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
110 #<FUNCTION "CLOSURE" {406974D5}>
111 It would be nice to make closures have a settable name slot,
112 and make things like DEFSTRUCT and FLET, which create closures,
113 set helpful values into this slot.
116 And as long as we're wishing, it would be awfully nice if INSPECT could
117 also report on closures, telling about the values of the bound variables.
120 The compiler assumes that any time a function of declared FTYPE
121 doesn't signal an error, its arguments were of the declared type.
122 E.g. compiling and loading
123 (DECLAIM (OPTIMIZE (SAFETY 3)))
124 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
126 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
128 (COND ((> (FACTORIAL X) 1.0E6)
129 (FORMAT T "too big~%"))
131 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
133 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
136 will cause the INTEGERP case to be selected, giving bogus output a la
138 This violates the "declarations are assertions" principle.
139 According to the ANSI spec, in the section "System Class FUNCTION",
140 this is a case of "lying to the compiler", but the lying is done
141 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
142 not by the unexpectedly general definition of FACTORIAL. In any case,
143 "declarations are assertions" means that lying to the compiler should
144 cause an error to be signalled, and should not cause a bogus
145 result to be returned. Thus, the compiler should not assume
146 that arbitrary functions check their argument types. (It might
147 make sense to add another flag (CHECKED?) to DEFKNOWN to
148 identify functions which *do* check their argument types.)
149 (Also, verify that the compiler handles declared function
150 return types as assertions.)
153 The definitions of SIGCONTEXT-FLOAT-REGISTER and
154 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
155 supported on FreeBSD because the floating point state is not saved,
156 but at least as of FreeBSD 4.0, the floating point state *is* saved,
157 so they could be supported after all. Very likely
158 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
161 The debugger LIST-LOCATIONS command doesn't work properly.
162 (How should it work properly?)
165 Compiling and loading
166 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
168 then requesting a BACKTRACE at the debugger prompt gives no information
169 about where in the user program the problem occurred.
171 (this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
172 while giving the backtrace the non-x86 systems complains about "unknown
173 source location: using block start", but apart from that the
174 backtrace seems reasonable. On x86 this is masked by bug 353. See
175 tests/debug.impure.lisp for a test case)
178 Using the pretty-printer from the command prompt gives funny
179 results, apparently because the pretty-printer doesn't know
180 about user's command input, including the user's carriage return
181 that the user, and therefore the pretty-printer thinks that
182 the new output block should start indented 2 or more characters
183 rightward of the correct location.
186 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
187 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
188 crashes SBCL. In general tracing anything which is used in the
189 implementation of TRACE is likely to have the same problem.
192 ANSI says in one place that type declarations can be abbreviated even
193 when the type name is not a symbol, e.g.
194 (DECLAIM ((VECTOR T) *FOOVECTOR*))
195 SBCL doesn't support this. But ANSI says in another place that this
196 isn't allowed. So it's not clear this is a bug after all. (See the
197 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
201 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
202 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
203 the mean of the distribution can be systematically O(0.1%) wrong.
204 Just increasing R-I-E-B is probably not a good solution, since
205 it would decrease efficiency more than is probably necessary. Perhaps
206 using some sort of accept/reject method would be better.
209 Internally the compiler sometimes evaluates
210 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
211 (I stumbled across this when I added an
212 (assert (not (eq type1 *wild-type*)))
213 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
214 in a type context should probably be translated to T, and so it's
215 probably wrong to ask whether it's equal to the T type and then (using
216 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
217 (I haven't tried to investigate this bug enough to guess whether
218 there might be any user-level symptoms.)
220 In fact, the type system is likely to depend on this inequality not
221 holding... * is not equivalent to T in many cases, such as
222 (VECTOR *) /= (VECTOR T).
225 The facility for dumping a running Lisp image to disk gets confused
226 when run without the PURIFY option, and creates an unnecessarily large
227 core file (apparently representing memory usage up to the previous
228 high-water mark). Moreover, when the file is loaded, it confuses the
229 GC, so that thereafter memory usage can never be reduced below that
232 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
233 The interaction between gencgc and the variables used by
234 save-lisp-and-die is still nonoptimal, though, so no respite from
238 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
239 CL), out-of-line structure slot setters are horribly inefficient
240 whenever the type of the slot is declared, because out-of-line
241 structure slot setters are implemented as closures to save space,
242 so the compiler doesn't compile the type test into code, but
243 instead just saves the type in a lexical closure and interprets it
245 To exercise the problem, compile and load
246 (cl:in-package :cl-user)
248 (bar (error "missing") :type bar))
251 (loop (setf (foo-bar *foo*) x)))
253 (defvar *bar* (make-bar))
254 (defvar *foo* (make-foo :bar *bar*))
255 (defvar *setf-foo-bar* #'(setf foo-bar))
257 (loop (funcall *setf-foo-bar* x *foo*)))
258 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
259 use BACKTRACE, to see it's spending all essentially all its time
260 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
261 One possible solution would be simply to give up on
262 representing structure slot accessors as functions, and represent
263 them as macroexpansions instead. This can be inconvenient for users,
264 but it's not clear that it's worse than trying to help by expanding
265 into a horribly inefficient implementation.
266 As a workaround for the problem, #'(SETF FOO) expressions
267 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
268 (defmacro efficient-setf-function (place-function-name)
269 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
270 ;; a workaround for the problem, encouraging the
271 ;; inline expansion of the structure accessor, so
272 ;; that the compiler can optimize its type test
273 (let ((new-value (gensym "NEW-VALUE-"))
274 (structure-value (gensym "STRUCTURE-VALUE-")))
275 `(lambda (,new-value ,structure-value)
276 (setf (,place-function-name ,structure-value)
278 ;; no problem, can just use the ordinary expansion
279 `(function (setf ,place-function-name))))
282 There's apparently a bug in CEILING optimization which caused
283 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
284 applied the patches to SBCL and they didn't seem to cause problems
285 (as reported sbcl-devel 2001-05-04). However, since the patches
286 modify nontrivial code which was apparently written incorrectly
287 the first time around, until regression tests are written I'm not
288 comfortable merging the patches in the CVS version of SBCL.
291 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
292 a clean, just-started SBCL system. And it seems to be right:
293 (ROOM T) can bring a small computer to its knees for a *long*
294 time trying to GC afterwards. Surely there's some more economical
295 way to implement (ROOM T).
297 Daniel Barlow doesn't know what fixed this, but observes that it
298 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
299 in a fresh SBCL causes
301 debugger invoked on a SB-INT:BUG in thread 5911:
302 failed AVER: "(SAP= CURRENT END)"
304 unless a GC has happened beforehand.
307 When the compiler inline expands functions, it may be that different
308 kinds of return values are generated from different code branches.
309 E.g. an inline expansion of POSITION generates integer results
310 from one branch, and NIL results from another. When that inline
311 expansion is used in a context where only one of those results
314 (aref *a1* (position x *a2*)))
315 and the compiler can't prove that the unacceptable branch is
316 never taken, then bogus type mismatch warnings can be generated.
317 If you need to suppress the type mismatch warnings, you can
318 suppress the inline expansion,
320 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
321 (aref *a1* (position x *a2*)))
322 or, sometimes, suppress them by declaring the result to be of an
325 (aref *a1* (the integer (position x *a2*))))
327 This is not a new compiler problem in 0.7.0, but the new compiler
328 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
329 more conspicuous. If you don't need performance from these functions,
330 and the bogus warnings are a nuisance for you, you can return to
331 your pre-0.7.0 state of grace with
332 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
337 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
338 the entire lexical environment at the point of MACROLET available
339 in the bodies of the macroexpander functions. In particular, it
340 allows the function bodies (which run at compile time) to try to
341 access lexical variables (which are only defined at runtime).
342 It doesn't even issue a warning, which is bad.
344 The SBCL behavior arguably conforms to the ANSI spec (since the
345 spec says that the behavior is undefined, ergo anything conforms).
346 However, it would be better to issue a compile-time error.
347 Unfortunately I (WHN) don't see any simple way to detect this
348 condition in order to issue such an error, so for the meantime
349 SBCL just does this weird broken "conforming" thing.
351 The ANSI standard says, in the definition of the special operator
353 The macro-expansion functions defined by MACROLET are defined
354 in the lexical environment in which the MACROLET form appears.
