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).
36 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
37 instead of just punting when a contradictory structure definition
38 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
39 changes their layout, you probably have to rebuild your entire
40 program, even if you know or guess enough about the internals of
41 SBCL to wager that this (undefined in ANSI) operation would be safe.
43 3: "type checking of structure slots"
45 ANSI specifies that a type mismatch in a structure slot
46 initialization value should not cause a warning.
48 This one might not be fixed for a while because while we're big
49 believers in ANSI compatibility and all, (1) there's no obvious
50 simple way to do it (short of disabling all warnings for type
51 mismatches everywhere), and (2) there's a good portable
52 workaround, and (3) by their own reasoning, it looks as though
53 ANSI may have gotten it wrong. ANSI justifies this specification
55 The restriction against issuing a warning for type mismatches
56 between a slot-initform and the corresponding slot's :TYPE
57 option is necessary because a slot-initform must be specified
58 in order to specify slot options; in some cases, no suitable
60 However, in SBCL (as in CMU CL or, for that matter, any compiler
61 which really understands Common Lisp types) a suitable default
62 does exist, in all cases, because the compiler understands the
63 concept of functions which never return (i.e. has return type NIL).
64 Thus, as a portable workaround, you can use a call to some
65 known-never-to-return function as the default. E.g.
67 (BAR (ERROR "missing :BAR argument")
68 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
70 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
71 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
72 (ERROR "missing required argument"))
74 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
75 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
76 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
77 Such code should compile without complaint and work correctly either
78 on SBCL or on any other completely compliant Common Lisp system.
80 b: &AUX argument in a boa-constructor without a default value means
81 "do not initilize this slot" and does not cause type error. But
82 an error may be signalled at read time and it would be good if
88 The "compiling top-level form:" output ought to be condensed.
89 Perhaps any number of such consecutive lines ought to turn into a
90 single "compiling top-level forms:" line.
93 (I *think* this is a bug. It certainly seems like strange behavior. But
94 the ANSI spec is scary, dark, and deep.. -- WHN)
95 (FORMAT NIL "~,1G" 1.4) => "1. "
96 (FORMAT NIL "~3,1G" 1.4) => "1. "
99 Sometimes (SB-EXT:QUIT) fails with
100 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
101 Process inferior-lisp exited abnormally with code 1
102 I haven't noticed a repeatable case of this yet.
105 The printer doesn't report closures very well. This is true in
109 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
110 It would be nice to make closures have a settable name slot,
111 and make things like DEFSTRUCT and FLET, which create closures,
112 set helpful values into this slot.
115 And as long as we're wishing, it would be awfully nice if INSPECT could
116 also report on closures, telling about the values of the bound variables.
119 The compiler assumes that any time a function of declared FTYPE
120 doesn't signal an error, its arguments were of the declared type.
121 E.g. compiling and loading
122 (DECLAIM (OPTIMIZE (SAFETY 3)))
123 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
125 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
127 (COND ((> (FACTORIAL X) 1.0E6)
128 (FORMAT T "too big~%"))
130 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
132 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
135 will cause the INTEGERP case to be selected, giving bogus output a la
137 This violates the "declarations are assertions" principle.
138 According to the ANSI spec, in the section "System Class FUNCTION",
139 this is a case of "lying to the compiler", but the lying is done
140 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
141 not by the unexpectedly general definition of FACTORIAL. In any case,
142 "declarations are assertions" means that lying to the compiler should
143 cause an error to be signalled, and should not cause a bogus
144 result to be returned. Thus, the compiler should not assume
145 that arbitrary functions check their argument types. (It might
146 make sense to add another flag (CHECKED?) to DEFKNOWN to
147 identify functions which *do* check their argument types.)
148 (Also, verify that the compiler handles declared function
149 return types as assertions.)
152 The definitions of SIGCONTEXT-FLOAT-REGISTER and
153 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
154 supported on FreeBSD because the floating point state is not saved,
155 but at least as of FreeBSD 4.0, the floating point state *is* saved,
156 so they could be supported after all. Very likely
157 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
160 The debugger LIST-LOCATIONS command doesn't work properly.
161 (How should it work properly?)
164 Compiling and loading
165 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
167 then requesting a BACKTRACE at the debugger prompt gives no information
168 about where in the user program the problem occurred.
170 (this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
171 while giving the backtrace the non-x86 systems complains about "unknown
172 source location: using block start", but apart from that the
173 backtrace seems reasonable. On x86 this is masked by bug 353. See
174 tests/debug.impure.lisp for a test case)
177 Using the pretty-printer from the command prompt gives funny
178 results, apparently because the pretty-printer doesn't know
179 about user's command input, including the user's carriage return
180 that the user, and therefore the pretty-printer thinks that
181 the new output block should start indented 2 or more characters
182 rightward of the correct location.
185 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
186 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
187 crashes SBCL. In general tracing anything which is used in the
188 implementation of TRACE is likely to have the same problem.
191 ANSI says in one place that type declarations can be abbreviated even
192 when the type name is not a symbol, e.g.
193 (DECLAIM ((VECTOR T) *FOOVECTOR*))
194 SBCL doesn't support this. But ANSI says in another place that this
195 isn't allowed. So it's not clear this is a bug after all. (See the
196 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
200 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
201 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
202 the mean of the distribution can be systematically O(0.1%) wrong.
203 Just increasing R-I-E-B is probably not a good solution, since
204 it would decrease efficiency more than is probably necessary. Perhaps
205 using some sort of accept/reject method would be better.
208 Internally the compiler sometimes evaluates
209 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
210 (I stumbled across this when I added an
211 (assert (not (eq type1 *wild-type*)))
212 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
213 in a type context should probably be translated to T, and so it's
214 probably wrong to ask whether it's equal to the T type and then (using
215 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
216 (I haven't tried to investigate this bug enough to guess whether
217 there might be any user-level symptoms.)
219 In fact, the type system is likely to depend on this inequality not
220 holding... * is not equivalent to T in many cases, such as
221 (VECTOR *) /= (VECTOR T).
