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 Sometimes (SB-EXT:QUIT) fails with
89 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
90 Process inferior-lisp exited abnormally with code 1
91 I haven't noticed a repeatable case of this yet.
94 The printer doesn't report closures very well. This is true in
99 #<FUNCTION "CLOSURE" {406974D5}>
100 It would be nice to make closures have a settable name slot,
101 and make things like DEFSTRUCT and FLET, which create closures,
102 set helpful values into this slot.
105 And as long as we're wishing, it would be awfully nice if INSPECT could
106 also report on closures, telling about the values of the bound variables.
109 The compiler assumes that any time a function of declared FTYPE
110 doesn't signal an error, its arguments were of the declared type.
111 E.g. compiling and loading
112 (DECLAIM (OPTIMIZE (SAFETY 3)))
113 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
115 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
117 (COND ((> (FACTORIAL X) 1.0E6)
118 (FORMAT T "too big~%"))
120 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
122 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
125 will cause the INTEGERP case to be selected, giving bogus output a la
127 This violates the "declarations are assertions" principle.
128 According to the ANSI spec, in the section "System Class FUNCTION",
129 this is a case of "lying to the compiler", but the lying is done
130 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
131 not by the unexpectedly general definition of FACTORIAL. In any case,
132 "declarations are assertions" means that lying to the compiler should
133 cause an error to be signalled, and should not cause a bogus
134 result to be returned. Thus, the compiler should not assume
135 that arbitrary functions check their argument types. (It might
136 make sense to add another flag (CHECKED?) to DEFKNOWN to
137 identify functions which *do* check their argument types.)
138 (Also, verify that the compiler handles declared function
139 return types as assertions.)
142 The definitions of SIGCONTEXT-FLOAT-REGISTER and
143 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
144 supported on FreeBSD because the floating point state is not saved,
145 but at least as of FreeBSD 4.0, the floating point state *is* saved,
146 so they could be supported after all. Very likely
147 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
150 The debugger LIST-LOCATIONS command doesn't work properly.
151 (How should it work properly?)
154 Compiling and loading
155 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
157 then requesting a BACKTRACE at the debugger prompt gives no information
158 about where in the user program the problem occurred.
160 (this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
161 while giving the backtrace the non-x86 systems complains about "unknown
162 source location: using block start", but apart from that the
163 backtrace seems reasonable. On x86 this is masked by bug 353. See
164 tests/debug.impure.lisp for a test case)
167 Using the pretty-printer from the command prompt gives funny
168 results, apparently because the pretty-printer doesn't know
169 about user's command input, including the user's carriage return
170 that the user, and therefore the pretty-printer thinks that
171 the new output block should start indented 2 or more characters
172 rightward of the correct location.
175 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
176 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
177 crashes SBCL. In general tracing anything which is used in the
178 implementation of TRACE is likely to have the same problem.
181 ANSI says in one place that type declarations can be abbreviated even
182 when the type name is not a symbol, e.g.
183 (DECLAIM ((VECTOR T) *FOOVECTOR*))
184 SBCL doesn't support this. But ANSI says in another place that this
185 isn't allowed. So it's not clear this is a bug after all. (See the
186 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
190 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
191 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
192 the mean of the distribution can be systematically O(0.1%) wrong.
193 Just increasing R-I-E-B is probably not a good solution, since
194 it would decrease efficiency more than is probably necessary. Perhaps
195 using some sort of accept/reject method would be better.
198 Internally the compiler sometimes evaluates
199 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
200 (I stumbled across this when I added an
201 (assert (not (eq type1 *wild-type*)))
202 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
203 in a type context should probably be translated to T, and so it's
204 probably wrong to ask whether it's equal to the T type and then (using
205 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
206 (I haven't tried to investigate this bug enough to guess whether
207 there might be any user-level symptoms.)
209 In fact, the type system is likely to depend on this inequality not
210 holding... * is not equivalent to T in many cases, such as
211 (VECTOR *) /= (VECTOR T).
214 The facility for dumping a running Lisp image to disk gets confused
215 when run without the PURIFY option, and creates an unnecessarily large
216 core file (apparently representing memory usage up to the previous
217 high-water mark). Moreover, when the file is loaded, it confuses the
218 GC, so that thereafter memory usage can never be reduced below that
221 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
222 The interaction between gencgc and the variables used by
223 save-lisp-and-die is still nonoptimal, though, so no respite from
227 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
228 CL), out-of-line structure slot setters are horribly inefficient
229 whenever the type of the slot is declared, because out-of-line
230 structure slot setters are implemented as closures to save space,
231 so the compiler doesn't compile the type test into code, but
232 instead just saves the type in a lexical closure and interprets it
234 To exercise the problem, compile and load
235 (cl:in-package :cl-user)
237 (bar (error "missing") :type bar))
240 (loop (setf (foo-bar *foo*) x)))
242 (defvar *bar* (make-bar))
243 (defvar *foo* (make-foo :bar *bar*))
244 (defvar *setf-foo-bar* #'(setf foo-bar))
246 (loop (funcall *setf-foo-bar* x *foo*)))
247 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
248 use BACKTRACE, to see it's spending all essentially all its time
249 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
250 One possible solution would be simply to give up on
251 representing structure slot accessors as functions, and represent
252 them as macroexpansions instead. This can be inconvenient for users,
253 but it's not clear that it's worse than trying to help by expanding
254 into a horribly inefficient implementation.
