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 from DTC on the CMU CL mailing list 25 Feb 2000:
131 ;;; Compiler fails when this file is compiled.
133 ;;; Problem shows up in delete-block within ir1util.lisp. The assertion
134 ;;; (assert (member (functional-kind lambda) '(:let :mv-let :assignment)))
135 ;;; fails within bind node branch.
137 ;;; Note that if c::*check-consistency* is enabled then an un-reached
138 ;;; entry is also reported.
141 (declare (values nil))
158 (let ((ttt #'(lambda () (go cccc))))
159 (declare (special ttt))
160 (return-from bbbb nil))
163 (return-from bbbb nil))))))
166 (I *think* this is a bug. It certainly seems like strange behavior. But
167 the ANSI spec is scary, dark, and deep..)
168 (FORMAT NIL "~,1G" 1.4) => "1. "
169 (FORMAT NIL "~3,1G" 1.4) => "1. "
172 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
174 (setf (find-class 'ccc1) (find-class 'ccc))
175 (defmethod zut ((c ccc1)) 123)
176 DTC's recommended workaround from the mailing list 3 Mar 2000:
177 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
180 The ANSI spec, in section "22.3.5.2 Tilde Less-Than-Sign: Logical Block",
181 says that an error is signalled if ~W, ~_, ~<...~:>, ~I, or ~:T is used
182 inside "~<..~>" (without the colon modifier on the closing syntax).
183 However, SBCL doesn't do this:
184 * (FORMAT T "~<munge~wegnum~>" 12)
189 Sometimes (SB-EXT:QUIT) fails with
190 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
191 Process inferior-lisp exited abnormally with code 1
192 I haven't noticed a repeatable case of this yet.
195 some sort of bug in inlining and RETURN-FROM in sbcl-0.6.5: Compiling
198 (BLOCK USED-BY-SOME-Y?
201 (UNLESS (REJECTED? Y)
202 (RETURN-FROM USED-BY-SOME-Y? T)))))
203 (DECLARE (INLINE FROB))
208 error in function SB-KERNEL:ASSERT-ERROR:
209 The assertion (EQ (SB-C::CONTINUATION-KIND SB-C::CONT) :BLOCK-START) failed.
210 This is still present in sbcl-0.6.8.
213 In some cases the compiler believes type declarations on array
214 elements without checking them, e.g.
215 (DECLAIM (OPTIMIZE (SAFETY 3) (SPEED 1) (SPACE 1)))
218 (DECLARE (TYPE (SIMPLE-ARRAY CONS 1) X))
219 (WHEN (CONSP (AREF X 0))
221 (BAR (VECTOR (MAKE-FOO :A 11 :B 12)))
224 in SBCL 0.6.5 (and also in CMU CL 18b). This does not happen for
225 all cases, e.g. the type assumption *is* checked if the array
226 elements are declared to be of some structure type instead of CONS.
229 The printer doesn't report closures very well. This is true in
233 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
234 It would be nice to make closures have a settable name slot,
235 and make things like DEFSTRUCT and FLET, which create closures,
236 set helpful values into this slot.
239 And as long as we're wishing, it would be awfully nice if INSPECT could
240 also report on closures, telling about the values of the bound variables.
243 The compiler assumes that any time a function of declared FTYPE
244 doesn't signal an error, its arguments were of the declared type.
245 E.g. compiling and loading
246 (DECLAIM (OPTIMIZE (SAFETY 3)))
247 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
249 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
251 (COND ((> (FACTORIAL X) 1.0E6)
252 (FORMAT T "too big~%"))
254 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
256 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
259 will cause the INTEGERP case to be selected, giving bogus output a la
261 This violates the "declarations are assertions" principle.
262 According to the ANSI spec, in the section "System Class FUNCTION",
263 this is a case of "lying to the compiler", but the lying is done
264 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
265 not by the unexpectedly general definition of FACTORIAL. In any case,
266 "declarations are assertions" means that lying to the compiler should
267 cause an error to be signalled, and should not cause a bogus
268 result to be returned. Thus, the compiler should not assume
269 that arbitrary functions check their argument types. (It might
270 make sense to add another flag (CHECKED?) to DEFKNOWN to
271 identify functions which *do* check their argument types.)
272 (Also, verify that the compiler handles declared function
273 return types as assertions.)
276 DEFMETHOD doesn't check the syntax of &REST argument lists properly,
277 accepting &REST even when it's not followed by an argument name:
278 (DEFMETHOD FOO ((X T) &REST) NIL)
281 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
282 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
284 (DECLARE (TYPE INDEXOID X))
285 (THE (VALUES INDEXOID)
287 where the implementation of the type check in function FOO
288 includes a full call to %TYPEP. There are also some fundamental problems
289 with the interpretation of VALUES types (inherited from CMU CL, and
290 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
291 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
294 The definitions of SIGCONTEXT-FLOAT-REGISTER and
295 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
296 supported on FreeBSD because the floating point state is not saved,
297 but at least as of FreeBSD 4.0, the floating point state *is* saved,
298 so they could be supported after all. Very likely
299 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
302 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
303 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
304 issue with 'union, 'and, 'or etc. So even though according to the
305 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
306 (and now SBCL) interpret them as legal types.
309 ANSI specifies DEFINE-SYMBOL-MACRO, but it's not defined in SBCL.
310 CMU CL added it ca. Aug 13, 2000, after some discussion on the mailing
311 list, and it is probably possible to use substantially the same
312 patches to add it to SBCL.
315 a slew of floating-point-related errors reported by Peter Van Eynde
317 a: (fixed in sbcl-0.6.11.25)
318 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
319 should probably be 1.4012985e-45. In SBCL,
320 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
321 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
322 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
323 and LEAST-NEGATIVE-LONG-FLOAT.
324 c: Many expressions generate floating infinity:
329 PVE's regression tests want them to raise errors. SBCL
330 generates the infinities instead, which may or may not be
332 d: (in section12.erg) various forms a la
333 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
334 don't give the right behavior.
337 type safety errors reported by Peter Van Eynde July 25, 2000:
338 a: (COERCE (QUOTE (A B C)) (QUOTE (VECTOR * 4)))
340 In general lengths of array type specifications aren't
341 checked by COERCE, so it fails when the spec is
342 (VECTOR 4), (STRING 2), (SIMPLE-BIT-VECTOR 3), or whatever.