224 The facility for dumping a running Lisp image to disk gets confused
225 when run without the PURIFY option, and creates an unnecessarily large
226 core file (apparently representing memory usage up to the previous
227 high-water mark). Moreover, when the file is loaded, it confuses the
228 GC, so that thereafter memory usage can never be reduced below that
231 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
232 The interaction between gencgc and the variables used by
233 save-lisp-and-die is still nonoptimal, though, so no respite from
237 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
238 CL), out-of-line structure slot setters are horribly inefficient
239 whenever the type of the slot is declared, because out-of-line
240 structure slot setters are implemented as closures to save space,
241 so the compiler doesn't compile the type test into code, but
242 instead just saves the type in a lexical closure and interprets it
244 To exercise the problem, compile and load
245 (cl:in-package :cl-user)
247 (bar (error "missing") :type bar))
250 (loop (setf (foo-bar *foo*) x)))
252 (defvar *bar* (make-bar))
253 (defvar *foo* (make-foo :bar *bar*))
254 (defvar *setf-foo-bar* #'(setf foo-bar))
256 (loop (funcall *setf-foo-bar* x *foo*)))
257 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
258 use BACKTRACE, to see it's spending all essentially all its time
259 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
260 One possible solution would be simply to give up on
261 representing structure slot accessors as functions, and represent
262 them as macroexpansions instead. This can be inconvenient for users,
263 but it's not clear that it's worse than trying to help by expanding
264 into a horribly inefficient implementation.
265 As a workaround for the problem, #'(SETF FOO) expressions
266 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
267 (defmacro efficient-setf-function (place-function-name)
268 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
269 ;; a workaround for the problem, encouraging the
270 ;; inline expansion of the structure accessor, so
271 ;; that the compiler can optimize its type test
272 (let ((new-value (gensym "NEW-VALUE-"))
273 (structure-value (gensym "STRUCTURE-VALUE-")))
274 `(lambda (,new-value ,structure-value)
275 (setf (,place-function-name ,structure-value)
277 ;; no problem, can just use the ordinary expansion
278 `(function (setf ,place-function-name))))
281 There's apparently a bug in CEILING optimization which caused
282 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
283 applied the patches to SBCL and they didn't seem to cause problems
284 (as reported sbcl-devel 2001-05-04). However, since the patches
285 modify nontrivial code which was apparently written incorrectly
286 the first time around, until regression tests are written I'm not
287 comfortable merging the patches in the CVS version of SBCL.
290 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
291 a clean, just-started SBCL system. And it seems to be right:
292 (ROOM T) can bring a small computer to its knees for a *long*
293 time trying to GC afterwards. Surely there's some more economical
294 way to implement (ROOM T).
296 Daniel Barlow doesn't know what fixed this, but observes that it
297 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
298 in a fresh SBCL causes
300 debugger invoked on a SB-INT:BUG in thread 5911:
301 failed AVER: "(SAP= CURRENT END)"
303 unless a GC has happened beforehand.
306 When the compiler inline expands functions, it may be that different
307 kinds of return values are generated from different code branches.
308 E.g. an inline expansion of POSITION generates integer results
309 from one branch, and NIL results from another. When that inline
310 expansion is used in a context where only one of those results
313 (aref *a1* (position x *a2*)))
314 and the compiler can't prove that the unacceptable branch is
315 never taken, then bogus type mismatch warnings can be generated.
316 If you need to suppress the type mismatch warnings, you can
317 suppress the inline expansion,
319 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
320 (aref *a1* (position x *a2*)))
321 or, sometimes, suppress them by declaring the result to be of an
324 (aref *a1* (the integer (position x *a2*))))
326 This is not a new compiler problem in 0.7.0, but the new compiler
327 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
328 more conspicuous. If you don't need performance from these functions,
329 and the bogus warnings are a nuisance for you, you can return to
330 your pre-0.7.0 state of grace with
331 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
336 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
337 the entire lexical environment at the point of MACROLET available
338 in the bodies of the macroexpander functions. In particular, it
339 allows the function bodies (which run at compile time) to try to
340 access lexical variables (which are only defined at runtime).
341 It doesn't even issue a warning, which is bad.
343 The SBCL behavior arguably conforms to the ANSI spec (since the
344 spec says that the behavior is undefined, ergo anything conforms).
345 However, it would be better to issue a compile-time error.
346 Unfortunately I (WHN) don't see any simple way to detect this
347 condition in order to issue such an error, so for the meantime
348 SBCL just does this weird broken "conforming" thing.
350 The ANSI standard says, in the definition of the special operator
352 The macro-expansion functions defined by MACROLET are defined
353 in the lexical environment in which the MACROLET form appears.
354 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
355 the local macro definitions in a MACROLET, but the consequences
356 are undefined if the local macro definitions reference any
357 local variable or function bindings that are visible in that
359 Then it seems to contradict itself by giving the example
361 (macrolet ((fudge (z)
362 ;The parameters x and flag are not accessible
363 ; at this point; a reference to flag would be to
364 ; the global variable of that name.
365 ` (if flag (* ,z ,z) ,z)))
366 ;The parameters x and flag are accessible here.
370 The comment "a reference to flag would be to the global variable
371 of the same name" sounds like good behavior for the system to have.
372 but actual specification quoted above says that the actual behavior
375 (Since 0.7.8.23 macroexpanders are defined in a restricted version
376 of the lexical environment, containing no lexical variables and
377 functions, which seems to conform to ANSI and CLtL2, but signalling
378 a STYLE-WARNING for references to variables similar to locals might
382 (as reported by Gabe Garza on cmucl-help 2001-09-21)
384 (defun test-pred (x y)
388 (func (lambda () x)))
389 (print (eq func func))
390 (print (test-pred func func))
391 (delete func (list func))))
392 Now calling (TEST-CASE) gives output
395 (#<FUNCTION {500A9EF9}>)
396 Evidently Python thinks of the lambda as a code transformation so
397 much that it forgets that it's also an object.
400 Ideally, uninterning a symbol would allow it, and its associated
401 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
402 at least as of sbcl-0.7.0, this isn't the case. Information about
403 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
404 essentially in ordinary (non-weak) hash tables keyed by symbols.
