that thereafter memory usage can never be reduced below that
868 The TRACE facility can't be used on some kinds of functions.
869 (Basically, the breakpoint facility was incompletely implemented
870 in the X86 port of CMU CL, and hasn't been fixed in SBCL.)
873 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
874 CL), out-of-line structure slot setters are horribly inefficient
875 whenever the type of the slot is declared, because out-of-line
876 structure slot setters are implemented as closures to save space,
877 so the compiler doesn't compile the type test into code, but
878 instead just saves the type in a lexical closure and interprets it
880 A proper solution involves deciding whether it's really worth
881 saving space by implementing structure slot accessors as closures.
882 (If it's not worth it, the problem vanishes automatically. If it
883 is worth it, there are hacks we could use to force type tests to
884 be compiled anyway, and even shared. E.g. we could implement
885 an EQUAL hash table mapping from types to compiled type tests,
886 and save the appropriate compiled type test as part of each lexical
887 closure; or we could make the lexical closures be placeholders
888 which overwrite their old definition as a lexical closure with
889 a new compiled definition the first time that they're called.)
890 As a workaround for the problem, #'(SETF FOO) expressions can
891 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
892 (defmacro efficient-setf-function (place-function-name)
893 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
894 ;; a workaround for the problem, encouraging the
895 ;; inline expansion of the structure accessor, so
896 ;; that the compiler can optimize its type test
897 (let ((new-value (gensym "NEW-VALUE-"))
898 (structure-value (gensym "STRUCTURE-VALUE-")))
899 `(lambda (,new-value ,structure-value)
900 (setf (,place-function-name ,structure-value)
902 ;; no problem, can just use the ordinary expansion
903 `(function (setf ,place-function-name))))
906 DESCRIBE interacts poorly with *PRINT-CIRCLE*, e.g. the output from
907 (let ((*print-circle* t)) (describe (make-hash-table)))
909 #<HASH-TABLE :TEST EQL :COUNT 0 {90BBFC5}> is an . (EQL)
911 Its REHASH-SIZE is 1.5. Its REHASH-THRESHOLD is . (1.0)
912 It holds 0 key/value pairs.
913 where the ". (EQL)" and ". (1.0)" substrings are screwups.
914 (This is likely a pretty-printer problem which happens to
915 be exercised by DESCRIBE, not actually a DESCRIBE problem.)
918 There's apparently a bug in CEILING optimization which caused
919 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
920 applied the patches to SBCL and they didn't seem to cause problems
921 (as reported sbcl-devel 2001-05-04). However, since the patches
922 modify nontrivial code which was apparently written incorrectly
923 the first time around, until regression tests are written I'm not
924 comfortable merging the patches in the CVS version of SBCL.
927 The error message for calls to structure accessors with the
928 wrong number of arguments is confusing and of the wrong
929 condition class (TYPE-ERROR instead of PROGRAM-ERROR):
930 * (defstruct foo x y)
932 debugger invoked on condition of type SIMPLE-TYPE-ERROR:
933 Structure for accessor FOO-X is not a FOO:
937 As reported by Arthur Lemmens sbcl-devel 2001-05-05, ANSI
938 requires that SYMBOL-MACROLET refuse to rebind special variables,
939 but SBCL doesn't do this. (Also as reported by AL in the same
940 message, SBCL depended on this nonconforming behavior to build
941 itself, because of the way that **CURRENT-SEGMENT** was implemented.
942 As of sbcl-0.6.12.x, this dependence on the nonconforming behavior
943 has been fixed, but the nonconforming behavior remains.)
946 As reported by Arthur Lemmens sbcl-devel 2001-05-05, ANSI's
947 definition of (LOOP .. DO ..) requires that the terms following
948 DO all be compound forms. SBCL's implementation of LOOP allows
949 non-compound forms (like the bare symbol COUNT, in his example)
953 (DESCRIBE 'SB-ALIEN:DEF-ALIEN-TYPE) reports the macro argument list
957 in #<PACKAGE "SB-ALIEN">.