355 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
356 the local macro definitions in a MACROLET, but the consequences
357 are undefined if the local macro definitions reference any
358 local variable or function bindings that are visible in that
360 Then it seems to contradict itself by giving the example
362 (macrolet ((fudge (z)
363 ;The parameters x and flag are not accessible
364 ; at this point; a reference to flag would be to
365 ; the global variable of that name.
366 ` (if flag (* ,z ,z) ,z)))
367 ;The parameters x and flag are accessible here.
371 The comment "a reference to flag would be to the global variable
372 of the same name" sounds like good behavior for the system to have.
373 but actual specification quoted above says that the actual behavior
376 (Since 0.7.8.23 macroexpanders are defined in a restricted version
377 of the lexical environment, containing no lexical variables and
378 functions, which seems to conform to ANSI and CLtL2, but signalling
379 a STYLE-WARNING for references to variables similar to locals might
383 Ideally, uninterning a symbol would allow it, and its associated
384 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
385 at least as of sbcl-0.7.0, this isn't the case. Information about
386 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
387 essentially in ordinary (non-weak) hash tables keyed by symbols.
388 Thus, once a system has an entry in this system, it tends to live
389 forever, even when it is uninterned and all other references to it
393 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
394 prominent SourceForge web/db bug tracking system, which is
395 unfortunately not a reliable way to get a timely response from
396 the SBCL maintainers)
397 In the course of trying to build a test case for an
398 application error, I encountered this behavior:
399 If you start up sbcl, and then lay on CTRL-C for a
400 minute or two, the lisp process will eventually say:
401 %PRIMITIVE HALT called; the party is over.
402 and throw you into the monitor. If I start up lisp,
403 attach to the process with strace, and then do the same
404 (abusive) thing, I get instead:
405 access failure in heap page not marked as write-protected
406 and the monitor again. I don't know enough to have the
407 faintest idea of what is going on here.
408 This is with sbcl 6.12, uname -a reports:
409 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
410 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
411 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
412 it took more than two minutes (but less than five) for me.
416 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
417 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
418 COMPLEX implementation didn't deal with this, and hasn't been
419 upgraded to do so. (This doesn't seem to be a high priority
420 conformance problem, since seems hard to construct useful code
423 [ partially fixed by CSR in 0.8.17.17 because of a PFD ansi-tests
424 report that (COMPLEX RATIO) was failing; still failing on types of
425 the form (AND NUMBER (SATISFIES REALP) (SATISFIES ZEROP)). ]
427 b. (fixed in 0.8.3.43)
430 Floating point errors are reported poorly. E.g. on x86 OpenBSD
433 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
434 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
435 No traps are enabled? How can this be?
436 It should be possible to be much more specific (overflow, division
437 by zero, etc.) and of course the "How can this be?" should be fixable.
439 See also bugs #45.c and #183
442 (reported by Robert E. Brown 2002-04-16)
443 When a function is called with too few arguments, causing the
444 debugger to be entered, the uninitialized slots in the bad call frame
445 seem to cause GCish problems, being interpreted as tagged data even
446 though they're not. In particular, executing ROOM in the
447 debugger at that point causes AVER failures:
450 * (lisp-implementation-version)
456 failed AVER: "(SAP= CURRENT END)"
457 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
458 isn't too surprising since there are many differences in stack
459 implementation and GC conservatism between the X86 and other ports.)
461 This is probably the same bug as 216
464 The compiler sometimes tries to constant-fold expressions before
465 it checks to see whether they can be reached. This can lead to
466 bogus warnings about errors in the constant folding, e.g. in code
469 (WRITE-STRING (> X 0) "+" "0"))
470 compiled in a context where the compiler can prove that X is NIL,
471 and the compiler complains that (> X 0) causes a type error because
472 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
473 caused a full WARNING, which made the bug really annoying because then
474 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
475 code. Since then the warning has been downgraded to STYLE-WARNING,
476 so it's still a bug but at least it's a little less annoying.
478 183: "IEEE floating point issues"
479 Even where floating point handling is being dealt with relatively
480 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
481 accrued-exceptions and current-exceptions part of the fp control
482 word don't seem to bear much relation to reality. E.g. on
486 debugger invoked on condition of type DIVISION-BY-ZERO:
487 arithmetic error DIVISION-BY-ZERO signalled
488 0] (sb-vm::get-floating-point-modes)
490 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
491 :ROUNDING-MODE :NEAREST
492 :CURRENT-EXCEPTIONS NIL
493 :ACCRUED-EXCEPTIONS (:INEXACT)
496 * (sb-vm::get-floating-point-modes)
497 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
498 :ROUNDING-MODE :NEAREST
499 :CURRENT-EXCEPTIONS (:INEXACT)
500 :ACCRUED-EXCEPTIONS (:INEXACT)
503 188: "compiler performance fiasco involving type inference and UNION-TYPE"
507 (declare (optimize (safety 3)))
508 (declare (optimize (compilation-speed 2)))
509 (declare (optimize (speed 1) (debug 1) (space 1)))
511 (declare (type (integer 0) start))
512 (print (incf start 22))
513 (print (incf start 26))
514 (print (incf start 28)))
516 (declare (type (integer 0) start))
517 (print (incf start 22))
518 (print (incf start 26)))
520 (declare (type (integer 0) start))
521 (print (incf start 22))
522 (print (incf start 26))))))
524 This example could be solved with clever enough constraint
525 propagation or with SSA, but consider
530 The careful type of X is {2k} :-(. Is it really important to be
531 able to work with unions of many intervals?
533 191: "Miscellaneous PCL deficiencies"
534 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
535 a. DEFCLASS does not inform the compiler about generated
536 functions. Compiling a file with
540 (WITH-SLOTS (A-CLASS-X) A
542 results in a STYLE-WARNING:
544 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
546 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
547 Mai points out that the declamation of functions is in fact
548 incorrect in some cases (most notably for structure
549 classes). This means that at present erroneous attempts to use
550 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
551 won't get the corresponding STYLE-WARNING.
552 c. (fixed in 0.8.4.23)
554 201: "Incautious type inference from compound types"
555 a. (reported by APD sbcl-devel 2002-09-17)
557 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
559 (FORMAT NIL "~S IS ~S, Y = ~S"
566 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
570 (declare (type (array * (4 4)) x))
572 (setq x (make-array '(4 4)))
573 (adjust-array y '(3 5))
574 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
576 * (foo (make-array '(4 4) :adjustable t))
579 205: "environment issues in cross compiler"
580 (These bugs have no impact on user code, but should be fixed or
582 a. Macroexpanders introduced with MACROLET are defined in the null
584 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
585 the null lexical environment.
586 c. The cross-compiler cannot inline functions defined in a non-null
589 206: ":SB-FLUID feature broken"
590 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
591 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
594 207: "poorly distributed SXHASH results for compound data"
595 SBCL's SXHASH could probably try a little harder. ANSI: "the
596 intent is that an implementation should make a good-faith
597 effort to produce hash-codes that are well distributed
598 within the range of non-negative fixnums". But
599 (let ((hits (make-hash-table)))
602 (let* ((ij (cons i j))
603 (newlist (push ij (gethash (sxhash ij) hits))))
605 (format t "~&collision: ~S~%" newlist))))))
606 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
607 would be an obvious way of fix. Maybe it would be acceptably efficient
608 to redo MIX using a lookup into a 256-entry s-box containing
609 29-bit pseudorandom numbers?
611 211: "keywords processing"
612 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
613 number of keyword arguments.
616 (flet ((foo (&key y) (list y)))
617 (list (foo :y 1 :y 2)))
619 issues confusing message
624 ; caught STYLE-WARNING:
625 ; The variable #:G15 is defined but never used.
627 212: "Sequence functions and circular arguments"
628 COERCE, MERGE and CONCATENATE go into an infinite loop when given
629 circular arguments; it would be good for the user if they could be
630 given an error instead (ANSI 17.1.1 allows this behaviour on the part
631 of the implementation, as conforming code cannot give non-proper
632 sequences to these functions. MAP also has this problem (and
633 solution), though arguably the convenience of being able to do
634 (MAP 'LIST '+ FOO '#1=(1 . #1#))
635 might be classed as more important (though signalling an error when
636 all of the arguments are circular is probably desireable).