405 Thus, once a system has an entry in this system, it tends to live
406 forever, even when it is uninterned and all other references to it
410 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
411 prominent SourceForge web/db bug tracking system, which is
412 unfortunately not a reliable way to get a timely response from
413 the SBCL maintainers)
414 In the course of trying to build a test case for an
415 application error, I encountered this behavior:
416 If you start up sbcl, and then lay on CTRL-C for a
417 minute or two, the lisp process will eventually say:
418 %PRIMITIVE HALT called; the party is over.
419 and throw you into the monitor. If I start up lisp,
420 attach to the process with strace, and then do the same
421 (abusive) thing, I get instead:
422 access failure in heap page not marked as write-protected
423 and the monitor again. I don't know enough to have the
424 faintest idea of what is going on here.
425 This is with sbcl 6.12, uname -a reports:
426 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
427 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
428 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
429 it took more than two minutes (but less than five) for me.
433 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
434 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
435 COMPLEX implementation didn't deal with this, and hasn't been
436 upgraded to do so. (This doesn't seem to be a high priority
437 conformance problem, since seems hard to construct useful code
440 b. (fixed in 0.8.3.43)
443 Floating point errors are reported poorly. E.g. on x86 OpenBSD
446 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
447 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
448 No traps are enabled? How can this be?
449 It should be possible to be much more specific (overflow, division
450 by zero, etc.) and of course the "How can this be?" should be fixable.
452 See also bugs #45.c and #183
455 (reported by Robert E. Brown 2002-04-16)
456 When a function is called with too few arguments, causing the
457 debugger to be entered, the uninitialized slots in the bad call frame
458 seem to cause GCish problems, being interpreted as tagged data even
459 though they're not. In particular, executing ROOM in the
460 debugger at that point causes AVER failures:
463 * (lisp-implementation-version)
469 failed AVER: "(SAP= CURRENT END)"
470 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
471 isn't too surprising since there are many differences in stack
472 implementation and GC conservatism between the X86 and other ports.)
474 This is probably the same bug as 216
477 The compiler sometimes tries to constant-fold expressions before
478 it checks to see whether they can be reached. This can lead to
479 bogus warnings about errors in the constant folding, e.g. in code
482 (WRITE-STRING (> X 0) "+" "0"))
483 compiled in a context where the compiler can prove that X is NIL,
484 and the compiler complains that (> X 0) causes a type error because
485 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
486 caused a full WARNING, which made the bug really annoying because then
487 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
488 code. Since then the warning has been downgraded to STYLE-WARNING,
489 so it's still a bug but at least it's a little less annoying.
491 183: "IEEE floating point issues"
492 Even where floating point handling is being dealt with relatively
493 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
494 accrued-exceptions and current-exceptions part of the fp control
495 word don't seem to bear much relation to reality. E.g. on
499 debugger invoked on condition of type DIVISION-BY-ZERO:
500 arithmetic error DIVISION-BY-ZERO signalled
501 0] (sb-vm::get-floating-point-modes)
503 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
504 :ROUNDING-MODE :NEAREST
505 :CURRENT-EXCEPTIONS NIL
506 :ACCRUED-EXCEPTIONS (:INEXACT)
509 * (sb-vm::get-floating-point-modes)
510 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
511 :ROUNDING-MODE :NEAREST
512 :CURRENT-EXCEPTIONS (:INEXACT)
513 :ACCRUED-EXCEPTIONS (:INEXACT)
516 188: "compiler performance fiasco involving type inference and UNION-TYPE"
520 (declare (optimize (safety 3)))
521 (declare (optimize (compilation-speed 2)))
522 (declare (optimize (speed 1) (debug 1) (space 1)))
524 (declare (type (integer 0) start))
525 (print (incf start 22))
526 (print (incf start 26))
527 (print (incf start 28)))
529 (declare (type (integer 0) start))
530 (print (incf start 22))
531 (print (incf start 26)))
533 (declare (type (integer 0) start))
534 (print (incf start 22))
535 (print (incf start 26))))))
537 This example could be solved with clever enough constraint
538 propagation or with SSA, but consider
543 The careful type of X is {2k} :-(. Is it really important to be
544 able to work with unions of many intervals?
546 191: "Miscellaneous PCL deficiencies"
547 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
548 a. DEFCLASS does not inform the compiler about generated
549 functions. Compiling a file with
553 (WITH-SLOTS (A-CLASS-X) A
555 results in a STYLE-WARNING:
557 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
559 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
560 Mai points out that the declamation of functions is in fact
561 incorrect in some cases (most notably for structure
562 classes). This means that at present erroneous attempts to use
563 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
564 won't get the corresponding STYLE-WARNING.
565 c. (fixed in 0.8.4.23)
567 201: "Incautious type inference from compound types"
568 a. (reported by APD sbcl-devel 2002-09-17)
570 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
572 (FORMAT NIL "~S IS ~S, Y = ~S"
579 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
583 (declare (type (array * (4 4)) x))
585 (setq x (make-array '(4 4)))
586 (adjust-array y '(3 5))
587 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
589 * (foo (make-array '(4 4) :adjustable t))
592 205: "environment issues in cross compiler"
593 (These bugs have no impact on user code, but should be fixed or
595 a. Macroexpanders introduced with MACROLET are defined in the null
597 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
598 the null lexical environment.
599 c. The cross-compiler cannot inline functions defined in a non-null
602 206: ":SB-FLUID feature broken"
603 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
604 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
607 207: "poorly distributed SXHASH results for compound data"
608 SBCL's SXHASH could probably try a little harder. ANSI: "the
609 intent is that an implementation should make a good-faith
610 effort to produce hash-codes that are well distributed
611 within the range of non-negative fixnums". But
612 (let ((hits (make-hash-table)))
615 (let* ((ij (cons i j))
616 (newlist (push ij (gethash (sxhash ij) hits))))
618 (format t "~&collision: ~S~%" newlist))))))
619 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
620 would be an obvious way of fix. Maybe it would be acceptably efficient
621 to redo MIX using a lookup into a 256-entry s-box containing
622 29-bit pseudorandom numbers?