343 b: CONCATENATE has the same problem of not checking the length
344 of specified output array types. MAKE-SEQUENCE and MAP and
345 MERGE also have the same problem.
346 c: (COERCE 'AND 'FUNCTION) returns something related to
347 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
348 f: (FLOAT-RADIX 2/3) should signal an error instead of
350 g: (LOAD "*.lsp") should signal FILE-ERROR.
351 h: (MAKE-CONCATENATED-STREAM (MAKE-STRING-OUTPUT-STREAM))
352 should signal TYPE-ERROR.
353 i: MAKE-TWO-WAY-STREAM doesn't check that its arguments can
354 be used for input and output as needed. It should fail with
355 TYPE-ERROR when handed e.g. the results of
356 MAKE-STRING-INPUT-STREAM or MAKE-STRING-OUTPUT-STREAM in
357 the inappropriate positions, but doesn't.
358 j: (PARSE-NAMESTRING (COERCE (LIST #\f #\o #\o (CODE-CHAR 0) #\4 #\8)
360 should probably signal an error instead of making a pathname with
362 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
363 not a binary input stream, but instead cheerfully reads from
364 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
367 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
368 a: (DEFCLASS FOO () (A B A)) should signal a PROGRAM-ERROR, and
370 b: (DEFCLASS FOO () (A B A) (:DEFAULT-INITARGS X A X B)) should
371 signal a PROGRAM-ERROR, and doesn't.
372 c: (DEFCLASS FOO07 NIL ((A :ALLOCATION :CLASS :ALLOCATION :CLASS))),
373 and other DEFCLASS forms with duplicate specifications in their
374 slots, should signal a PROGRAM-ERROR, and doesn't.
375 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
376 causes a COMPILER-ERROR.
379 SYMBOL-MACROLET bugs reported by Peter Van Eynde July 25, 2000:
380 a: (SYMBOL-MACROLET ((T TRUE)) ..) should probably signal
381 PROGRAM-ERROR, but SBCL accepts it instead.
382 b: SYMBOL-MACROLET should refuse to bind something which is
383 declared as a global variable, signalling PROGRAM-ERROR.
384 c: SYMBOL-MACROLET should signal PROGRAM-ERROR if something
385 it binds is declared SPECIAL inside.
388 type system errors reported by Peter Van Eynde July 25, 2000:
389 c: (SUBTYPEP '(INTEGER (0) (0)) 'NIL) dies with nested errors.
390 d: In general, the system doesn't like '(INTEGER (0) (0)) -- it
391 blows up at the level of SPECIFIER-TYPE with
392 "Lower bound (0) is greater than upper bound (0)." Probably
393 SPECIFIER-TYPE should return NIL instead.
394 e: (TYPEP 0 '(COMPLEX (EQL 0)) fails with
395 "Component type for Complex is not numeric: (EQL 0)."
396 This might be easy to fix; the type system already knows
397 that (SUBTYPEP '(EQL 0) 'NUMBER) is true.
398 g: The type system isn't all that smart about relationships
399 between hairy types, as shown in the type.erg test results,
400 e.g. (SUBTYPEP 'CONS '(NOT ATOM)) => NIL, NIL.
403 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
405 (DEFGENERIC FOO02 (X))
406 (DEFMETHOD FOO02 ((X NUMBER)) T)
407 (LET ((M (FIND-METHOD (FUNCTION FOO02)
409 (LIST (FIND-CLASS (QUOTE NUMBER))))))
410 (REMOVE-METHOD (FUNCTION FOO02) M)
411 (DEFGENERIC FOO03 (X))
412 (ADD-METHOD (FUNCTION FOO03) M)))
413 should give an error, but SBCL allows it.
414 b: READ should probably return READER-ERROR, not the bare
415 arithmetic error, when input a la "1/0" or "1e1000" causes
419 It has been reported (e.g. by Peter Van Eynde) that there are
420 several metaobject protocol "errors". (In order to fix them, we might
421 need to document exactly what metaobject protocol specification
422 we're following -- the current code is just inherited from PCL.)
425 The implementation of #'+ returns its single argument without
426 type checking, e.g. (+ "illegal") => "illegal".
429 Attempting to use COMPILE on something defined by DEFMACRO fails:
430 (DEFMACRO FOO (X) (CONS X X))
432 Error in function C::GET-LAMBDA-TO-COMPILE:
433 #<Closure Over Function "DEFUN (SETF MACRO-FUNCTION)" {480E21B1}> was defined in a non-null environment.
436 (SUBTYPEP '(AND ZILCH INTEGER) 'ZILCH) => NIL, NIL
437 Note: I looked into fixing this in 0.6.11.15, but gave up. The
438 problem seems to be that there are two relevant type methods for
439 the subtypep operation, HAIRY :COMPLEX-SUBTYPEP-ARG2 and
440 INTERSECTION :COMPLEX-SUBTYPEP-ARG1, and only the first is
441 called. This could be fixed, but type dispatch is messy and
442 confusing enough already, I don't want to complicate it further.
443 Perhaps someday we can make CLOS cross-compiled (instead of compiled
444 after bootstrapping) so that we don't need to have the type system
445 available before CLOS, and then we can rewrite the type methods to
446 CLOS methods, and then expressing the solutions to stuff like this
447 should become much more straightforward. -- WHN 2001-03-14
450 The debugger LIST-LOCATIONS command doesn't work properly.
453 Compiling and loading
454 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
456 then requesting a BACKTRACE at the debugger prompt gives no information
457 about where in the user program the problem occurred.
460 The compiler is supposed to do type inference well enough that
463 ((SIMPLE-ARRAY SINGLE-FLOAT)
465 (DECLARE (TYPE (SIMPLE-ARRAY SINGLE-FLOAT) X))
468 is redundant. However, as reported by Juan Jose Garcia Ripoll for
469 CMU CL, it sometimes doesn't. Adding declarations is a pretty good
470 workaround for the problem for now, but can't be done by the TYPECASE
471 macros themselves, since it's too hard for the macro to detect
472 assignments to the variable within the clause.