638 213: "Sequence functions and type checking"
639 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
640 check that it will return a sequence of the given type. Fixing
641 it along the same lines as the others (cf. work done around
642 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
643 entirely straightforward.
644 c. All of these functions will silently accept a type of the form
646 whether or not the return value is of this type. This is
647 probably permitted by ANSI (see "Exceptional Situations" under
648 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
649 know about this escape clause, so code of the form
650 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
651 can erroneously return T.
653 215: ":TEST-NOT handling by functions"
654 a. FIND and POSITION currently signal errors when given non-NIL for
655 both their :TEST and (deprecated) :TEST-NOT arguments, but by
656 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
657 means that the effect is "unpredictable but harmless". It's not
658 clear what that actually means; it may preclude conforming
659 implementations from signalling errors.
660 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
661 when conflict occurs. As a quality of implementation issue, it
662 might be preferable to treat :TEST and :TEST-NOT as being in some
663 sense the same &KEY, and effectively take the first test function in
665 c. Again, a quality of implementation issue: it would be good to issue a
666 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
667 WARNING for calls with both :TEST and :TEST-NOT; possibly this
668 latter should be WARNed about at execute-time too.
670 216: "debugger confused by frames with invalid number of arguments"
671 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
672 leaves the system confused, enough so that (QUIT) no longer works.
673 It's as though the process of working with the uninitialized slot in
674 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
675 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
677 This is probably the same bug as 162
679 217: "Bad type operations with FUNCTION types"
682 * (values-type-union (specifier-type '(function (base-char)))
683 (specifier-type '(function (integer))))
685 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
687 It causes insertion of wrong type assertions into generated
691 (let ((f (etypecase x
692 (character #'write-char)
693 (integer #'write-byte))))
696 (character (write-char x s))
697 (integer (write-byte x s)))))
699 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
701 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
702 produce invalid code, but type checking is not accurate.)
704 233: bugs in constraint propagation
706 (declaim (optimize (speed 2) (safety 3)))
708 (if (typep (prog1 x (setq x y)) 'double-float)
711 (foo 1d0 5) => segmentation violation
713 235: "type system and inline expansion"
715 (declaim (ftype (function (cons) number) acc))
716 (declaim (inline acc))
718 (the number (car c)))
721 (values (locally (declare (optimize (safety 0)))
723 (locally (declare (optimize (safety 3)))
726 (foo '(nil) '(t)) => NIL, T.
728 237: "Environment arguments to type functions"
729 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
730 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
731 argument, but they ignore it completely. This is almost
732 certainly not correct.
733 b. Also, the compiler's optimizers for TYPEP have not been informed
734 about the new argument; consequently, they will not transform
735 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
736 just as optimizeable as (TYPEP 1 'INTEGER).
738 238: "REPL compiler overenthusiasm for CLOS code"
740 * (defclass foo () ())
741 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
742 causes approximately 100 lines of code deletion notes. Some
743 discussion on this issue happened under the title 'Three "interesting"
744 bugs in PCL', resulting in a fix for this oververbosity from the
745 compiler proper; however, the problem persists in the interactor
746 because the notion of original source is not preserved: for the
747 compiler, the original source of the above expression is (DEFMETHOD
748 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
749 compiler gets its hands on the code needing compilation from the REPL,
750 it has been macroexpanded several times.
752 A symptom of the same underlying problem, reported by Tony Martinez:
754 (with-input-from-string (*query-io* " no")
756 (simple-type-error () 'error))
758 ; (SB-KERNEL:FLOAT-WAIT)
760 ; note: deleting unreachable code
761 ; compilation unit finished
764 242: "WRITE-SEQUENCE suboptimality"
765 (observed from clx performance)
766 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
767 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
768 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
769 rather than writing the sequence in one go, leading to severe
770 performance degradation.
772 243: "STYLE-WARNING overenthusiasm for unused variables"
773 (observed from clx compilation)
774 In sbcl-0.7.14, in the presence of the macros
775 (DEFMACRO FOO (X) `(BAR ,X))
776 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
777 somewhat surprising style warnings are emitted for
778 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
780 ; (LAMBDA (Y) (FOO Y))
782 ; caught STYLE-WARNING:
783 ; The variable Y is defined but never used.
785 245: bugs in disassembler
786 b. On X86 operand size prefix is not recognized.
789 (defun foo (&key (a :x))
793 does not cause a warning. (BTW: old SBCL issued a warning, but for a
794 function, which was never called!)
797 Compiler does not emit warnings for
799 a. (lambda () (svref (make-array 8 :adjustable t) 1))
802 (list (let ((y (the real x)))
803 (unless (floatp y) (error ""))
808 (declare (optimize (debug 0)))
809 (declare (type vector x))
810 (list (fill-pointer x)
814 Complex array type does not have corresponding type specifier.
816 This is a problem because the compiler emits optimization notes when
817 you use a non-simple array, and without a type specifier for hairy
818 array types, there's no good way to tell it you're doing it
819 intentionally so that it should shut up and just compile the code.
821 Another problem is confusing error message "asserted type ARRAY
822 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
823 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
825 The last problem is that when type assertions are converted to type
826 checks, types are represented with type specifiers, so we could lose
827 complex attribute. (Now this is probably not important, because
828 currently checks for complex arrays seem to be performed by
832 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
833 warning. Analogous cases with the index and length being equal and
834 greater than 0 are warned for; the problem here seems to be that the
835 type required for an array reference of this type is (INTEGER 0 (0))
836 which is canonicalized to NIL.
841 (t1 (specifier-type s)))
842 (eval `(defstruct ,s))
843 (type= t1 (specifier-type s)))
848 b. The same for CSUBTYPEP.
850 262: "yet another bug in inline expansion of local functions"
851 During inline expansion of a local function Python can try to
852 reference optimized away objects (functions, variables, CTRANs from
853 tags and blocks), which later may lead to problems. Some of the
854 cases are worked around by forbidding expansion in such cases, but
855 the better way would be to reimplement inline expansion by copying
859 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
860 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
861 looks reasonable, if not obviously correct; however, it caused the
862 PPC/Linux port to segfault during warm-init while loading
863 src/pcl/std-class.fasl. A workaround patch was made, but it would
864 be nice to understand why the first patch caused problems, and to
865 fix the cause if possible.
867 268: "wrong free declaration scope"
868 The following code must signal type error:
870 (locally (declare (optimize (safety 3)))
871 (flet ((foo (x &optional (y (car x)))
872 (declare (optimize (safety 0)))
874 (funcall (eval #'foo) 1)))
877 In the following function constraint propagator optimizes nothing:
880 (declare (integer x))
881 (declare (optimize speed))
889 Compilation of the following two forms causes "X is unbound" error:
891 (symbol-macrolet ((x pi))
892 (macrolet ((foo (y) (+ x y)))
893 (declaim (inline bar))
899 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
902 CLHS says that type declaration of a symbol macro should not affect
903 its expansion, but in SBCL it does. (If you like magic and want to
904 fix it, don't forget to change all uses of MACROEXPAND to
908 The following code (taken from CLOCC) takes a lot of time to compile:
911 (declare (type (integer 0 #.large-constant) n))
914 (fixed in 0.8.2.51, but a test case would be good)
919 (declare (optimize speed))
920 (loop for i of-type (integer 0) from 0 by 2 below 10
923 uses generic arithmetic.
925 b. (fixed in 0.8.3.6)
927 279: type propagation error -- correctly inferred type goes astray?
928 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
929 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
930 &OPTIONAL), not a (INTEGER 1 536870911)
931 is emitted when compiling this file:
932 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
933 (integer #.most-negative-fixnum 0))
938 (let* (;; Uncomment this for a type mismatch warning indicating
939 ;; that the type of (FOO X) is correctly understood.
940 #+nil (fs-foo (float-sign (foo x)))
941 ;; Uncomment this for a type mismatch warning
942 ;; indicating that the type of (ABS (FOO X)) is
943 ;; correctly understood.
944 #+nil (fs-abs-foo (float-sign (abs (foo x))))
945 ;; something wrong with this one though
946 (abs-foo (abs (foo x))))
947 (declare (type (integer 1 100) abs-foo))
952 281: COMPUTE-EFFECTIVE-METHOD error signalling.