624 211: "keywords processing"
625 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
626 number of keyword arguments.
629 (flet ((foo (&key y) (list y)))
630 (list (foo :y 1 :y 2)))
632 issues confusing message
637 ; caught STYLE-WARNING:
638 ; The variable #:G15 is defined but never used.
640 212: "Sequence functions and circular arguments"
641 COERCE, MERGE and CONCATENATE go into an infinite loop when given
642 circular arguments; it would be good for the user if they could be
643 given an error instead (ANSI 17.1.1 allows this behaviour on the part
644 of the implementation, as conforming code cannot give non-proper
645 sequences to these functions. MAP also has this problem (and
646 solution), though arguably the convenience of being able to do
647 (MAP 'LIST '+ FOO '#1=(1 . #1#))
648 might be classed as more important (though signalling an error when
649 all of the arguments are circular is probably desireable).
651 213: "Sequence functions and type checking"
652 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
653 check that it will return a sequence of the given type. Fixing
654 it along the same lines as the others (cf. work done around
655 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
656 entirely straightforward.
657 c. All of these functions will silently accept a type of the form
659 whether or not the return value is of this type. This is
660 probably permitted by ANSI (see "Exceptional Situations" under
661 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
662 know about this escape clause, so code of the form
663 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
664 can erroneously return T.
666 215: ":TEST-NOT handling by functions"
667 a. FIND and POSITION currently signal errors when given non-NIL for
668 both their :TEST and (deprecated) :TEST-NOT arguments, but by
669 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
670 means that the effect is "unpredictable but harmless". It's not
671 clear what that actually means; it may preclude conforming
672 implementations from signalling errors.
673 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
674 when conflict occurs. As a quality of implementation issue, it
675 might be preferable to treat :TEST and :TEST-NOT as being in some
676 sense the same &KEY, and effectively take the first test function in
678 c. Again, a quality of implementation issue: it would be good to issue a
679 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
680 WARNING for calls with both :TEST and :TEST-NOT; possibly this
681 latter should be WARNed about at execute-time too.
683 216: "debugger confused by frames with invalid number of arguments"
684 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
685 leaves the system confused, enough so that (QUIT) no longer works.
686 It's as though the process of working with the uninitialized slot in
687 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
688 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
690 This is probably the same bug as 162
692 217: "Bad type operations with FUNCTION types"
695 * (values-type-union (specifier-type '(function (base-char)))
696 (specifier-type '(function (integer))))
698 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
700 It causes insertion of wrong type assertions into generated
704 (let ((f (etypecase x
705 (character #'write-char)
706 (integer #'write-byte))))
709 (character (write-char x s))
710 (integer (write-byte x s)))))
712 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
714 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
715 produce invalid code, but type checking is not accurate.)
717 233: bugs in constraint propagation
719 (declaim (optimize (speed 2) (safety 3)))
721 (if (typep (prog1 x (setq x y)) 'double-float)
724 (foo 1d0 5) => segmentation violation
726 235: "type system and inline expansion"
728 (declaim (ftype (function (cons) number) acc))
729 (declaim (inline acc))
731 (the number (car c)))
734 (values (locally (declare (optimize (safety 0)))
736 (locally (declare (optimize (safety 3)))
739 (foo '(nil) '(t)) => NIL, T.
741 237: "Environment arguments to type functions"
742 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
743 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
744 argument, but they ignore it completely. This is almost
745 certainly not correct.
746 b. Also, the compiler's optimizers for TYPEP have not been informed
747 about the new argument; consequently, they will not transform
748 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
749 just as optimizeable as (TYPEP 1 'INTEGER).
751 238: "REPL compiler overenthusiasm for CLOS code"
753 * (defclass foo () ())
754 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
755 causes approximately 100 lines of code deletion notes. Some
756 discussion on this issue happened under the title 'Three "interesting"
757 bugs in PCL', resulting in a fix for this oververbosity from the
758 compiler proper; however, the problem persists in the interactor
759 because the notion of original source is not preserved: for the
760 compiler, the original source of the above expression is (DEFMETHOD
761 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
762 compiler gets its hands on the code needing compilation from the REPL,
763 it has been macroexpanded several times.
765 A symptom of the same underlying problem, reported by Tony Martinez:
767 (with-input-from-string (*query-io* " no")
769 (simple-type-error () 'error))
771 ; (SB-KERNEL:FLOAT-WAIT)
773 ; note: deleting unreachable code
774 ; compilation unit finished
777 242: "WRITE-SEQUENCE suboptimality"
778 (observed from clx performance)
779 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
780 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
781 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
782 rather than writing the sequence in one go, leading to severe
783 performance degradation.
785 243: "STYLE-WARNING overenthusiasm for unused variables"
786 (observed from clx compilation)
787 In sbcl-0.7.14, in the presence of the macros
788 (DEFMACRO FOO (X) `(BAR ,X))
789 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
790 somewhat surprising style warnings are emitted for
791 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
793 ; (LAMBDA (Y) (FOO Y))
795 ; caught STYLE-WARNING:
796 ; The variable Y is defined but never used.
798 245: bugs in disassembler
799 b. On X86 operand size prefix is not recognized.
802 (defun foo (&key (a :x))
806 does not cause a warning. (BTW: old SBCL issued a warning, but for a
807 function, which was never called!)
810 Compiler does not emit warnings for
812 a. (lambda () (svref (make-array 8 :adjustable t) 1))
815 (list (let ((y (the real x)))
816 (unless (floatp y) (error ""))
821 (declare (optimize (debug 0)))
822 (declare (type vector x))
823 (list (fill-pointer x)
827 Complex array type does not have corresponding type specifier.
829 This is a problem because the compiler emits optimization notes when
830 you use a non-simple array, and without a type specifier for hairy
831 array types, there's no good way to tell it you're doing it
832 intentionally so that it should shut up and just compile the code.