473 Note: The compiler *is* smart enough to do the type inference in
474 many cases. This case, derived from a couple of MACROEXPAND-1
475 calls on Ripoll's original test case,
477 (DECLARE (OPTIMIZE SPEED (SAFETY 0)))
478 (COND ((TYPEP A '(SIMPLE-ARRAY SINGLE-FLOAT)) NIL
479 (LET ((LENGTH (ARRAY-TOTAL-SIZE A)))
480 (LET ((I 0) (G2554 LENGTH))
481 (DECLARE (TYPE REAL G2554) (TYPE REAL I))
484 (WHEN (>= I G2554) (GO SB-LOOP::END-LOOP))
485 (SETF (ROW-MAJOR-AREF A I) (- (ROW-MAJOR-AREF A I)))
486 (GO SB-LOOP::NEXT-LOOP)
487 SB-LOOP::END-LOOP))))))
488 demonstrates the problem; but the problem goes away if the TAGBODY
489 and GO forms are removed (leaving the SETF in ordinary, non-looping
490 code), or if the TAGBODY and GO forms are retained, but the
491 assigned value becomes 0.0 instead of (- (ROW-MAJOR-AREF A I)).
494 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
495 I am looking into this problem that showed up on the cmucl-help
496 list. It seems to me that the "implementation specific environment
497 hacking functions" found in pcl/walker.lisp are completely messed
498 up. The good thing is that they appear to be barely used within
499 PCL and the munged environment object is passed to cmucl only
500 in calls to macroexpand-1, which is probably why this case fails.
501 SBCL uses essentially the same code, so if the environment hacking
502 is screwed up, it affects us too.
505 Using the pretty-printer from the command prompt gives funny
506 results, apparently because the pretty-printer doesn't know
507 about user's command input, including the user's carriage return
508 that the user, and therefore the pretty-printer thinks that
509 the new output block should start indented 2 or more characters
510 rightward of the correct location.
513 (probably related to bug #70; maybe related to bug #109)
514 As reported by Carl Witty on submit@bugs.debian.org 1999-05-08,
516 (in-package "CL-USER")
517 (defun equal-terms (termx termy)
519 ((alpha-equal-bound-term-lists (listx listy)
520 (or (and (null listx) (null listy))
522 (let ((bindings-x (bindings-of-bound-term (car listx)))
523 (bindings-y (bindings-of-bound-term (car listy))))
524 (if (and (null bindings-x) (null bindings-y))
525 (alpha-equal-terms (term-of-bound-term (car listx))
526 (term-of-bound-term (car listy)))
527 (and (= (length bindings-x) (length bindings-y))
529 (enter-binding-pairs (bindings-of-bound-term (car listx))
530 (bindings-of-bound-term (car listy)))
531 (alpha-equal-terms (term-of-bound-term (car listx))
532 (term-of-bound-term (car listy)))
533 (exit-binding-pairs (bindings-of-bound-term (car listx))
534 (bindings-of-bound-term (car listy)))))))
535 (alpha-equal-bound-term-lists (cdr listx) (cdr listy)))))
537 (alpha-equal-terms (termx termy)
538 (if (and (variable-p termx)
540 (equal-bindings (id-of-variable-term termx)
541 (id-of-variable-term termy))
542 (and (equal-operators-p (operator-of-term termx) (operator-of-term termy))
543 (alpha-equal-bound-term-lists (bound-terms-of-term termx)
544 (bound-terms-of-term termy))))))
548 (with-variable-invocation (alpha-equal-terms termx termy))))))
549 causes an assertion failure
550 The assertion (EQ (C::LAMBDA-TAIL-SET C::CALLER)
551 (C::LAMBDA-TAIL-SET (C::LAMBDA-HOME C::CALLEE))) failed.
553 Bob Rogers reports (1999-07-28 on cmucl-imp@cons.org) a smaller test
554 case with the same problem:
555 (defun parse-fssp-alignment ()
556 ;; Given an FSSP alignment file named by the argument . . .
557 (labels ((get-fssp-char ()
561 ;; Stub body, enough to tickle the bug.
562 (list (read-fssp-char)
566 ANSI specifies that the RESULT-TYPE argument of CONCATENATE must be
567 a subtype of SEQUENCE, but CONCATENATE doesn't check this properly:
568 (CONCATENATE 'SIMPLE-ARRAY #(1 2) '(3)) => #(1 2 3)
569 This also leads to funny behavior when derived type specifiers
570 are used, as originally reported by Milan Zamazal for CMU CL (on the
571 Debian bugs mailing list (?) 2000-02-27), then reported by Martin
572 Atzmueller for SBCL (2000-10-01 on sbcl-devel@lists.sourceforge.net):
573 (DEFTYPE FOO () 'SIMPLE-ARRAY)
574 (CONCATENATE 'FOO #(1 2) '(3))
575 => #<ARRAY-TYPE SIMPLE-ARRAY> is a bad type specifier for
577 The derived type specifier FOO should act the same way as the
578 built-in type SIMPLE-ARRAY here, but it doesn't. That problem
579 doesn't seem to exist for sequence types:
580 (DEFTYPE BAR () 'SIMPLE-VECTOR)
581 (CONCATENATE 'BAR #(1 2) '(3)) => #(1 2 3)
584 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
585 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
586 crashes SBCL. In general tracing anything which is used in the
587 implementation of TRACE is likely to have the same problem.
590 As reported by Daniel Solaz on cmucl-help@cons.org 2000-11-23,
591 SXHASH returns the same value for all non-STRUCTURE-OBJECT instances,
592 notably including all PCL instances. There's a limit to how much
593 SXHASH can do to return unique values for instances, but at least
594 it should probably look at the class name, the way that it does
595 for STRUCTURE-OBJECTs.
598 As reported by Martin Atzmueller on the sbcl-devel list 2000-11-22,
599 > There remains one issue, that is a bug in SBCL:
600 > According to my interpretation of the spec, the ":" and "@" modifiers
601 > should appear _after_ the comma-seperated arguments.
602 > Well, SBCL (and CMUCL for that matter) accept
603 > (ASSERT (STRING= (FORMAT NIL "~:8D" 1) " 1"))
604 > where the correct way (IMHO) should be
605 > (ASSERT (STRING= (FORMAT NIL "~8:D" 1) " 1"))
606 Probably SBCL should stop accepting the "~:8D"-style format arguments,
607 or at least issue a warning.