958 Macro-function: #<FUNCTION "DEF!MACRO DEF-ALIEN-TYPE" {19F4A39}>
959 Macro arguments: (#:whole-470 #:environment-471)
960 On Sat, May 26, 2001 09:45:57 AM CDT it was compiled from:
961 /usr/stuff/sbcl/src/code/host-alieneval.lisp
962 Created: Monday, March 12, 2001 07:47:43 AM CST
965 (DESCRIBE 'STREAM-READ-BYTE)
968 (reported by Eric Marsden on cmucl-imp 2001-06-15)
970 (TYPEP 0 '(COMPLEX (EQL 0)))
971 signals an error in sbcl-0.6.12.34,
972 The component type for COMPLEX is not numeric: (EQL 0)
973 This is funny since sbcl-0.6.12.34 knows
974 (SUBTYPEP '(EQL 0) 'NUMBER) => T
977 (reported as a CMU CL bug by Erik Naggum on comp.lang.lisp
979 * (write #*101 :radix t :base 36)
985 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
986 a clean, just-started SBCL system. And it seems to be right:
987 (ROOM T) can bring a small computer to its knees for a *long*
988 time trying to GC afterwards. Surely there's some more economical
989 way to implement (ROOM T).
991 KNOWN BUGS RELATED TO THE IR1 INTERPRETER
993 (Note: At some point, the pure interpreter (actually a semi-pure
994 interpreter aka "the IR1 interpreter") will probably go away, replaced
996 (DEFUN EVAL (X) (FUNCALL (COMPILE NIL (LAMBDA ..)))))
997 and at that time these bugs should either go away automatically or
998 become more tractable to fix. Until then, they'll probably remain,
999 since some of them aren't considered urgent, and the rest are too hard
1000 to fix as long as so many special cases remain. After the IR1
1001 interpreter goes away is also the preferred time to start
1002 systematically exterminating cases where debugging functionality
1003 (backtrace, breakpoint, etc.) breaks down, since getting rid of the
1004 IR1 interpreter will reduce the number of special cases we need to
1008 The FUNCTION special operator doesn't check properly whether its
1009 argument is a function name. E.g. (FUNCTION (X Y)) returns a value
1010 instead of failing with an error. (Later attempting to funcall the
1011 value does cause an error.)
1014 COMPILED-FUNCTION-P bogusly reports T for interpreted functions:
1015 * (DEFUN FOO (X) (- 12 X))
1017 * (COMPILED-FUNCTION-P #'FOO)
1022 (DEFVAR *SUPPRESS-P* T)
1023 (EVAL '(UNLESS *SUPPRESS-P*
1024 (EVAL-WHEN (:COMPILE-TOPLEVEL :LOAD-TOPLEVEL :EXECUTE)
1025 (FORMAT T "surprise!"))))
1026 prints "surprise!". Probably the entire EVAL-WHEN mechanism ought to be
1027 rewritten from scratch to conform to the ANSI definition, abandoning
1028 the *ALREADY-EVALED-THIS* hack which is used in sbcl-0.6.8.9 (and
1029 in the original CMU CL source, too). This should be easier to do --
1030 though still nontrivial -- once the various IR1 interpreter special
1034 EVAL-WHEN's idea of what's a toplevel form is even more screwed up
1035 than the example in IR1-3 would suggest, since COMPILE-FILE and
1036 COMPILE both print both "right now!" messages when compiling the
1040 (EVAL-WHEN (:COMPILE-TOPLEVEL :LOAD-TOPLEVEL :EXECUTE)
1041 (PRINT "yes! right now!"))
1044 (EVAL-WHEN (:COMPILE-TOPLEVEL :LOAD-TOPLEVEL :EXECUTE)
1045 (PRINT "no! right now!"))
1047 and while EVAL doesn't print the "right now!" messages, the first
1048 FUNCALL on the value returned by EVAL causes both of them to be printed.
1051 The system accepts DECLAIM in most places where DECLARE would be
1052 accepted, without even issuing a warning. ANSI allows this, but since
1053 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
1054 meaning is rather different, and it's unlikely that the user
1055 has a good reason for doing DECLAIM not at top level, it would be
1056 good to issue a STYLE-WARNING when this happens. A possible
1057 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
1058 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
1059 [This is considered an IR1-interpreter-related bug because until
1060 EVAL-WHEN is rewritten, which won't happen until after the IR1
1061 interpreter is gone, the system's notion of what's a top-level form
1062 and what's not will remain too confused to fix this problem.]