834 Another problem is confusing error message "asserted type ARRAY
835 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
836 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
838 The last problem is that when type assertions are converted to type
839 checks, types are represented with type specifiers, so we could lose
840 complex attribute. (Now this is probably not important, because
841 currently checks for complex arrays seem to be performed by
845 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
846 warning. Analogous cases with the index and length being equal and
847 greater than 0 are warned for; the problem here seems to be that the
848 type required for an array reference of this type is (INTEGER 0 (0))
849 which is canonicalized to NIL.
854 (t1 (specifier-type s)))
855 (eval `(defstruct ,s))
856 (type= t1 (specifier-type s)))
861 b. The same for CSUBTYPEP.
863 262: "yet another bug in inline expansion of local functions"
864 During inline expansion of a local function Python can try to
865 reference optimized away objects (functions, variables, CTRANs from
866 tags and blocks), which later may lead to problems. Some of the
867 cases are worked around by forbidding expansion in such cases, but
868 the better way would be to reimplement inline expansion by copying
872 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
873 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
874 looks reasonable, if not obviously correct; however, it caused the
875 PPC/Linux port to segfault during warm-init while loading
876 src/pcl/std-class.fasl. A workaround patch was made, but it would
877 be nice to understand why the first patch caused problems, and to
878 fix the cause if possible.
880 268: "wrong free declaration scope"
881 The following code must signal type error:
883 (locally (declare (optimize (safety 3)))
884 (flet ((foo (x &optional (y (car x)))
885 (declare (optimize (safety 0)))
887 (funcall (eval #'foo) 1)))
890 In the following function constraint propagator optimizes nothing:
893 (declare (integer x))
894 (declare (optimize speed))
902 Compilation of the following two forms causes "X is unbound" error:
904 (symbol-macrolet ((x pi))
905 (macrolet ((foo (y) (+ x y)))
906 (declaim (inline bar))
912 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
915 CLHS says that type declaration of a symbol macro should not affect
916 its expansion, but in SBCL it does. (If you like magic and want to
917 fix it, don't forget to change all uses of MACROEXPAND to
921 The following code (taken from CLOCC) takes a lot of time to compile:
924 (declare (type (integer 0 #.large-constant) n))
927 (fixed in 0.8.2.51, but a test case would be good)
932 (declare (optimize speed))
933 (loop for i of-type (integer 0) from 0 by 2 below 10
936 uses generic arithmetic.
938 b. (fixed in 0.8.3.6)
940 279: type propagation error -- correctly inferred type goes astray?
941 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
942 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
943 &OPTIONAL), not a (INTEGER 1 536870911)
944 is emitted when compiling this file:
945 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
946 (integer #.most-negative-fixnum 0))
951 (let* (;; Uncomment this for a type mismatch warning indicating
952 ;; that the type of (FOO X) is correctly understood.
953 #+nil (fs-foo (float-sign (foo x)))
954 ;; Uncomment this for a type mismatch warning
955 ;; indicating that the type of (ABS (FOO X)) is
956 ;; correctly understood.
957 #+nil (fs-abs-foo (float-sign (abs (foo x))))
958 ;; something wrong with this one though
959 (abs-foo (abs (foo x))))
960 (declare (type (integer 1 100) abs-foo))
965 281: COMPUTE-EFFECTIVE-METHOD error signalling.
966 (slightly obscured by a non-0 default value for
967 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
968 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
969 when it finds a method with invalid qualifiers. However, it
970 shouldn't signal errors when any such methods are not applicable to
971 the particular call being evaluated, and certainly it shouldn't when
972 simply precomputing effective methods that may never be called.
973 (setf sb-pcl::*max-emf-precompute-methods* 0)
975 (:method-combination +)
976 (:method ((x symbol)) 1)
977 (:method + ((x number)) x))
978 (foo 1) -> ERROR, but should simply return 1
980 The issue seems to be that construction of a discriminating function
981 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
983 283: Thread safety: libc functions
984 There are places that we call unsafe-for-threading libc functions
985 that we should find alternatives for, or put locks around. Known or
986 strongly suspected problems, as of 0.8.3.10: please update this
987 bug instead of creating new ones
989 localtime() - called for timezone calculations in code/time.lisp
991 284: Thread safety: special variables
992 There are lots of special variables in SBCL, and I feel sure that at
993 least some of them are indicative of potentially thread-unsafe
994 parts of the system. See doc/internals/notes/threading-specials
996 286: "recursive known functions"
997 Self-call recognition conflicts with known function
998 recognition. Currently cross compiler and target COMPILE do not
999 recognize recursion, and in target compiler it can be disabled. We
1000 can always disable it for known functions with RECURSIVE attribute,
1001 but there remains a possibility of a function with a
1002 (tail)-recursive simplification pass and transforms/VOPs for base
1005 287: PPC/Linux miscompilation or corruption in first GC
1006 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
1007 segmentation fault is reported at the point of first triggered GC,
1008 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
1009 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
1010 that this merely obscures, not solves, the underlying problem; as and when
1011 underlying problems are fixed, it would be worth trying again to provoke
1014 288: fundamental cross-compilation issues (from old UGLINESS file)
1015 Using host floating point numbers to represent target floating point
1016 numbers, or host characters to represent target characters, is
1017 theoretically shaky. (The characters are OK as long as the characters
1018 are in the ANSI-guaranteed character set, though, so they aren't a
1019 real problem as long as the sources don't need anything but that;
1020 the floats are a real problem.)
1022 289: "type checking and source-transforms"
1024 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1027 Our policy is to check argument types at the moment of a call. It
1028 disagrees with ANSI, which says that type assertions are put
1029 immediately onto argument expressions, but is easier to implement in
1030 IR1 and is more compatible to type inference, inline expansion,
1031 etc. IR1-transforms automatically keep this policy, but source
1032 transforms for associative functions (such as +), being applied
1033 during IR1-convertion, do not. It may be tolerable for direct calls
1034 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1036 b. Another aspect of this problem is efficiency. [x y + z +]
1037 requires less registers than [x y z + +]. This transformation is
1038 currently performed with source transforms, but it would be good to
1039 also perform it in IR1 optimization phase.