610 (probably related to bug #65; maybe related to bug #109)
611 The compiler doesn't like &OPTIONAL arguments in LABELS and FLET
613 (DEFUN FIND-BEFORE (ITEM SEQUENCE &KEY (TEST #'EQL))
614 (LABELS ((FIND-ITEM (OBJ SEQ TEST &OPTIONAL (VAL NIL))
615 (LET ((ITEM (FIRST SEQ)))
618 ((FUNCALL TEST OBJ ITEM)
621 (FIND-ITEM OBJ (REST SEQ) TEST (NCONC VAL `(,ITEM))))))))
622 (FIND-ITEM ITEM SEQUENCE TEST)))
623 from David Young's bug report on cmucl-help@cons.org 30 Nov 2000
624 causes sbcl-0.6.9 to fail with
625 error in function SB-KERNEL:ASSERT-ERROR:
626 The assertion (EQ (SB-C::LAMBDA-TAIL-SET SB-C::CALLER)
627 (SB-C::LAMBDA-TAIL-SET
628 (SB-C::LAMBDA-HOME SB-C::CALLEE))) failed.
631 (DECLAIM (OPTIMIZE ..)) doesn't work. E.g. even after
632 (DECLAIM (OPTIMIZE (SPEED 3))), things are still optimized with
633 the previous SPEED policy. This bug will probably get fixed in
634 0.6.9.x in a general cleanup of optimization policy.
637 (DECLAIM (OPTIMIZE ..)) doesn't work properly inside LOCALLY forms.
640 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
641 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
642 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
643 WITH-OUTPUT-TO-STRING.
646 ANSI says in one place that type declarations can be abbreviated even
647 when the type name is not a symbol, e.g.
648 (DECLAIM ((VECTOR T) *FOOVECTOR*))
649 SBCL doesn't support this. But ANSI says in another place that this
650 isn't allowed. So it's not clear this is a bug after all. (See the
651 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
655 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
656 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
657 an easily guessable temporary filename in a way which might open
658 applications using LOAD-FOREIGN to hijacking by malicious users
659 on the same machine. Incantations for doing this safely are
660 floating around the net in various "how to write secure programs
661 despite Unix" documents, and it would be good to (1) fix this in
662 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
663 files and make it share the same new safe logic.
666 Functions are assigned names based on the context in which they're
667 defined. This is less than ideal for the functions which are
668 used to implement CLOS methods. E.g. the output of
669 (DESCRIBE 'PRINT-OBJECT) lists functions like
670 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
672 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
673 It would be better if these functions' names always identified
674 them as methods, and identified their generic functions and
678 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
679 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
680 the mean of the distribution can be systematically O(0.1%) wrong.
681 Just increasing R-I-E-B is probably not a good solution, since
682 it would decrease efficiency more than is probably necessary. Perhaps
683 using some sort of accept/reject method would be better.
686 (SUBTYPEP '(SATISFIES SOME-UNDEFINED-FUN) NIL)=>NIL,T (should be NIL,NIL)
689 Internally the compiler sometimes evaluates
690 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
691 (I stumbled across this when I added an
692 (assert (not (eq type1 *wild-type*)))
693 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
694 in a type context should probably be translated to T, and so it's
695 probably to ask whether it's equal to the T type and then (using the
696 EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
697 (I haven't tried to investigate this bug enough to guess whether
698 there might be any user-level symptoms.)
701 a latent cross-compilation/bootstrapping bug: The cross-compilation
702 host's CL:CHAR-CODE-LIMIT is used in target code in readtable.lisp
703 and possibly elsewhere. Instead, we should use the target system's
704 CHAR-CODE-LIMIT. This will probably cause problems if we try to
705 bootstrap on a system which uses a different value of CHAR-CODE-LIMIT
709 (subtypep '(or (integer -1 1)
713 (integer -1 1))) => NIL,T
714 An analogous problem with SINGLE-FLOAT and REAL types was fixed in
715 sbcl-0.6.11.22, but some peculiarites of the RATIO type make it
716 awkward to generalize the fix to INTEGER and RATIONAL. It's not
717 clear what's the best fix. (See the "bug in type handling" discussion
718 on cmucl-imp ca. 2001-03-22 and ca. 2001-02-12.)
721 In sbcl-0.6.11.26, (COMPILE 'IN-HOST-COMPILATION-MODE) in
722 src/cold/shared.lisp doesn't correctly translate the
724 (defun in-host-compilation-mode (fn)
725 (let ((*features* (cons :sb-xc-host *features*))
726 ;; the CROSS-FLOAT-INFINITY-KLUDGE, as documented in
727 ;; base-target-features.lisp-expr:
728 (*shebang-features* (set-difference *shebang-features*
729 '(:sb-propagate-float-type
730 :sb-propagate-fun-type))))
731 (with-additional-nickname ("SB-XC" "SB!XC")
733 No error is reported by the compiler, but when the function is executed,
735 TYPE-ERROR in SB-KERNEL::OBJECT-NOT-TYPE-ERROR-HANDLER:
736 (:LINUX :X86 :IEEE-FLOATING-POINT :SB-CONSTRAIN-FLOAT-TYPE :SB-TEST
737 :SB-INTERPRETER :SB-DOC :UNIX ...) is not of type SYMBOL.
740 Inconsistencies between derived and declared VALUES return types for
741 DEFUN aren't checked very well. E.g. the logic which successfully
742 catches problems like
743 (declaim (ftype (function (fixnum) float) foo))
745 (declare (type integer x))
746 (values x)) ; wrong return type, detected, gives warning, good!
748 (declaim (ftype (function (t) (values t t)) bar))
750 (values x)) ; wrong number of return values, no warning, bad!
751 The cause of this is seems to be that (1) the internal function
752 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
753 arguments symmetrically, and (2) when the type checking code was
754 written back when when SBCL's code was still CMU CL, the intent
756 (declaim (ftype (function (t) t) bar))
758 (values x x)) ; wrong number of return values; should give warning?
759 not be warned for, because a two-valued return value is considered
760 to be compatible with callers who expects a single value to be
761 returned. That intent is probably not appropriate for modern ANSI
762 Common Lisp, but fixing this might be complicated because of other
763 divergences between auld-style and new-style handling of
764 multiple-VALUES types. (Some issues related to this were discussed
765 on cmucl-imp at some length sometime in 2000.)
768 The facility for dumping a running Lisp image to disk gets confused
769 when run without the PURIFY option, and creates an unnecessarily large
770 core file (apparently representing memory usage up to the previous
771 high-water mark). Moreover, when the file is loaded, it confuses the
772 GC, so that thereafter memory usage can never be reduced below that
776 The TRACE facility can't be used on some kinds of functions.
777 (Basically, the breakpoint facility was incompletely implemented
778 in the X86 port of CMU CL, and hasn't been fixed in SBCL.)