953 (slightly obscured by a non-0 default value for
954 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
955 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
956 when it finds a method with invalid qualifiers. However, it
957 shouldn't signal errors when any such methods are not applicable to
958 the particular call being evaluated, and certainly it shouldn't when
959 simply precomputing effective methods that may never be called.
960 (setf sb-pcl::*max-emf-precompute-methods* 0)
962 (:method-combination +)
963 (:method ((x symbol)) 1)
964 (:method + ((x number)) x))
965 (foo 1) -> ERROR, but should simply return 1
967 The issue seems to be that construction of a discriminating function
968 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
970 283: Thread safety: libc functions
971 There are places that we call unsafe-for-threading libc functions
972 that we should find alternatives for, or put locks around. Known or
973 strongly suspected problems, as of 0.8.3.10: please update this
974 bug instead of creating new ones
976 localtime() - called for timezone calculations in code/time.lisp
978 284: Thread safety: special variables
979 There are lots of special variables in SBCL, and I feel sure that at
980 least some of them are indicative of potentially thread-unsafe
981 parts of the system. See doc/internals/notes/threading-specials
983 286: "recursive known functions"
984 Self-call recognition conflicts with known function
985 recognition. Currently cross compiler and target COMPILE do not
986 recognize recursion, and in target compiler it can be disabled. We
987 can always disable it for known functions with RECURSIVE attribute,
988 but there remains a possibility of a function with a
989 (tail)-recursive simplification pass and transforms/VOPs for base
992 287: PPC/Linux miscompilation or corruption in first GC
993 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
994 segmentation fault is reported at the point of first triggered GC,
995 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
996 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
997 that this merely obscures, not solves, the underlying problem; as and when
998 underlying problems are fixed, it would be worth trying again to provoke
1001 288: fundamental cross-compilation issues (from old UGLINESS file)
1002 Using host floating point numbers to represent target floating point
1003 numbers, or host characters to represent target characters, is
1004 theoretically shaky. (The characters are OK as long as the characters
1005 are in the ANSI-guaranteed character set, though, so they aren't a
1006 real problem as long as the sources don't need anything but that;
1007 the floats are a real problem.)
1009 289: "type checking and source-transforms"
1011 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1014 Our policy is to check argument types at the moment of a call. It
1015 disagrees with ANSI, which says that type assertions are put
1016 immediately onto argument expressions, but is easier to implement in
1017 IR1 and is more compatible to type inference, inline expansion,
1018 etc. IR1-transforms automatically keep this policy, but source
1019 transforms for associative functions (such as +), being applied
1020 during IR1-convertion, do not. It may be tolerable for direct calls
1021 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1023 b. Another aspect of this problem is efficiency. [x y + z +]
1024 requires less registers than [x y z + +]. This transformation is
1025 currently performed with source transforms, but it would be good to
1026 also perform it in IR1 optimization phase.
1028 290: Alpha floating point and denormalized traps
1029 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1030 hardware or kernel deficiency: the status of the enable/disable
1031 denormalized-float traps bit seems to be ambiguous; by the time we
1032 get to os_restore_fp_control after a trap, denormalized traps seem
1033 to be enabled. Since we don't want a trap every time someone uses a
1034 denormalized float, in general, we mask out that bit when we restore
1035 the control word; however, this clobbers any change the user might
1039 (reported by Adam Warner, sbcl-devel 2003-09-23)
1041 The --load toplevel argument does not perform any sanitization of its
1042 argument. As a result, files with Lisp pathname pattern characters
1043 (#\* or #\?, for instance) or quotation marks can cause the system
1044 to perform arbitrary behaviour.
1047 LOOP with non-constant arithmetic step clauses suffers from overzealous
1048 type constraint: code of the form
1049 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1050 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1051 (0))). However, an integral value of X should be legal, because
1052 successive adds of integers to double-floats produces double-floats,
1053 so none of the type restrictions in the code is violated.
1055 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1056 argument type only after having read a character. This is caused
1057 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1058 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1059 less error prone to have EXPLICIT-CHECK be a local declaration,
1060 being put into the definition, instead of an attribute being kept in
1061 a separate file; maybe also put it into SB-EXT?
1063 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1064 in NOTE-ASSUMED-TYPES
1065 In sbcl-0.8.7.32, compiling the file
1067 (declare (type integer x))
1068 (declare (type (vector (or hash-table bit)) y))
1071 (declare (type integer x))
1072 (declare (type (simple-array base (2)) y))
1075 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1077 303: "nonlinear LVARs" (aka MISC.293)
1079 (multiple-value-call #'list
1081 (multiple-value-prog1
1082 (eval '(values :a :b :c))
1088 (throw 'bar (values 3 4)))))))))))
1090 (BUU 1) returns garbage.
1092 The problem is that both EVALs sequentially write to the same LVAR.
1095 (Reported by Dave Roberts.)
1096 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1099 (declare (ftype (function () (integer 0 10)) fee)
1103 uses generic arithmetic with INLINE and fixnum without.
1105 306: "Imprecise unions of array types"
1107 (declare (optimize speed)
1108 (type (or (array cons) (array vector)) x))
1110 (foo #((0))) => TYPE-ERROR
1117 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1118 collect `(array ,(sb-vm:saetp-specifier x)))))
1119 => NIL, T (when it should be T, T)
1121 309: "Dubious values for implementation limits"
1122 (reported by Bruno Haible sbcl-devel "Incorrect value of
1123 multiple-values-limit" 2004-04-19)
1124 (values-list (make-list 1000000)), on x86/linux, signals a stack
1125 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1126 significantly larger than 1000000. There are probably similar
1127 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1128 around the same time regarding a call to LIST on sparc with 1000
1129 arguments) and other implementation limit constants.
1131 311: "Tokeniser not thread-safe"
1132 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1134 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1135 spurious errors should two threads attempt to tokenise at the same
1138 314: "LOOP :INITIALLY clauses and scope of initializers"
1139 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1140 test suite, originally by Thomas F. Burdick.
1141 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1142 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1143 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1144 ;; initially clauses are evaluated in the loop prologue, which precedes all
1145 ;; loop code except for the initial settings provided by with, for, or as.
1146 (loop :for x = 0 :then (1+ x)
1147 :for y = (1+ x) :then (ash y 1)
1148 :for z :across #(1 3 9 27 81 243)
1150 :initially (assert (zerop x)) :initially (assert (= 2 w))
1151 :until (>= w 100) :collect w)
1152 Expected: (2 6 15 38)
1155 317: "FORMAT of floating point numbers"
1156 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1158 (format nil "~1F" 10) => "0." ; "10." expected
1159 (format nil "~0F" 10) => "0." ; "10." expected
1160 (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected
1161 it would be nice if whatever fixed this also untangled the two
1162 competing implementations of floating point printing (Steele and
1163 White, and Burger and Dybvig) present in src/code/print.lisp
1165 318: "stack overflow in compiler warning with redefined class"
1166 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1169 (setf (find-class 'foo) nil)
1170 (defstruct foo slot-1)
1171 This used to give a stack overflow from within the printer, which has
1172 been fixed as of 0.8.16.11. Current result:
1174 ; can't compile TYPEP of anonymous or undefined class:
1175 ; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
1177 debugger invoked on a TYPE-ERROR in thread 19973:
1178 The value NIL is not of type FUNCTION.
1180 CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
1181 meant to be enough to delete structure classes from the system?
1183 319: "backquote with comma inside array"
1184 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1186 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1188 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1189 which probably isn't intentional.
1191 323: "REPLACE, BIT-BASH and large strings"
1192 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1193 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1195 (declare (optimize speed (safety 1)))
1196 (let ((x (make-string 140000000))
1197 (y (make-string 140000000)))
1198 (length (replace x y))))
1201 debugger invoked on a TYPE-ERROR in thread 2412:
1202 The value 1120000000 is not of type (MOD 536870911).
1203 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1205 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1206 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1207 for all positive integral <x>. At present, it only works for <x> up
1208 to about 1024 (and similarly for signed-byte), so
1209 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1210 gives an error in sbcl-0.8.10.
1212 325: "CLOSE :ABORT T on supeseding streams"
1213 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1214 file on disk, even if one existed before opening.