1041 290: Alpha floating point and denormalized traps
1042 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1043 hardware or kernel deficiency: the status of the enable/disable
1044 denormalized-float traps bit seems to be ambiguous; by the time we
1045 get to os_restore_fp_control after a trap, denormalized traps seem
1046 to be enabled. Since we don't want a trap every time someone uses a
1047 denormalized float, in general, we mask out that bit when we restore
1048 the control word; however, this clobbers any change the user might
1052 (reported by Adam Warner, sbcl-devel 2003-09-23)
1054 The --load toplevel argument does not perform any sanitization of its
1055 argument. As a result, files with Lisp pathname pattern characters
1056 (#\* or #\?, for instance) or quotation marks can cause the system
1057 to perform arbitrary behaviour.
1060 LOOP with non-constant arithmetic step clauses suffers from overzealous
1061 type constraint: code of the form
1062 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1063 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1064 (0))). However, an integral value of X should be legal, because
1065 successive adds of integers to double-floats produces double-floats,
1066 so none of the type restrictions in the code is violated.
1068 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1069 argument type only after having read a character. This is caused
1070 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1071 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1072 less error prone to have EXPLICIT-CHECK be a local declaration,
1073 being put into the definition, instead of an attribute being kept in
1074 a separate file; maybe also put it into SB-EXT?
1076 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1077 in NOTE-ASSUMED-TYPES
1078 In sbcl-0.8.7.32, compiling the file
1080 (declare (type integer x))
1081 (declare (type (vector (or hash-table bit)) y))
1084 (declare (type integer x))
1085 (declare (type (simple-array base (2)) y))
1088 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1090 303: "nonlinear LVARs" (aka MISC.293)
1092 (multiple-value-call #'list
1094 (multiple-value-prog1
1095 (eval '(values :a :b :c))
1101 (throw 'bar (values 3 4)))))))))))
1103 (BUU 1) returns garbage.
1105 The problem is that both EVALs sequentially write to the same LVAR.
1108 (Reported by Dave Roberts.)
1109 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1112 (declare (ftype (function () (integer 0 10)) fee)
1116 uses generic arithmetic with INLINE and fixnum without.
1118 306: "Imprecise unions of array types"
1120 (declare (optimize speed)
1121 (type (or (array cons) (array vector)) x))
1123 (foo #((0))) => TYPE-ERROR
1130 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1131 collect `(array ,(sb-vm:saetp-specifier x)))))
1132 => NIL, T (when it should be T, T)
1134 309: "Dubious values for implementation limits"
1135 (reported by Bruno Haible sbcl-devel "Incorrect value of
1136 multiple-values-limit" 2004-04-19)
1137 (values-list (make-list 1000000)), on x86/linux, signals a stack
1138 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1139 significantly larger than 1000000. There are probably similar
1140 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1141 around the same time regarding a call to LIST on sparc with 1000
1142 arguments) and other implementation limit constants.
1144 311: "Tokeniser not thread-safe"
1145 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1147 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1148 spurious errors should two threads attempt to tokenise at the same
1151 314: "LOOP :INITIALLY clauses and scope of initializers"
1152 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1153 test suite, originally by Thomas F. Burdick.
1154 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1155 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1156 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1157 ;; initially clauses are evaluated in the loop prologue, which precedes all
1158 ;; loop code except for the initial settings provided by with, for, or as.
1159 (loop :for x = 0 :then (1+ x)
1160 :for y = (1+ x) :then (ash y 1)
1161 :for z :across #(1 3 9 27 81 243)
1163 :initially (assert (zerop x)) :initially (assert (= 2 w))
1164 :until (>= w 100) :collect w)
1165 Expected: (2 6 15 38)
1168 317: "FORMAT of floating point numbers"
1169 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1171 (format nil "~1F" 10) => "0." ; "10." expected
1172 (format nil "~0F" 10) => "0." ; "10." expected
1173 (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected
1174 it would be nice if whatever fixed this also untangled the two
1175 competing implementations of floating point printing (Steele and
1176 White, and Burger and Dybvig) present in src/code/print.lisp
1178 318: "stack overflow in compiler warning with redefined class"
1179 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1182 (setf (find-class 'foo) nil)
1183 (defstruct foo slot-1)
1184 This used to give a stack overflow from within the printer, which has
1185 been fixed as of 0.8.16.11. Current result:
1187 ; can't compile TYPEP of anonymous or undefined class:
1188 ; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
1190 debugger invoked on a TYPE-ERROR in thread 19973:
1191 The value NIL is not of type FUNCTION.
1193 CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
1194 meant to be enough to delete structure classes from the system?
1196 319: "backquote with comma inside array"
1197 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1199 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1201 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1202 which probably isn't intentional.
1204 323: "REPLACE, BIT-BASH and large strings"
1205 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1206 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1208 (declare (optimize speed (safety 1)))
1209 (let ((x (make-string 140000000))
1210 (y (make-string 140000000)))
1211 (length (replace x y))))
1214 debugger invoked on a TYPE-ERROR in thread 2412:
1215 The value 1120000000 is not of type (MOD 536870911).
1216 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1218 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1219 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1220 for all positive integral <x>. At present, it only works for <x> up
1221 to about 1024 (and similarly for signed-byte), so
1222 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1223 gives an error in sbcl-0.8.10.
1225 325: "CLOSE :ABORT T on supeseding streams"
1226 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1227 file on disk, even if one existed before opening.
1229 The illegality of this is not crystal clear, as the ANSI dictionary
1230 entry for CLOSE says that when :ABORT is T superseded files are not
1231 superseded (ie. the original should be restored), whereas the OPEN
1232 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1233 implementation should not destroy the old file until the new stream
1234 is closed." -- implying that even though undesirable, early deletion
1235 is legal. Restoring the original would none the less be the polite
1238 326: "*PRINT-CIRCLE* crosstalk between streams"
1239 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1240 mixed between streams when output operations are intermingled closely
1241 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1242 test case below), so that e.g. the references #2# appears on a stream
1243 with no preceding #2= on that stream to define it (because the #2= was
1244 sent to another stream).