781 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
782 CL), out-of-line structure slot setters are horribly inefficient
783 whenever the type of the slot is declared, because out-of-line
784 structure slot setters are implemented as closures to save space,
785 so the compiler doesn't compile the type test into code, but
786 instead just saves the type in a lexical closure and interprets it
788 A proper solution involves deciding whether it's really worth
789 saving space by implementing structure slot accessors as closures.
790 (If it's not worth it, the problem vanishes automatically. If it
791 is worth it, there are hacks we could use to force type tests to
792 be compiled anyway, and even shared. E.g. we could implement
793 an EQUAL hash table mapping from types to compiled type tests,
794 and save the appropriate compiled type test as part of each lexical
795 closure; or we could make the lexical closures be placeholders
796 which overwrite their old definition as a lexical closure with
797 a new compiled definition the first time that they're called.)
798 As a workaround for the problem, #'(SETF FOO) expressions can
799 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
800 (defmacro efficient-setf-function (place-function-name)
801 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
802 ;; a workaround for the problem, encouraging the
803 ;; inline expansion of the structure accessor, so
804 ;; that the compiler can optimize its type test
805 (let ((new-value (gensym "NEW-VALUE-"))
806 (structure-value (gensym "STRUCTURE-VALUE-")))
807 `(lambda (,new-value ,structure-value)
808 (setf (,place-function-name ,structure-value)
810 ;; no problem, can just use the ordinary expansion
811 `(function (setf ,place-function-name))))
814 DESCRIBE interacts poorly with *PRINT-CIRCLE*, e.g. the output from
815 (let ((*print-circle* t)) (describe (make-hash-table)))
817 #<HASH-TABLE :TEST EQL :COUNT 0 {90BBFC5}> is an . (EQL)
819 Its REHASH-SIZE is 1.5. Its REHASH-THRESHOLD is . (1.0)
820 It holds 0 key/value pairs.
821 where the ". (EQL)" and ". (1.0)" substrings are screwups.
822 (This is likely a pretty-printer problem which happens to
823 be exercised by DESCRIBE, not actually a DESCRIBE problem.)
826 There's apparently a bug in CEILING optimization which caused
827 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
828 applied the patches to SBCL and they didn't seem to cause problems
829 (as reported sbcl-devel 2001-05-04). However, since the patches
830 modify nontrivial code which was apparently written incorrectly
831 the first time around, until regression tests are written I'm not
832 comfortable merging the patches in the CVS version of SBCL.
835 As reported by Arthur Lemmens sbcl-devel 2001-05-05, ANSI
836 requires that SYMBOL-MACROLET refuse to rebind special variables,
837 but SBCL doesn't do this. (Also as reported by AL in the same
838 message, SBCL depended on this nonconforming behavior to build
839 itself, because of the way that **CURRENT-SEGMENT** was implemented.
840 As of sbcl-0.6.12.x, this dependence on the nonconforming behavior
841 has been fixed, but the nonconforming behavior remains.)
844 (DESCRIBE 'SB-ALIEN:DEF-ALIEN-TYPE) reports the macro argument list
848 in #<PACKAGE "SB-ALIEN">.
849 Macro-function: #<FUNCTION "DEF!MACRO DEF-ALIEN-TYPE" {19F4A39}>
850 Macro arguments: (#:whole-470 #:environment-471)
851 On Sat, May 26, 2001 09:45:57 AM CDT it was compiled from:
852 /usr/stuff/sbcl/src/code/host-alieneval.lisp
853 Created: Monday, March 12, 2001 07:47:43 AM CST
856 (DESCRIBE 'STREAM-READ-BYTE)
859 (reported by Eric Marsden on cmucl-imp 2001-06-15)
860 (and APD pointed out on sbcl-devel 2001-12-29 that it's the same
864 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
865 a clean, just-started SBCL system. And it seems to be right:
866 (ROOM T) can bring a small computer to its knees for a *long*
867 time trying to GC afterwards. Surely there's some more economical
868 way to implement (ROOM T).
871 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
873 ;;; This file fails to compile.
874 ;;; Maybe this bug is related to bugs #65, #70 in the BUGS file.
875 (in-package :cl-user)
881 ;; Uncomment and it works
884 In SBCL 0.6.12.42, the problem is
885 internal error, failed AVER:
886 "(COMMON-LISP:EQ (SB!C::LAMBDA-TAIL-SET SB!C::CALLER)
887 (SB!C::LAMBDA-TAIL-SET (SB!C::LAMBDA-HOME SB!C::CALLEE)))"
890 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
892 ;;; The compiler is flushing the argument type test, and the default
893 ;;; case in the cond, so that calling with say a fixnum 0 causes a
895 (declaim (optimize (safety 2) (speed 3)))
897 (declare (type (or string stream) x))
898 (cond ((typep x 'string) 'string)
899 ((typep x 'stream) 'stream)
902 The symptom in sbcl-0.6.12.42 on OpenBSD is actually (TST 0)=>STREAM
903 (not the SIGBUS reported in the comment) but that's broken too;
904 type declarations are supposed to be treated as assertions unless
905 SAFETY 0, so we should be getting a TYPE-ERROR.
908 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
910 (in-package :cl-user)
911 ;;; Produces an assertion failures when compiled.
913 (declare (type (or (function (t) t) null) z))
914 (let ((z (or z #'identity)))
915 (declare (type (function (t) t) z))
917 The error in sbcl-0.6.12.42 is
918 internal error, failed AVER:
919 "(COMMON-LISP:NOT (COMMON-LISP:EQ SB!C::CHECK COMMON-LISP:T))"
922 reported by Martin Atzmueller 2001-06-25; taken from CMU CL bugs
923 collection; apparently originally reported by Bruno Haible
924 (in-package :cl-user)
925 ;;; From: Bruno Haible
926 ;;; Subject: scope of SPECIAL declarations
927 ;;; It seems CMUCL has a bug relating to the scope of SPECIAL
928 ;;; declarations. I observe this with "CMU Common Lisp 18a x86-linux
931 (declare (special x))
934 (declare (special x)) y)))
935 ;;; Gives: 0 (this should return 1 according to CLHS)
937 (declare (special x))
940 (declare (special x)) y)))
941 ;;; Gives: 1 (correct).
942 The reported results match what we get from the interpreter
946 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
948 (in-package :cl-user)
949 ;;; From: David Gadbois <gadbois@cyc.com>
951 ;;; Logical pathnames aren't externalizable.