255 As a workaround for the problem, #'(SETF FOO) expressions
256 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
257 (defmacro efficient-setf-function (place-function-name)
258 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
259 ;; a workaround for the problem, encouraging the
260 ;; inline expansion of the structure accessor, so
261 ;; that the compiler can optimize its type test
262 (let ((new-value (gensym "NEW-VALUE-"))
263 (structure-value (gensym "STRUCTURE-VALUE-")))
264 `(lambda (,new-value ,structure-value)
265 (setf (,place-function-name ,structure-value)
267 ;; no problem, can just use the ordinary expansion
268 `(function (setf ,place-function-name))))
271 There's apparently a bug in CEILING optimization which caused
272 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
273 applied the patches to SBCL and they didn't seem to cause problems
274 (as reported sbcl-devel 2001-05-04). However, since the patches
275 modify nontrivial code which was apparently written incorrectly
276 the first time around, until regression tests are written I'm not
277 comfortable merging the patches in the CVS version of SBCL.
280 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
281 a clean, just-started SBCL system. And it seems to be right:
282 (ROOM T) can bring a small computer to its knees for a *long*
283 time trying to GC afterwards. Surely there's some more economical
284 way to implement (ROOM T).
286 Daniel Barlow doesn't know what fixed this, but observes that it
287 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
288 in a fresh SBCL causes
290 debugger invoked on a SB-INT:BUG in thread 5911:
291 failed AVER: "(SAP= CURRENT END)"
293 unless a GC has happened beforehand.
296 When the compiler inline expands functions, it may be that different
297 kinds of return values are generated from different code branches.
298 E.g. an inline expansion of POSITION generates integer results
299 from one branch, and NIL results from another. When that inline
300 expansion is used in a context where only one of those results
303 (aref *a1* (position x *a2*)))
304 and the compiler can't prove that the unacceptable branch is
305 never taken, then bogus type mismatch warnings can be generated.
306 If you need to suppress the type mismatch warnings, you can
307 suppress the inline expansion,
309 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
310 (aref *a1* (position x *a2*)))
311 or, sometimes, suppress them by declaring the result to be of an
314 (aref *a1* (the integer (position x *a2*))))
316 This is not a new compiler problem in 0.7.0, but the new compiler
317 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
318 more conspicuous. If you don't need performance from these functions,
319 and the bogus warnings are a nuisance for you, you can return to
320 your pre-0.7.0 state of grace with
321 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
326 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
327 the entire lexical environment at the point of MACROLET available
328 in the bodies of the macroexpander functions. In particular, it
329 allows the function bodies (which run at compile time) to try to
330 access lexical variables (which are only defined at runtime).
331 It doesn't even issue a warning, which is bad.
333 The SBCL behavior arguably conforms to the ANSI spec (since the
334 spec says that the behavior is undefined, ergo anything conforms).
335 However, it would be better to issue a compile-time error.
336 Unfortunately I (WHN) don't see any simple way to detect this
337 condition in order to issue such an error, so for the meantime
338 SBCL just does this weird broken "conforming" thing.
340 The ANSI standard says, in the definition of the special operator
342 The macro-expansion functions defined by MACROLET are defined
343 in the lexical environment in which the MACROLET form appears.
344 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
345 the local macro definitions in a MACROLET, but the consequences
346 are undefined if the local macro definitions reference any
347 local variable or function bindings that are visible in that
349 Then it seems to contradict itself by giving the example
351 (macrolet ((fudge (z)
352 ;The parameters x and flag are not accessible
353 ; at this point; a reference to flag would be to
354 ; the global variable of that name.
355 ` (if flag (* ,z ,z) ,z)))
356 ;The parameters x and flag are accessible here.
360 The comment "a reference to flag would be to the global variable
361 of the same name" sounds like good behavior for the system to have.
362 but actual specification quoted above says that the actual behavior
365 (Since 0.7.8.23 macroexpanders are defined in a restricted version
366 of the lexical environment, containing no lexical variables and
367 functions, which seems to conform to ANSI and CLtL2, but signalling
368 a STYLE-WARNING for references to variables similar to locals might
372 Ideally, uninterning a symbol would allow it, and its associated
373 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
374 at least as of sbcl-0.7.0, this isn't the case. Information about
375 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
376 essentially in ordinary (non-weak) hash tables keyed by symbols.