1245 (cl:in-package :cl-user)
1246 (defstruct foo index)
1247 (defparameter *foo* (make-foo :index 4))
1249 (defparameter *bar* (make-bar))
1250 (defparameter *tangle* (list *foo* *bar* *foo*))
1251 (defmethod print-object ((foo foo) stream)
1252 (let ((index (foo-index foo)))
1253 (format *trace-output*
1254 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1256 (format stream "[FOO ~D]" index))
1258 (let ((tsos (make-string-output-stream))
1259 (ssos (make-string-output-stream)))
1260 (let ((*print-circle* t)
1261 (*trace-output* tsos)
1262 (*standard-output* ssos))
1263 (prin1 *tangle* *standard-output*))
1264 (let ((string (get-output-stream-string ssos)))
1265 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1266 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1267 (error "oops: ~S" string)))))
1268 It might be straightforward to fix this by turning the
1269 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1270 per-stream slots, but (1) it would probably be sort of messy faking
1271 up the special variable binding semantics using UNWIND-PROTECT and
1272 (2) it might be sort of a pain to test that no other bugs had been
1275 328: "Profiling generic functions", transplanted from #241
1276 (from tonyms on #lisp IRC 2003-02-25)
1277 In sbcl-0.7.12.55, typing
1278 (defclass foo () ((bar :accessor foo-bar)))
1281 (defclass foo () ((bar :accessor foo-bar)))
1282 gives the error message
1283 "#:FOO-BAR already names an ordinary function or a macro."
1285 Problem: when a generic function is profiled, it appears as an ordinary
1286 function to PCL. (Remembering the uninterned accessor is OK, as the
1287 redefinition must be able to remove old accessors from their generic
1290 329: "Sequential class redefinition"
1291 reported by Bruno Haible:
1292 (defclass reactor () ((max-temp :initform 10000000)))
1293 (defvar *r1* (make-instance 'reactor))
1294 (defvar *r2* (make-instance 'reactor))
1295 (slot-value *r1* 'max-temp)
1296 (slot-value *r2* 'max-temp)
1297 (defclass reactor () ((uptime :initform 0)))
1298 (slot-value *r1* 'uptime)
1299 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1300 (slot-value *r1* 'max-temp) ; => 10000
1301 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1304 The method effective when the wrapper is obsoleted can be saved
1305 in the wrapper, and then to update the instance just run through
1306 all the old wrappers in order from oldest to newest.
1308 332: "fasl stack inconsistency in structure redefinition"
1309 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1310 Even though structure redefinition is undefined by the standard, the
1311 following behaviour is suboptimal: running
1312 (defun stimulate-sbcl ()
1313 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1314 ;;create a file which redefines a structure incompatibly
1315 (with-open-file (f filename :direction :output :if-exists :supersede)
1316 (print '(defstruct astruct foo) f)
1317 (print '(defstruct astruct foo bar) f))
1318 ;;compile and load the file, then invoke the continue restart on
1319 ;;the structure redefinition error
1320 (handler-bind ((error (lambda (c) (continue c))))
1321 (load (compile-file filename)))))
1323 and choosing the CONTINUE restart yields the message
1324 debugger invoked on a SB-INT:BUG in thread 27726:
1325 fasl stack not empty when it should be
1327 336: "slot-definitions must retain the generic functions of accessors"
1328 reported by Tony Martinez:
1329 (defclass foo () ((bar :reader foo-bar)))
1330 (defun foo-bar (x) x)
1331 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1333 Note: just punting the accessor removal if the fdefinition
1334 is not a generic function is not enough:
1336 (defclass foo () ((bar :reader foo-bar)))
1337 (defvar *reader* #'foo-bar)
1338 (defun foo-bar (x) x)
1339 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1340 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1341 ; the method must be removed
1342 ; by the class redefinition
1344 Fixing this should also fix a subset of #328 -- update the
1345 description with a new test-case then.
1347 337: MAKE-METHOD and user-defined method classes
1348 (reported by Bruno Haible sbcl-devel 2004-06-11)
1352 (defclass user-method (standard-method) (myslot))
1353 (defmacro def-user-method (name &rest rest)
1354 (let* ((lambdalist-position (position-if #'listp rest))
1355 (qualifiers (subseq rest 0 lambdalist-position))
1356 (lambdalist (elt rest lambdalist-position))
1357 (body (subseq rest (+ lambdalist-position 1)))
1359 (subseq lambdalist 0 (or
1361 (lambda (x) (member x lambda-list-keywords))
1363 (length lambdalist))))
1364 (specializers (mapcar #'find-class
1365 (mapcar (lambda (x) (if (consp x) (second x) t))
1367 (unspecialized-required-part
1368 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1369 (unspecialized-lambdalist
1370 (append unspecialized-required-part
1371 (subseq lambdalist (length required-part)))))
1374 (MAKE-INSTANCE 'USER-METHOD
1375 :QUALIFIERS ',qualifiers
1376 :LAMBDA-LIST ',unspecialized-lambdalist
1377 :SPECIALIZERS ',specializers
1379 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1380 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1381 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1382 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1383 (IF (NULL NEXT-METHODS-LIST)
1384 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1385 (FUNCALL (SB-PCL:METHOD-FUNCTION
1386 (FIRST NEXT-METHODS-LIST))
1387 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1388 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1392 (defgeneric test-um03 (x))
1393 (defmethod test-um03 ((x integer))
1394 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1395 (def-user-method test-um03 ((x rational))
1396 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1397 (defmethod test-um03 ((x real))
1398 (list 'real x (not (null (next-method-p)))))
1403 (defgeneric test-um10 (x))
1404 (defmethod test-um10 ((x integer))
1405 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1406 (defmethod test-um10 ((x rational))
1407 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1408 (defmethod test-um10 ((x real))
1409 (list 'real x (not (null (next-method-p)))))
1410 (defmethod test-um10 :after ((x real)))
1411 (def-user-method test-um10 :around ((x integer))
1412 (list* 'around-integer x
1413 (not (null (next-method-p))) (call-next-method)))
1414 (defmethod test-um10 :around ((x rational))
1415 (list* 'around-rational x
1416 (not (null (next-method-p))) (call-next-method)))
1417 (defmethod test-um10 :around ((x real))
1418 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1420 fails with a type error, and
1423 (defgeneric test-um12 (x))
1424 (defmethod test-um12 ((x integer))
1425 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1426 (defmethod test-um12 ((x rational))
1427 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1428 (defmethod test-um12 ((x real))
1429 (list 'real x (not (null (next-method-p)))))
1430 (defmethod test-um12 :after ((x real)))
1431 (defmethod test-um12 :around ((x integer))
1432 (list* 'around-integer x
1433 (not (null (next-method-p))) (call-next-method)))
1434 (defmethod test-um12 :around ((x rational))
1435 (list* 'around-rational x
1436 (not (null (next-method-p))) (call-next-method)))
1437 (def-user-method test-um12 :around ((x real))
1438 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1440 fails with NO-APPLICABLE-METHOD.