953 (let ((tempfile "/tmp/test.lisp"))
954 (setf (logical-pathname-translations "XXX")
955 '(("XXX:**;*.*" "/tmp/**/*.*")))
956 (with-open-file (out tempfile :direction :output)
957 (write-string "(defvar *path* #P\"XXX:XXX;FOO.LISP\")" out))
958 (compile-file tempfile))
959 The error message in sbcl-0.6.12.42 is
961 ; (while making load form for #<SB-IMPL::LOGICAL-HOST "XXX">)
962 ; A logical host can't be dumped as a constant: #<SB-IMPL::LOGICAL-HOST "XXX">
965 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
967 (in-package :cl-user)
968 ;;; The following invokes a compiler error.
969 (declaim (optimize (speed 2) (debug 3)))
972 (unwind-protect nil)))
976 The error message in sbcl-0.6.12.42 is
977 internal error, failed AVER:
978 "(COMMON-LISP:EQ (SB!C::TN-ENVIRONMENT SB!C:TN) SB!C::TN-ENV)"
981 When the compiler inline expands functions, it may be that different
982 kinds of return values are generated from different code branches.
983 E.g. an inline expansion of POSITION generates integer results
984 from one branch, and NIL results from another. When that inline
985 expansion is used in a context where only one of those results
988 (aref *a1* (position x *a2*)))
989 and the compiler can't prove that the unacceptable branch is
990 never taken, then bogus type mismatch warnings can be generated.
991 If you need to suppress the type mismatch warnings, you can
992 suppress the inline expansion,
994 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
995 (aref *a1* (position x *a2*)))
996 or, sometimes, suppress them by declaring the result to be of an
999 (aref *a1* (the integer (position x *a2*))))
1001 This is not a new compiler problem in 0.7.0, but the new compiler
1002 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
1003 more conspicuous. If you don't need performance from these functions,
1004 and the bogus warnings are a nuisance for you, you can return to
1005 your pre-0.7.0 state of grace with
1006 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
1009 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
1010 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
1011 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
1012 when of course it should be NIL. (He says it only fails for X86,
1013 not SPARC; dunno about Alpha.)
1015 Also, "the same problem exists for LONG-FLOAT-EPSILON,
1016 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
1017 for the -negative- the + is replaced by a - in the test)."
1019 Raymond Toy comments that this is tricky on the X86 since its FPU
1020 uses 80-bit precision internally.
1023 The compiler incorrectly figures the return type of
1024 (DEFUN FOO (FRAME UP-FRAME)
1031 This problem exists in CMU CL 18c too. When I reported it on
1032 cmucl-imp@cons.org, Raymond Toy replied 23 Aug 2001 with
1033 a partial explanation, but no fix has been found yet.
1036 Even in sbcl-0.pre7.x, which is supposed to be free of the old
1037 non-ANSI behavior of treating the function return type inferred
1038 from the current function definition as a declaration of the
1039 return type from any function of that name, the return type of NIL
1040 is attached to FOO in 120a above, and used to optimize code which
1044 There was some sort of screwup in handling of
1045 (IF (NOT (IGNORE-ERRORS ..))). E.g.
1047 (if (not (ignore-errors
1048 (make-pathname :host "foo" :directory "!bla" :name "bar")))
1050 (error "notunlessnot")))
1051 The (NOT (IGNORE-ERRORS ..)) form evaluates to T, so this should be
1052 printing "ok", but instead it's going to the ERROR. This problem
1053 seems to've been introduced by MNA's HANDLER-CASE patch (sbcl-devel
1054 2001-07-17) and as a workaround (put in sbcl-0.pre7.14.flaky4.12)
1055 I reverted back to the old weird HANDLER-CASE code. However, I
1056 think the problem looks like a compiler bug in handling RETURN-FROM,
1057 so I left the MNA-patched code in HANDLER-CASE (suppressed with
1058 #+NIL) and I'd like to go back to see whether this really is
1059 a compiler bug before I delete this BUGS entry.
1062 The *USE-IMPLEMENTATION-TYPES* hack causes bugs, particularly
1063 (IN-PACKAGE :SB-KERNEL)
1064 (TYPE= (SPECIFIER-TYPE '(VECTOR T))
1065 (SPECIFIER-TYPE '(VECTOR UNDEFTYPE)))
1066 Then because of this, the compiler bogusly optimizes
1067 (TYPEP #(11) '(SIMPLE-ARRAY UNDEF-TYPE 1))
1068 to T. Unfortunately, just setting *USE-IMPLEMENTATION-TYPES* to
1069 NIL around sbcl-0.pre7.14.flaky4.12 didn't work: the compiler complained
1070 about type mismatches (probably harmlessly, another instance of bug 117);
1071 and then cold init died with a segmentation fault.
1074 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
1075 the entire lexical environment at the point of MACROLET available
1076 in the bodies of the macroexpander functions. In particular, it
1077 allows the function bodies (which run at compile time) to try to
1078 access lexical variables (which are only defined at runtime).
1079 It doesn't even issue a warning, which is bad.
1081 The SBCL behavior arguably conforms to the ANSI spec (since the
1082 spec says that the behavior is undefined, ergo anything conforms).
1083 However, it would be better to issue a compile-time error.
1084 Unfortunately I (WHN) don't see any simple way to detect this
1085 condition in order to issue such an error, so for the meantime
1086 SBCL just does this weird broken "conforming" thing.
1088 The ANSI standard says, in the definition of the special operator
1090 The macro-expansion functions defined by MACROLET are defined
1091 in the lexical environment in which the MACROLET form appears.
1092 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
1093 the local macro definitions in a MACROLET, but the consequences
1094 are undefined if the local macro definitions reference any
1095 local variable or function bindings that are visible in that
1096 lexical environment.
1097 Then it seems to contradict itself by giving the example
1099 (macrolet ((fudge (z)
1100 ;The parameters x and flag are not accessible
1101 ; at this point; a reference to flag would be to
1102 ; the global variable of that name.
1103 ` (if flag (* ,z ,z) ,z)))
1104 ;The parameters x and flag are accessible here.
1108 The comment "a reference to flag would be to the global variable
1109 of the same name" sounds like good behavior for the system to have.