1216 The illegality of this is not crystal clear, as the ANSI dictionary
1217 entry for CLOSE says that when :ABORT is T superseded files are not
1218 superseded (ie. the original should be restored), whereas the OPEN
1219 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1220 implementation should not destroy the old file until the new stream
1221 is closed." -- implying that even though undesirable, early deletion
1222 is legal. Restoring the original would none the less be the polite
1225 326: "*PRINT-CIRCLE* crosstalk between streams"
1226 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1227 mixed between streams when output operations are intermingled closely
1228 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1229 test case below), so that e.g. the references #2# appears on a stream
1230 with no preceding #2= on that stream to define it (because the #2= was
1231 sent to another stream).
1232 (cl:in-package :cl-user)
1233 (defstruct foo index)
1234 (defparameter *foo* (make-foo :index 4))
1236 (defparameter *bar* (make-bar))
1237 (defparameter *tangle* (list *foo* *bar* *foo*))
1238 (defmethod print-object ((foo foo) stream)
1239 (let ((index (foo-index foo)))
1240 (format *trace-output*
1241 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1243 (format stream "[FOO ~D]" index))
1245 (let ((tsos (make-string-output-stream))
1246 (ssos (make-string-output-stream)))
1247 (let ((*print-circle* t)
1248 (*trace-output* tsos)
1249 (*standard-output* ssos))
1250 (prin1 *tangle* *standard-output*))
1251 (let ((string (get-output-stream-string ssos)))
1252 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1253 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1254 (error "oops: ~S" string)))))
1255 It might be straightforward to fix this by turning the
1256 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1257 per-stream slots, but (1) it would probably be sort of messy faking
1258 up the special variable binding semantics using UNWIND-PROTECT and
1259 (2) it might be sort of a pain to test that no other bugs had been
1262 328: "Profiling generic functions", transplanted from #241
1263 (from tonyms on #lisp IRC 2003-02-25)
1264 In sbcl-0.7.12.55, typing
1265 (defclass foo () ((bar :accessor foo-bar)))
1268 (defclass foo () ((bar :accessor foo-bar)))
1269 gives the error message
1270 "#:FOO-BAR already names an ordinary function or a macro."
1272 Problem: when a generic function is profiled, it appears as an ordinary
1273 function to PCL. (Remembering the uninterned accessor is OK, as the
1274 redefinition must be able to remove old accessors from their generic
1277 329: "Sequential class redefinition"
1278 reported by Bruno Haible:
1279 (defclass reactor () ((max-temp :initform 10000000)))
1280 (defvar *r1* (make-instance 'reactor))
1281 (defvar *r2* (make-instance 'reactor))
1282 (slot-value *r1* 'max-temp)
1283 (slot-value *r2* 'max-temp)
1284 (defclass reactor () ((uptime :initform 0)))
1285 (slot-value *r1* 'uptime)
1286 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1287 (slot-value *r1* 'max-temp) ; => 10000
1288 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1291 The method effective when the wrapper is obsoleted can be saved
1292 in the wrapper, and then to update the instance just run through
1293 all the old wrappers in order from oldest to newest.
1295 332: "fasl stack inconsistency in structure redefinition"
1296 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1297 Even though structure redefinition is undefined by the standard, the
1298 following behaviour is suboptimal: running
1299 (defun stimulate-sbcl ()
1300 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1301 ;;create a file which redefines a structure incompatibly
1302 (with-open-file (f filename :direction :output :if-exists :supersede)
1303 (print '(defstruct astruct foo) f)
1304 (print '(defstruct astruct foo bar) f))
1305 ;;compile and load the file, then invoke the continue restart on
1306 ;;the structure redefinition error
1307 (handler-bind ((error (lambda (c) (continue c))))
1308 (load (compile-file filename)))))
1310 and choosing the CONTINUE restart yields the message
1311 debugger invoked on a SB-INT:BUG in thread 27726:
1312 fasl stack not empty when it should be
1314 336: "slot-definitions must retain the generic functions of accessors"
1315 reported by Tony Martinez:
1316 (defclass foo () ((bar :reader foo-bar)))
1317 (defun foo-bar (x) x)
1318 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1320 Note: just punting the accessor removal if the fdefinition
1321 is not a generic function is not enough:
1323 (defclass foo () ((bar :reader foo-bar)))
1324 (defvar *reader* #'foo-bar)
1325 (defun foo-bar (x) x)
1326 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1327 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1328 ; the method must be removed
1329 ; by the class redefinition
1331 Fixing this should also fix a subset of #328 -- update the
1332 description with a new test-case then.
1334 337: MAKE-METHOD and user-defined method classes
1335 (reported by Bruno Haible sbcl-devel 2004-06-11)
1339 (defclass user-method (standard-method) (myslot))
1340 (defmacro def-user-method (name &rest rest)
1341 (let* ((lambdalist-position (position-if #'listp rest))
1342 (qualifiers (subseq rest 0 lambdalist-position))
1343 (lambdalist (elt rest lambdalist-position))
1344 (body (subseq rest (+ lambdalist-position 1)))
1346 (subseq lambdalist 0 (or
1348 (lambda (x) (member x lambda-list-keywords))
1350 (length lambdalist))))
1351 (specializers (mapcar #'find-class
1352 (mapcar (lambda (x) (if (consp x) (second x) t))
1354 (unspecialized-required-part
1355 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1356 (unspecialized-lambdalist
1357 (append unspecialized-required-part
1358 (subseq lambdalist (length required-part)))))
1361 (MAKE-INSTANCE 'USER-METHOD
1362 :QUALIFIERS ',qualifiers
1363 :LAMBDA-LIST ',unspecialized-lambdalist
1364 :SPECIALIZERS ',specializers
1366 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1367 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1368 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1369 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1370 (IF (NULL NEXT-METHODS-LIST)
1371 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1372 (FUNCALL (SB-PCL:METHOD-FUNCTION
1373 (FIRST NEXT-METHODS-LIST))
1374 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1375 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1379 (defgeneric test-um03 (x))
1380 (defmethod test-um03 ((x integer))
1381 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1382 (def-user-method test-um03 ((x rational))
1383 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1384 (defmethod test-um03 ((x real))
1385 (list 'real x (not (null (next-method-p)))))
1390 (defgeneric test-um10 (x))
1391 (defmethod test-um10 ((x integer))
1392 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1393 (defmethod test-um10 ((x rational))
1394 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1395 (defmethod test-um10 ((x real))
1396 (list 'real x (not (null (next-method-p)))))
1397 (defmethod test-um10 :after ((x real)))
1398 (def-user-method test-um10 :around ((x integer))
1399 (list* 'around-integer x
1400 (not (null (next-method-p))) (call-next-method)))
1401 (defmethod test-um10 :around ((x rational))
1402 (list* 'around-rational x
1403 (not (null (next-method-p))) (call-next-method)))
1404 (defmethod test-um10 :around ((x real))
1405 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1407 fails with a type error, and
1410 (defgeneric test-um12 (x))
1411 (defmethod test-um12 ((x integer))
1412 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1413 (defmethod test-um12 ((x rational))
1414 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1415 (defmethod test-um12 ((x real))
1416 (list 'real x (not (null (next-method-p)))))
1417 (defmethod test-um12 :after ((x real)))
1418 (defmethod test-um12 :around ((x integer))
1419 (list* 'around-integer x
1420 (not (null (next-method-p))) (call-next-method)))
1421 (defmethod test-um12 :around ((x rational))
1422 (list* 'around-rational x
1423 (not (null (next-method-p))) (call-next-method)))
1424 (def-user-method test-um12 :around ((x real))
1425 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1427 fails with NO-APPLICABLE-METHOD.
1429 339: "DEFINE-METHOD-COMBINATION bugs"
1430 (reported by Bruno Haible via the clisp test suite)
1432 a. Syntax checking laxity (should produce errors):
1433 i. (define-method-combination foo :documentation :operator)
1434 ii. (define-method-combination foo :documentation nil)
1435 iii. (define-method-combination foo nil)
1436 iv. (define-method-combination foo nil nil
1437 (:arguments order &aux &key))
1438 v. (define-method-combination foo nil nil (:arguments &whole))
1439 vi. (define-method-combination foo nil nil (:generic-function))
1440 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1441 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1442 ix. (define-method-combination foo nil ((3)))
1443 x. (define-method-combination foo nil ((a)))
1445 b. define-method-combination arguments lambda list badness
1446 i. &aux args are currently unsupported;
1447 ii. default values of &optional and &key arguments are ignored;
1448 iii. supplied-p variables for &optional and &key arguments are not
1451 c. qualifier matching incorrect
1453 (define-method-combination mc27 ()
1455 (ignored (:ignore :unused)))
1457 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1458 (defgeneric test-mc27 (x)
1459 (:method-combination mc27)
1460 (:method :ignore ((x number)) (/ 0)))
1463 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1464 method is applicable, and yet matches neither of the method group
1467 341: PPRINT-LOGICAL-BLOCK / PPRINT-FILL / PPRINT-LINEAR sharing detection.