1442 339: "DEFINE-METHOD-COMBINATION bugs"
1443 (reported by Bruno Haible via the clisp test suite)
1445 a. Syntax checking laxity (should produce errors):
1446 i. (define-method-combination foo :documentation :operator)
1447 ii. (define-method-combination foo :documentation nil)
1448 iii. (define-method-combination foo nil)
1449 iv. (define-method-combination foo nil nil
1450 (:arguments order &aux &key))
1451 v. (define-method-combination foo nil nil (:arguments &whole))
1452 vi. (define-method-combination foo nil nil (:generic-function))
1453 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1454 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1455 ix. (define-method-combination foo nil ((3)))
1456 x. (define-method-combination foo nil ((a)))
1458 b. define-method-combination arguments lambda list badness
1459 i. &aux args are currently unsupported;
1460 ii. default values of &optional and &key arguments are ignored;
1461 iii. supplied-p variables for &optional and &key arguments are not
1464 c. qualifier matching incorrect
1466 (define-method-combination mc27 ()
1468 (ignored (:ignore :unused)))
1470 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1471 (defgeneric test-mc27 (x)
1472 (:method-combination mc27)
1473 (:method :ignore ((x number)) (/ 0)))
1476 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1477 method is applicable, and yet matches neither of the method group
1480 341: PPRINT-LOGICAL-BLOCK / PPRINT-FILL / PPRINT-LINEAR sharing detection.
1481 (from Paul Dietz' test suite)
1483 CLHS on PPRINT-LINEAR and PPRINT-FILL (and PPRINT-TABULAR, though
1484 that's slightly different) states that these functions perform
1485 circular and shared structure detection on their object. Therefore,
1487 a.(let ((*print-circle* t))
1488 (pprint-linear *standard-output* (let ((x '(a))) (list x x))))
1489 should print "(#1=(A) #1#)"
1491 b.(let ((*print-circle* t))
1492 (pprint-linear *standard-output*
1493 (let ((x (cons nil nil))) (setf (cdr x) x) x)))
1494 should print "#1=(NIL . #1#)"
1496 (it is likely that the fault lies in PPRINT-LOGICAL-BLOCK, as
1497 suggested by the suggested implementation of PPRINT-TABULAR)
1499 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1500 Even the simplest possible overriding of
1501 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1502 as "canonical", does not work:
1503 (defclass my-generic-function (standard-generic-function) ()
1504 (:metaclass funcallable-standard-class))
1505 (defmethod compute-discriminating-function ((gf my-generic-function))
1506 (let ((dfun (call-next-method)))
1507 (lambda (&rest args)
1508 (apply dfun args))))
1510 (:generic-function-class my-generic-function))
1511 (defmethod foo (x) (+ x x))
1513 signals an error. This error is the same even if the LAMBDA is
1514 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1515 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1516 src/code/target-defstruct.lisp is broken? This seems to be broken
1517 in CMUCL 18e, so it's not caused by a recent change.
1519 344: more (?) ROOM T problems (possibly part of bug 108)
1520 In sbcl-0.8.12.51, and off and on leading up to it, the
1521 SB!VM:MEMORY-USAGE operations in ROOM T caused
1522 unhandled condition (of type SB-INT:BUG):
1523 failed AVER: "(SAP= CURRENT END)"
1524 Several clever people have taken a shot at this without fixing
1525 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1526 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1527 out how to make them work reliably with the rest of the GC.
1529 (Note: there's at least one dubious thing in room.lisp: see the
1530 comment in VALID-OBJ)
1532 346: alpha backtrace
1533 In sbcl-0.8.13, all backtraces from errors caused by internal errors
1534 on the alpha seem to have a "bogus stack frame".
1537 Structure slot setters do not preserve evaluation order:
1541 (let ((i (eval '-2))
1543 (funcall #'(setf foo-x)
1545 (aref (vector x) (incf i)))
1549 349: PPRINT-INDENT rounding implementation decisions
1550 At present, pprint-indent (and indeed the whole pretty printer)
1551 more-or-less assumes that it's using a monospace font. That's
1552 probably not too silly an assumption, but one piece of information
1553 the current implementation loses is from requests to indent by a
1554 non-integral amount. As of sbcl-0.8.15.9, the system silently
1555 truncates the indentation to an integer at the point of request, but
1556 maybe the non-integral value should be propagated through the
1557 pprinter and only truncated at output? (So that indenting by 1/2
1558 then 3/2 would indent by two spaces, not one?)
1560 352: forward-referenced-class trouble
1561 reported by Bruno Haible on sbcl-devel
1563 (setf (class-name (find-class 'a)) 'b)
1567 Expected: an instance of c, with a slot named x
1568 Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
1569 While computing the class precedence list of the class named C.
1570 The class named B is a forward referenced class.
1571 The class named B is a direct superclass of the class named C.
1573 353: debugger suboptimalities on x86
1574 On x86 backtraces for undefined functions start with a bogus stack
1575 frame, and backtraces for throws to unknown catch tags with a "no
1576 debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
1577 (used on non-x86 platforms) being a more complete solution then what
1580 More generally, the debugger internals suffer from excessive x86/non-x86
1581 conditionalization and OAOOMization: refactoring the common parts would
1584 354: XEPs in backtraces
1585 Under default compilation policy
1589 Has the XEP for TEST in the backtrace, not the TEST frame itself.
1590 (sparc and x86 at least)