1110 but actual specification quoted above says that the actual behavior
1114 (as reported by Gabe Garza on cmucl-help 2001-09-21)
1116 (defun test-pred (x y)
1120 (func (lambda () x)))
1121 (print (eq func func))
1122 (print (test-pred func func))
1123 (delete func (list func))))
1124 Now calling (TEST-CASE) gives output
1127 (#<FUNCTION {500A9EF9}>)
1128 Evidently Python thinks of the lambda as a code transformation so
1129 much that it forgets that it's also an object.
1132 (fixed in 0.pre7.41)
1135 The DEFSTRUCT section of the ANSI spec, in the :CONC-NAME section,
1136 specifies a precedence rule for name collisions between slot accessors of
1137 structure classes related by inheritance. As of 0.7.0, SBCL still
1141 insufficient syntax checking in MACROLET:
1143 (macrolet ((defmacro bar (z) `(+ z z)))
1145 shouldn't compile without error (because of the extra DEFMACRO symbol).
1148 reported by Alexey Dejneka on sbcl-devel 2001-11-03
1150 "Return X if X is a non-negative integer."
1151 (let ((step (lambda (%funcall)
1154 (t (1+ (funcall %funcall (1- n)))))))))
1157 (funcall step (lambda (n)
1158 (funcall (funcall a a) n))))
1160 (funcall step (lambda (n)
1161 (funcall (funcall a a) n)))))
1163 This function returns its argument. But after removing percents it
1164 does not work: "Result of (1- n) is not a function".
1167 As of sbcl-0.pre7.86.flaky7.3, the cross-compiler, and probably
1168 the CL:COMPILE function (which is based on the same %COMPILE
1169 mechanism) get confused by
1171 (labels ((sxhash-number (x)
1173 (fixnum (sxhash x)) ; through DEFTRANSFORM
1174 (integer (sb!bignum:sxhash-bignum x))
1175 (single-float (sxhash x)) ; through DEFTRANSFORM
1176 (double-float (sxhash x)) ; through DEFTRANSFORM
1177 #!+long-float (long-float (error "stub: no LONG-FLOAT"))
1178 (ratio (let ((result 127810327))
1179 (declare (type fixnum result))
1180 (mixf result (sxhash-number (numerator x)))
1181 (mixf result (sxhash-number (denominator x)))
1183 (complex (let ((result 535698211))
1184 (declare (type fixnum result))
1185 (mixf result (sxhash-number (realpart x)))
1186 (mixf result (sxhash-number (imagpart x)))
1188 (sxhash-recurse (x &optional (depthoid +max-hash-depthoid+))
1189 (declare (type index depthoid))
1192 (if (plusp depthoid)
1193 (mix (sxhash-recurse (car x) (1- depthoid))
1194 (sxhash-recurse (cdr x) (1- depthoid)))
1197 (if (typep x 'structure-object)
1199 (sxhash ; through DEFTRANSFORM
1200 (class-name (layout-class (%instance-layout x)))))
1202 (symbol (sxhash x)) ; through DEFTRANSFORM
1203 (number (sxhash-number x))
1206 (simple-string (sxhash x)) ; through DEFTRANSFORM
1207 (string (%sxhash-substring x))
1208 (bit-vector (let ((result 410823708))
1209 (declare (type fixnum result))
1210 (dotimes (i (min depthoid (length x)))
1211 (mixf result (aref x i)))
1213 (t (logxor 191020317 (sxhash (array-rank x))))))
1216 (sxhash (char-code x)))) ; through DEFTRANSFORM
1218 (sxhash-recurse x)))
1219 complaining "function called with two arguments, but wants exactly
1220 one" about SXHASH-RECURSE. (This might not be strictly a new bug,
1221 since IIRC post-fork CMU CL has also had problems with &OPTIONAL
1222 arguments in FLET/LABELS: it might be an old Python bug which is
1223 only exercised by the new arrangement of the SBCL compiler.)
1227 (DEFUN FOO () (CATCH 0 (PRINT 1331)))
1229 #<SB-C:TN '0!1> is not valid as the second argument to VOP:
1230 SB-C:MAKE-CATCH-BLOCK,
1231 since the TN's primitive type SB-VM::POSITIVE-FIXNUM doesn't allow
1232 any of the SCs allowed by the operand restriction:
1233 (SB-VM::DESCRIPTOR-REG)
1234 The (CATCH 0 ...) construct is bad style (because of unportability
1235 of EQ testing of numbers) but it is legal, and shouldn't cause an
1236 internal compiler error. (This error occurs in sbcl-0.6.13 and in
1237 0.pre7.86.flaky7.14.)
1240 Trying to compile something like
1241 (sb!alien:def-alien-routine "breakpoint_remove" sb!c-call:void
1242 (code-obj sb!c-call:unsigned-long)
1243 (pc-offset sb!c-call:int)
1244 (old-inst sb!c-call:unsigned-long))
1245 in SBCL-0.pre7.86.flaky7.22 after warm init fails with an error
1246 cannot use values types here
1247 probably because the SB-C-CALL:VOID type gets translated to (VALUES).
1248 It should be valid to use VOID for a function return type, so perhaps
1249 instead of calling SPECIFIER-TYPE (which excludes all VALUES types
1250 automatically) we should call VALUES-SPECIFIER-TYPE and handle VALUES
1251 types manually, allowing the special case (VALUES) but still excluding
1252 all more-complex VALUES types.
1255 Ideally, uninterning a symbol would allow it, and its associated
1256 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
1257 at least as of sbcl-0.7.0, this isn't the case. Information about
1258 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
1259 essentially in ordinary (non-weak) hash tables keyed by symbols.
1260 Thus, once a system has an entry in this system, it tends to live
1261 forever, even when it is uninterned and all other references to it
1265 (reported by Arnaud Rouanet on cmucl-imp 2001-12-18)
1266 (defmethod foo ((x integer))
1268 (defmethod foo :around ((x integer))
1270 (call-next-method)))
1271 Now (FOO 3) should return 3, but instead it returns 4.
1274 (SB-DEBUG:BACKTRACE) output should start with something
1275 including the name BACKTRACE, not (as in 0.pre7.88)
1276 just "0: (\"hairy arg processor\" ...)". Until about
1277 sbcl-0.pre7.109, the names in BACKTRACE were all screwed
1278 up compared to the nice useful names in sbcl-0.6.13.
1279 Around sbcl-0.pre7.109, they were mostly fixed by using
1280 NAMED-LAMBDA to implement DEFUN. However, there are still
1281 some screwups left, e.g. as of sbcl-0.pre7.109, there are
1282 still some functions named "hairy arg processor" and
1283 "SB-INT:&MORE processor".