1468 (from Paul Dietz' test suite)
1470 CLHS on PPRINT-LINEAR and PPRINT-FILL (and PPRINT-TABULAR, though
1471 that's slightly different) states that these functions perform
1472 circular and shared structure detection on their object. Therefore,
1474 a.(let ((*print-circle* t))
1475 (pprint-linear *standard-output* (let ((x '(a))) (list x x))))
1476 should print "(#1=(A) #1#)"
1478 b.(let ((*print-circle* t))
1479 (pprint-linear *standard-output*
1480 (let ((x (cons nil nil))) (setf (cdr x) x) x)))
1481 should print "#1=(NIL . #1#)"
1483 (it is likely that the fault lies in PPRINT-LOGICAL-BLOCK, as
1484 suggested by the suggested implementation of PPRINT-TABULAR)
1486 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1487 Even the simplest possible overriding of
1488 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1489 as "canonical", does not work:
1490 (defclass my-generic-function (standard-generic-function) ()
1491 (:metaclass funcallable-standard-class))
1492 (defmethod compute-discriminating-function ((gf my-generic-function))
1493 (let ((dfun (call-next-method)))
1494 (lambda (&rest args)
1495 (apply dfun args))))
1497 (:generic-function-class my-generic-function))
1498 (defmethod foo (x) (+ x x))
1500 signals an error. This error is the same even if the LAMBDA is
1501 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1502 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1503 src/code/target-defstruct.lisp is broken? This seems to be broken
1504 in CMUCL 18e, so it's not caused by a recent change.
1506 344: more (?) ROOM T problems (possibly part of bug 108)
1507 In sbcl-0.8.12.51, and off and on leading up to it, the
1508 SB!VM:MEMORY-USAGE operations in ROOM T caused
1509 unhandled condition (of type SB-INT:BUG):
1510 failed AVER: "(SAP= CURRENT END)"
1511 Several clever people have taken a shot at this without fixing
1512 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1513 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1514 out how to make them work reliably with the rest of the GC.
1516 (Note: there's at least one dubious thing in room.lisp: see the
1517 comment in VALID-OBJ)
1519 346: alpha backtrace
1520 In sbcl-0.8.13, all backtraces from errors caused by internal errors
1521 on the alpha seem to have a "bogus stack frame".
1523 349: PPRINT-INDENT rounding implementation decisions
1524 At present, pprint-indent (and indeed the whole pretty printer)
1525 more-or-less assumes that it's using a monospace font. That's
1526 probably not too silly an assumption, but one piece of information
1527 the current implementation loses is from requests to indent by a
1528 non-integral amount. As of sbcl-0.8.15.9, the system silently
1529 truncates the indentation to an integer at the point of request, but
1530 maybe the non-integral value should be propagated through the
1531 pprinter and only truncated at output? (So that indenting by 1/2
1532 then 3/2 would indent by two spaces, not one?)
1534 352: forward-referenced-class trouble
1535 reported by Bruno Haible on sbcl-devel
1537 (setf (class-name (find-class 'a)) 'b)
1541 Expected: an instance of c, with a slot named x
1542 Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
1543 While computing the class precedence list of the class named C.
1544 The class named B is a forward referenced class.
1545 The class named B is a direct superclass of the class named C.
1547 353: debugger suboptimalities on x86
1548 On x86 backtraces for undefined functions start with a bogus stack
1549 frame, and backtraces for throws to unknown catch tags with a "no
1550 debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
1551 (used on non-x86 platforms) being a more complete solution then what
1554 More generally, the debugger internals suffer from excessive x86/non-x86
1555 conditionalization and OAOOMization: refactoring the common parts would
1558 354: XEPs in backtraces
1559 Under default compilation policy
1563 Has the XEP for TEST in the backtrace, not the TEST frame itself.
1564 (sparc and x86 at least)
1566 355: change-class of generic-function
1567 (reported by Bruno Haible)
1568 The MOP doesn't support change-class on a generic-function. However, SBCL
1569 apparently supports it, since it doesn't give an error or warning when doing
1570 so so. Then, however, it produces wrong results for calls to this generic
1572 ;;; The effective-methods cache:
1574 (defgeneric testgf35 (x))
1575 (defmethod testgf35 ((x integer))
1576 (cons 'integer (if (next-method-p) (call-next-method))))
1577 (defmethod testgf35 ((x real))
1578 (cons 'real (if (next-method-p) (call-next-method))))
1579 (defclass customized5-generic-function (standard-generic-function)
1581 (:metaclass sb-pcl:funcallable-standard-class))
1582 (defmethod sb-pcl:compute-effective-method ((gf customized5-generic-function) method-combination methods)
1583 `(REVERSE ,(call-next-method)))
1587 (change-class #'testgf35 'customized5-generic-function)
1589 Expected: ((INTEGER REAL) (REAL INTEGER))
1590 Got: ((INTEGER REAL) (INTEGER REAL))
1591 ;;; The discriminating-function cache:
1593 (defgeneric testgf36 (x))
1594 (defmethod testgf36 ((x integer))
1595 (cons 'integer (if (next-method-p) (call-next-method))))
1596 (defmethod testgf36 ((x real))
1597 (cons 'real (if (next-method-p) (call-next-method))))
1598 (defclass customized6-generic-function (standard-generic-function)
1600 (:metaclass sb-pcl:funcallable-standard-class))
1601 (defmethod sb-pcl:compute-discriminating-function ((gf customized6-generic-function))
1602 (let ((orig-df (call-next-method)))
1603 #'(lambda (&rest arguments)
1604 (reverse (apply orig-df arguments)))))
1608 (change-class #'testgf36 'customized6-generic-function)
1610 Expected: ((INTEGER REAL) (REAL INTEGER))
1611 Got: ((INTEGER REAL) (INTEGER REAL))
1614 (reported by Bruno Haible)
1615 After the "layout depth conflict" error, the CLOS is left in a state where
1616 it's not possible to define new standard-class subclasses any more.
1618 (defclass prioritized-dispatcher ()
1619 ((dependents :type list :initform nil)))
1620 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1621 (c2 (eql (find-class 'prioritized-dispatcher))))
1623 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1625 (:metaclass sb-pcl:funcallable-standard-class))
1626 ;; ERROR, Quit the debugger with ABORT
1627 (defclass typechecking-reader-class (standard-class)
1629 Expected: #<STANDARD-CLASS TYPECHECKING-READER-CLASS>
1630 Got: ERROR "The assertion SB-PCL::WRAPPERS failed."
1632 357: defstruct inheritance of initforms
1633 (reported by Bruno Haible)
1634 When defstruct and defclass (with :metaclass structure-class) are mixed,
1635 1. some slot initforms are ignored by the DEFSTRUCT generated constructor
1637 2. all slot initforms are ignored by MAKE-INSTANCE. (This can be arguably
1638 OK for initforms that were given in a DEFSTRUCT form, but for those
1639 given in a DEFCLASS form, I think it qualifies as a bug.)
1641 (defstruct structure02a
1645 (defclass structure02b (structure02a)
1646 ((slot4 :initform -44)
1649 (slot7 :initform (floor (* pi pi)))
1650 (slot8 :initform 88))
1651 (:metaclass structure-class))
1652 (defstruct (structure02c (:include structure02b (slot8 -88)))
1655 (slot11 (floor (exp 3))))
1657 (let ((a (make-structure02c)))
1658 (list (structure02c-slot4 a)
1659 (structure02c-slot5 a)
1660 (structure02c-slot6 a)
1661 (structure02c-slot7 a)))
1662 Expected: (-44 nil t 9)
1663 Got: (SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..
1664 SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..)
1666 (let ((b (make-instance 'structure02c)))
1667 (list (structure02c-slot2 b)
1668 (structure02c-slot3 b)
1669 (structure02c-slot4 b)
1670 (structure02c-slot6 b)
1671 (structure02c-slot7 b)
1672 (structure02c-slot8 b)
1673 (structure02c-slot10 b)
1674 (structure02c-slot11 b)))
1675 Expected: (t 3 -44 t 9 -88 t 20)
1676 Got: (0 0 0 0 0 0 0 0)
1678 358: :DECLARE argument to ENSURE-GENERIC-FUNCTION
1679 (reported by Bruno Haible)
1680 According to ANSI CL, ensure-generic-function must accept a :DECLARE
1681 keyword argument. In SBCL 0.8.16 it does not.
1684 (ensure-generic-function 'foo113 :declare '((optimize (speed 3))))
1685 (sb-pcl:generic-function-declarations #'foo113))
1686 Expected: ((OPTIMIZE (SPEED 3)))
1688 Invalid initialization argument:
1690 in call for class #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>.
1692 The ANSI Standard, Section 7.1.2
1694 Bruno notes: The MOP specifies that ensure-generic-function accepts :DECLARATIONS.
1695 The easiest way to be compliant to both specs is to accept both (exclusively
1698 359: wrong default value for ensure-generic-function's :generic-function-class argument
1699 (reported by Bruno Haible)
1700 ANSI CL is silent on this, but the MOP's specification of ENSURE-GENERIC-FUNCTION says:
1701 "The remaining arguments are the complete set of keyword arguments
1702 received by ENSURE-GENERIC-FUNCTION."
1703 and the spec of ENSURE-GENERIC-FUNCTION-USING-CLASS:
1704 ":GENERIC-FUNCTION-CLASS - a class metaobject or a class name. If it is not
1705 supplied, it defaults to the class named STANDARD-GENERIC-FUNCTION."
1706 This is not the case in SBCL. Test case:
1707 (defclass my-generic-function (standard-generic-function)
1709 (:metaclass sb-pcl:funcallable-standard-class))
1710 (setf (fdefinition 'foo1)
1711 (make-instance 'my-generic-function :name 'foo1))
1712 (ensure-generic-function 'foo1
1713 :generic-function-class (find-class 'standard-generic-function))
1715 ; => #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1716 (setf (fdefinition 'foo2)
1717 (make-instance 'my-generic-function :name 'foo2))
1718 (ensure-generic-function 'foo2)
1720 Expected: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1721 Got: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS MY-GENERIC-FUNCTION>
1723 360: CALL-METHOD not recognized in method-combination body
1724 (reported by Bruno Haible)
1725 This method combination, which adds 'redo' and 'return' restarts for each
1726 method invocation to standard method combination, gives an error in SBCL.
1727 (defun prompt-for-new-values ()
1728 (format *debug-io* "~&New values: ")
1729 (list (read *debug-io*)))
1730 (defun add-method-restarts (form method)
1731 (let ((block (gensym))
1739 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Try calling ~S again." ,method))
1742 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Specify return values for ~S call." ,method))
1743 :INTERACTIVE (LAMBDA () (PROMPT-FOR-NEW-VALUES))
1744 (RETURN-FROM ,block (VALUES-LIST L)))))))))
1745 (defun convert-effective-method (efm)
1747 (if (eq (car efm) 'CALL-METHOD)
1748 (let ((method-list (third efm)))
1749 (if (or (typep (first method-list) 'method) (rest method-list))
1750 ; Reduce the case of multiple methods to a single one.
1751 ; Make the call to the next-method explicit.
1752 (convert-effective-method
1753 `(CALL-METHOD ,(second efm)
1755 (CALL-METHOD ,(first method-list) ,(rest method-list))))))
1756 ; Now the case of at most one method.
1757 (if (typep (second efm) 'method)
1758 ; Wrap the method call in a RESTART-CASE.
1759 (add-method-restarts
1760 (cons (convert-effective-method (car efm))
1761 (convert-effective-method (cdr efm)))
1763 ; Normal recursive processing.
1764 (cons (convert-effective-method (car efm))
1765 (convert-effective-method (cdr efm))))))
1766 (cons (convert-effective-method (car efm))
1767 (convert-effective-method (cdr efm))))
1769 (define-method-combination standard-with-restarts ()
1772 (primary () :required t)
1774 (flet ((call-methods-sequentially (methods)
1775 (mapcar #'(lambda (method)
1776 `(CALL-METHOD ,method))
1778 (let ((form (if (or before after (rest primary))
1779 `(MULTIPLE-VALUE-PROG1
1781 ,@(call-methods-sequentially before)
1782 (CALL-METHOD ,(first primary) ,(rest primary)))
1783 ,@(call-methods-sequentially (reverse after)))
1784 `(CALL-METHOD ,(first primary)))))
1787 `(CALL-METHOD ,(first around)
1788 (,@(rest around) (MAKE-METHOD ,form)))))
1789 (convert-effective-method form))))
1790 (defgeneric testgf16 (x) (:method-combination standard-with-restarts))
1791 (defclass testclass16a () ())
1792 (defclass testclass16b (testclass16a) ())
1793 (defclass testclass16c (testclass16a) ())
1794 (defclass testclass16d (testclass16b testclass16c) ())
1795 (defmethod testgf16 ((x testclass16a))
1797 (not (null (find-restart 'method-redo)))
1798 (not (null (find-restart 'method-return)))))
1799 (defmethod testgf16 ((x testclass16b))
1800 (cons 'b (call-next-method)))
1801 (defmethod testgf16 ((x testclass16c))
1802 (cons 'c (call-next-method)))
1803 (defmethod testgf16 ((x testclass16d))
1804 (cons 'd (call-next-method)))
1805 (testgf16 (make-instance 'testclass16d))
1807 Expected: (D B C A T T)
1808 Got: ERROR CALL-METHOD outside of a effective method form
1810 This is a bug because ANSI CL HyperSpec/Body/locmac_call-m__make-method
1812 "The macro call-method invokes the specified method, supplying it with
1813 arguments and with definitions for call-next-method and for next-method-p.
1814 If the invocation of call-method is lexically inside of a make-method,
1815 the arguments are those that were supplied to that method. Otherwise
1816 the arguments are those that were supplied to the generic function."
1817 and the example uses nothing more than these two cases (as you can see by
1818 doing (trace convert-effective-method)).
1820 361: initialize-instance of standard-reader-method ignores :function argument
1821 (reported by Bruno Haible)
1822 Pass a custom :function argument to initialize-instance of a
1823 standard-reader-method instance, but it has no effect.
1824 ;; Check that it's possible to define reader methods that do typechecking.
1826 (defclass typechecking-reader-method (sb-pcl:standard-reader-method)
1828 (defmethod initialize-instance ((method typechecking-reader-method) &rest initargs
1829 &key slot-definition)
1830 (let ((name (sb-pcl:slot-definition-name slot-definition))
1831 (type (sb-pcl:slot-definition-type slot-definition)))
1832 (apply #'call-next-method method
1833 :function #'(lambda (args next-methods)
1834 (declare (ignore next-methods))
1835 (apply #'(lambda (instance)
1836 (let ((value (slot-value instance name)))
1837 (unless (typep value type)
1838 (error "Slot ~S of ~S is not of type ~S: ~S"
1839 name instance type value))
1843 (defclass typechecking-reader-class (standard-class)
1845 (defmethod sb-pcl:validate-superclass ((c1 typechecking-reader-class) (c2 standard-class))
1847 (defmethod reader-method-class ((class typechecking-reader-class) direct-slot &rest args)
1848 (find-class 'typechecking-reader-method))
1849 (defclass testclass25 ()
1850 ((pair :type (cons symbol (cons symbol null)) :initarg :pair :accessor testclass25-pair))
1851 (:metaclass typechecking-reader-class))
1852 (macrolet ((succeeds (form)
1853 `(not (nth-value 1 (ignore-errors ,form)))))
1854 (let ((p (list 'abc 'def))
1855 (x (make-instance 'testclass25)))
1856 (list (succeeds (make-instance 'testclass25 :pair '(seventeen 17)))
1857 (succeeds (setf (testclass25-pair x) p))
1858 (succeeds (setf (second p) 456))
1859 (succeeds (testclass25-pair x))
1860 (succeeds (slot-value x 'pair))))))
1861 Expected: (t t t nil t)
1864 (inspect (first (sb-pcl:generic-function-methods #'testclass25-pair)))
1865 shows that the method was created with a FAST-FUNCTION slot but with a
1866 FUNCTION slot of NIL.
1868 362: missing error when a slot-definition is created without a name
1869 (reported by Bruno Haible)
1870 The MOP says about slot-definition initialization:
1871 "The :NAME argument is a slot name. An ERROR is SIGNALled if this argument
1872 is not a symbol which can be used as a variable name. An ERROR is SIGNALled
1873 if this argument is not supplied."