1286 a pair of cross-compiler bugs in sbcl-0.pre7.107
1289 $ cat > /tmp/bug138.lisp << EOF
1290 (in-package "SB!KERNEL")
1291 (defun f-c-l (name parent-types)
1292 (let* ((cpl (mapcar (lambda (x)
1293 (condition-class-cpl x))
1296 (concatenate 'simple-vector
1297 (layout-inherits cond-layout))))
1298 (if (not (mismatch (layout-inherits olayout) new-inherits))
1302 $ sbcl --core output/after-xc.core
1304 * (target-compile-file "/tmp/bug138.lisp")
1306 internal error, failed AVER:
1307 "(COMMON-LISP:MEMBER SB!C::FUN (SB!C::COMPONENT-LAMBDAS SB!C:COMPONENT))"
1309 It seems as though this xc bug is likely to correspond to a bug in the
1310 ordinary compiler, but I haven't yet found a test case which causes
1311 this problem in the ordinary compiler.
1313 related weirdness: Using #'(LAMBDA (X) ...) instead of (LAMBDA (X) ...)
1314 makes the assertion failure go away.
1317 Even when you relax the AVER that fails in 138a, there's another
1318 problem cross-compiling the same code:
1319 internal error, failed AVER:
1321 (COMMON-LISP:HASH-TABLE-COUNT
1322 (SB!FASL::FASL-OUTPUT-PATCH-TABLE SB!FASL:FASL-OUTPUT)))"
1324 The same problem appears in the simpler test case
1325 (in-package "SB!KERNEL")
1327 (let ((cpl (foo (lambda (x)
1328 (condition-class-cpl x))))
1329 (new-inherits (layout-inherits cond-layout)))
1330 (layout-inherits olayout)))
1332 Changing CONDITION-CLASS-CPL or (either of the calls to) LAYOUT-INHERITS
1333 to arbitrary nonmagic not-defined-yet just-do-a-full-call functions makes
1334 the problem go away. Also, even in this simpler test case which fails
1335 on a very different AVER, the 138a weirdness about s/(lambda/#'(lambda/
1336 making the problem go away is preserved.
1338 I still haven't found any way to make this happen in the ordinary
1339 (not cross-) SBCL compiler, nor in CMU CL.
1342 In sbcl-0.pre7.111 I added an assertion upstream, in IR2-CONVERT-CLOSURE,
1343 which fails for the test case above but doesn't keep the system
1344 from cross-compiling itself or passing its tests.
1346 I traced IR1-CONVERT-LAMBDA (with :PRINT *CURRENT-COMPONENT*)
1347 and tracing various COMPONENT-manipulating functions like
1348 FIND-INITIAL-DFO, DFO-SCAVENGE-DEPENDENCY-GRAPH,
1349 JOIN-COMPONENTS, LOCALL-ANALYZE-COMPONENT, etc. From that,
1350 it looks as though the problem is that IR1-CONVERT-LAMBDA
1351 is being called by MAKE-EXTERNAL-ENTRY-POINT to
1352 create the mislaid LAMBDA in an environment set up by
1353 WITH-BELATED-IR1-ENVIRONMENT which has *CURRENT-COMPONENT* set
1354 to a component which is never seen again, and specifically never
1355 passed to LOCALL-ANALYZE-COMPONENT or JOIN-COMPONENTS, so that
1356 its NEW-FUNS list (where the mislaid LAMBDA is waiting patiently)
1357 gets lost. Thus, the LAMBDA is essentially being written into never
1358 never land. But I haven't figured out why. *CURRENT-COMPONENT* is set
1359 wrong? Something later on is dropping the ball and neglecting
1360 to look at all the components it should? Something else?
1363 In sbcl-0.pre7.107, (DIRECTORY "*.*") is broken, as reported by
1364 Nathan Froyd sbcl-devel 2001-12-28.
1366 Christophe Rhodes suggested (sbcl-devel 2001-12-30) converting
1367 the MERGED-PATHNAME expression in DEFUN DIRECTORY to
1368 (merged-pathname (merge-pathnames pathname
1369 *default-pathname-defaults*))
1370 This looks right, and fixes this bug, but it interacts with the NODES
1371 logic in %ENUMERATE-PATHNAMES to create a new bug, so that
1372 (DIRECTORY "../**/*.*") no longer shows files in the current working
1373 directory. Probably %ENUMERATE-PATHNAMES (or related logic like
1374 %ENUMERATE-MATCHES) needs to be patched as well.
1376 Note: The MERGED-PATHNAME change changes behavior incompatibly,
1377 making e.g. (DIRECTORY "*") no longer equivalent to (DIRECTORY "*.*"),
1378 so deserves a NEWS entry. E.g.
1379 * minor incompatible change (part of a bug fix by Christophe Rhodes
1380 to DIRECTORY behavior): DIRECTORY no longer implicitly promotes
1381 NIL slots of its pathname argument to :WILD, and in particular
1382 asking for the contents of a directory, which you used to be able
1383 to do without explicit wildcards, e.g. (DIRECTORY "/tmp/"),
1384 now needs explicit wildcards, e.g. (DIRECTORY "/tmp/*.*").
1387 (reported by Alexey Dejneka sbcl-devel 2002-01-03)
1389 SUBTYPEP does not work well with redefined classes:
1391 * (defclass a () ())
1393 * (defclass b () ())
1398 * (defclass b (a) ())
1403 * (defclass b () ())
1412 KNOWN BUGS RELATED TO THE IR1 INTERPRETER
1414 (Now that the IR1 interpreter has gone away, these should be
1415 relatively straightforward to fix.)
1418 The system accepts DECLAIM in most places where DECLARE would be
1419 accepted, without even issuing a warning. ANSI allows this, but since
1420 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
1421 meaning is rather different, and it's unlikely that the user
1422 has a good reason for doing DECLAIM not at top level, it would be
1423 good to issue a STYLE-WARNING when this happens. A possible
1424 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
1425 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
1426 [This is considered an IR1-interpreter-related bug because until
1427 EVAL-WHEN is rewritten, which won't happen until after the IR1
1428 interpreter is gone, the system's notion of what's a top-level form
1429 and what's not will remain too confused to fix this problem.]
1432 (another wishlist thing..) Reimplement DEFMACRO to be basically
1433 like DEFMACRO-MUNDANELY, just using EVAL-WHEN.