1875 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition))
1877 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION NIL>
1879 363: missing error when a slot-definition is created with a wrong documentation object
1880 (reported by Bruno Haible)
1881 The MOP says about slot-definition initialization:
1882 "The :DOCUMENTATION argument is a STRING or NIL. An ERROR is SIGNALled
1883 if it is not. This argument default to NIL during initialization."
1885 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition)
1887 :documentation 'not-a-string)
1889 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION FOO>
1891 364: does not support class objects as specializer names
1892 (reported by Bruno Haible)
1893 According to ANSI CL 7.6.2, class objects are valid specializer names,
1894 and "Parameter specializer names are used in macros intended as the
1895 user-level interface (defmethod)". DEFMETHOD's syntax section doesn't
1896 mention this possibility in the BNF for parameter-specializer-name;
1897 however, this appears to be an editorial omission, since the CLHS
1898 mentions issue CLASS-OBJECT-SPECIALIZER:AFFIRM as being approved
1899 by X3J13. SBCL doesn't support it:
1900 (defclass foo () ())
1901 (defmethod goo ((x #.(find-class 'foo))) x)
1902 Expected: #<STANDARD-METHOD GOO (#<STANDARD-CLASS FOO>)>
1903 Got: ERROR "#<STANDARD-CLASS FOO> is not a legal class name."
1905 365: mixin on generic-function subclass
1906 (reported by Bruno Haible)
1908 (defclass prioritized-dispatcher ()
1909 ((dependents :type list :initform nil)))
1910 on a generic-function subclass:
1911 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1913 (:metaclass sb-pcl:funcallable-standard-class))
1914 SBCL gives an error on this, telling to define a method on SB-MOP:VALIDATE-SUPERCLASS. If done,
1915 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1916 (c2 (eql (find-class 'prioritized-dispatcher))))
1919 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1921 (:metaclass sb-pcl:funcallable-standard-class))
1922 => debugger invoked on a SIMPLE-ERROR in thread 6687:
1923 layout depth conflict: #(#<SB-KERNEL:LAYOUT for T {500E1E9}> ...)
1925 Further discussion on this: http://thread.gmane.org/gmane.lisp.steel-bank.general/491
1927 366: cannot define two generic functions with user-defined class
1928 (reported by Bruno Haible)
1929 it is possible to define one generic function class and an instance
1930 of it. But attempting to do the same thing again, in the same session,
1931 leads to a "Control stack exhausted" error. Test case:
1932 (defclass my-generic-function-1 (standard-generic-function)
1934 (:metaclass sb-pcl:funcallable-standard-class))
1935 (defgeneric testgf-1 (x) (:generic-function-class my-generic-function-1)
1936 (:method ((x integer)) (cons 'integer nil)))
1937 (defclass my-generic-function-2 (standard-generic-function)
1939 (:metaclass sb-pcl:funcallable-standard-class))
1940 (defgeneric testgf-2 (x) (:generic-function-class my-generic-function-2)
1941 (:method ((x integer)) (cons 'integer nil)))
1942 => SB-KERNEL::CONTROL-STACK-EXHAUSTED
1944 367: TYPE-ERROR at compile time, undetected TYPE-ERROR at runtime
1946 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1950 (i367s (make-array 0 :fill-pointer t) :type (or (vector i367) null)))
1952 (g367 (error "missing :G367") :type g367 :read-only t))
1953 ;;; In sbcl-0.8.18, commenting out this (DECLAIM (FTYPE ... R367))
1954 ;;; gives an internal error at compile time:
1955 ;;; The value #<SB-KERNEL:NAMED-TYPE NIL> is not of
1956 ;;; type SB-KERNEL:VALUES-TYPE.
1957 (declaim (ftype (function ((vector i367) e367) (or s367 null)) r367))
1958 (declaim (ftype (function ((vector e367)) (values)) h367))
1960 (let ((x (g367-i367s (make-g367))))
1961 (let* ((y (or (r367 x w)
1964 (format t "~&Y=~S Z=~S~%" y z)
1966 (defun r367 (x y) (declare (ignore x y)) nil)
1967 (defun h367 (x) (declare (ignore x)) (values))
1968 ;;; In sbcl-0.8.18, executing this form causes an low-level error
1969 ;;; segmentation violation at #X9B0E1F4
1970 ;;; (instead of the TYPE-ERROR that one might like).
1971 (frob 0 (make-e367))
1972 can be made to cause two different problems, as noted in the comments:
1973 bug 367a: Compile and load the file. No TYPE-ERROR is signalled at
1974 run time (in the (S367-G367 Y) form of FROB, when Y is NIL
1975 instead of an instance of S367). Instead (on x86/Linux at least)
1976 we end up with a segfault.
1977 bug 367b: Comment out the (DECLAIM (FTYPE ... R367)), and compile
1978 the file. The compiler fails with TYPE-ERROR at compile time.
1980 368: miscompiled OR (perhaps related to bug 367)
1981 Trying to relax type declarations to find a workaround for bug 367,
1982 it turns out that even when the return type isn't declared (or
1983 declared to be T, anyway) the system remains confused about type
1984 inference in code similar to that for bug 367:
1985 (in-package :cl-user)
1986 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1990 (i368s (make-array 0 :fill-pointer t) :type (or (vector i368) null)))
1992 (g368 (error "missing :G368") :type g368 :read-only t))
1993 (declaim (ftype (function (fixnum (vector i368) e368) t) r368))
1994 (declaim (ftype (function (fixnum (vector e368)) t) h368))
1995 (defparameter *h368-was-called-p* nil)
1996 (defun nsu (vertices e368)
1997 (let ((i368s (g368-i368s (make-g368))))
1998 (let ((fuis (r368 0 i368s e368)))
1999 (format t "~&FUIS=~S~%" fuis)
2000 (or fuis (h368 0 i368s)))))
2002 (declare (ignore w x y))
2005 (declare (ignore w x))
2006 (setf *h368-was-called-p* t)
2007 (make-s368 :g368 (make-g368)))
2009 (format t "~&calling NSU~%")
2010 (let ((nsu (nsu #() (make-e368))))
2011 (format t "~&NSU returned ~S~%" nsu)
2012 (format t "~&*H368-WAS-CALLED-P*=~S~%" *h368-was-called-p*)
2013 (assert (s368-p nsu))
2014 (assert *h368-was-called-p*))
2015 In sbcl-0.8.18, both ASSERTs fail, and (DISASSEMBLE 'NSU) shows
2016 that no call to H368 is compiled.
2018 369: unlike-an-intersection behavior of VALUES-TYPE-INTERSECTION
2019 In sbcl-0.8.18.2, the identity $(x \cap y \cap y)=(x \cap y)$
2020 does not hold for VALUES-TYPE-INTERSECTION, even for types which
2021 can be intersected exactly, so that ASSERTs fail in this test case:
2022 (in-package :cl-user)
2023 (let ((types (mapcar #'sb-c::values-specifier-type
2024 '((values (vector package) &optional)
2025 (values (vector package) &rest t)
2026 (values (vector hash-table) &rest t)
2027 (values (vector hash-table) &optional)
2028 (values t &optional)
2030 (values nil &optional)
2031 (values nil &rest t)
2032 (values sequence &optional)
2033 (values sequence &rest t)
2034 (values list &optional)
2035 (values list &rest t)))))
2038 (let ((i (sb-c::values-type-intersection x y)))
2039 (assert (sb-c::type= i (sb-c::values-type-intersection i x)))
2040 (assert (sb-c::type= i (sb-c::values-type-intersection i y)))))))