377 Thus, once a system has an entry in this system, it tends to live
378 forever, even when it is uninterned and all other references to it
382 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
383 prominent SourceForge web/db bug tracking system, which is
384 unfortunately not a reliable way to get a timely response from
385 the SBCL maintainers)
386 In the course of trying to build a test case for an
387 application error, I encountered this behavior:
388 If you start up sbcl, and then lay on CTRL-C for a
389 minute or two, the lisp process will eventually say:
390 %PRIMITIVE HALT called; the party is over.
391 and throw you into the monitor. If I start up lisp,
392 attach to the process with strace, and then do the same
393 (abusive) thing, I get instead:
394 access failure in heap page not marked as write-protected
395 and the monitor again. I don't know enough to have the
396 faintest idea of what is going on here.
397 This is with sbcl 6.12, uname -a reports:
398 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
399 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
400 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
401 it took more than two minutes (but less than five) for me.
405 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
406 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
407 COMPLEX implementation didn't deal with this, and hasn't been
408 upgraded to do so. (This doesn't seem to be a high priority
409 conformance problem, since seems hard to construct useful code
412 [ partially fixed by CSR in 0.8.17.17 because of a PFD ansi-tests
413 report that (COMPLEX RATIO) was failing; still failing on types of
414 the form (AND NUMBER (SATISFIES REALP) (SATISFIES ZEROP)). ]
416 b. (fixed in 0.8.3.43)
419 Floating point errors are reported poorly. E.g. on x86 OpenBSD
422 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
423 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
424 No traps are enabled? How can this be?
425 It should be possible to be much more specific (overflow, division
426 by zero, etc.) and of course the "How can this be?" should be fixable.
428 See also bugs #45.c and #183
431 (reported by Robert E. Brown 2002-04-16)
432 When a function is called with too few arguments, causing the
433 debugger to be entered, the uninitialized slots in the bad call frame
434 seem to cause GCish problems, being interpreted as tagged data even
435 though they're not. In particular, executing ROOM in the
436 debugger at that point causes AVER failures:
439 * (lisp-implementation-version)
445 failed AVER: "(SAP= CURRENT END)"
446 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
447 isn't too surprising since there are many differences in stack
448 implementation and GC conservatism between the X86 and other ports.)
450 This is probably the same bug as 216
453 The compiler sometimes tries to constant-fold expressions before
454 it checks to see whether they can be reached. This can lead to
455 bogus warnings about errors in the constant folding, e.g. in code
458 (WRITE-STRING (> X 0) "+" "0"))
459 compiled in a context where the compiler can prove that X is NIL,
460 and the compiler complains that (> X 0) causes a type error because
461 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
462 caused a full WARNING, which made the bug really annoying because then
463 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
464 code. Since then the warning has been downgraded to STYLE-WARNING,
465 so it's still a bug but at least it's a little less annoying.
467 183: "IEEE floating point issues"
468 Even where floating point handling is being dealt with relatively
469 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
470 accrued-exceptions and current-exceptions part of the fp control
471 word don't seem to bear much relation to reality. E.g. on
475 debugger invoked on condition of type DIVISION-BY-ZERO:
476 arithmetic error DIVISION-BY-ZERO signalled
477 0] (sb-vm::get-floating-point-modes)
479 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
480 :ROUNDING-MODE :NEAREST
481 :CURRENT-EXCEPTIONS NIL
482 :ACCRUED-EXCEPTIONS (:INEXACT)
485 * (sb-vm::get-floating-point-modes)
486 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
487 :ROUNDING-MODE :NEAREST
488 :CURRENT-EXCEPTIONS (:INEXACT)
489 :ACCRUED-EXCEPTIONS (:INEXACT)
492 188: "compiler performance fiasco involving type inference and UNION-TYPE"
496 (declare (optimize (safety 3)))
497 (declare (optimize (compilation-speed 2)))
498 (declare (optimize (speed 1) (debug 1) (space 1)))
500 (declare (type (integer 0) start))
501 (print (incf start 22))
502 (print (incf start 26))
503 (print (incf start 28)))
505 (declare (type (integer 0) start))
506 (print (incf start 22))
507 (print (incf start 26)))
509 (declare (type (integer 0) start))
510 (print (incf start 22))
511 (print (incf start 26))))))
513 This example could be solved with clever enough constraint
514 propagation or with SSA, but consider
519 The careful type of X is {2k} :-(. Is it really important to be
520 able to work with unions of many intervals?
522 191: "Miscellaneous PCL deficiencies"
523 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
524 a. DEFCLASS does not inform the compiler about generated
525 functions. Compiling a file with
529 (WITH-SLOTS (A-CLASS-X) A
531 results in a STYLE-WARNING:
533 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
535 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
536 Mai points out that the declamation of functions is in fact
537 incorrect in some cases (most notably for structure
538 classes). This means that at present erroneous attempts to use
539 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
540 won't get the corresponding STYLE-WARNING.
541 c. (fixed in 0.8.4.23)
543 201: "Incautious type inference from compound types"
544 a. (reported by APD sbcl-devel 2002-09-17)
546 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
548 (FORMAT NIL "~S IS ~S, Y = ~S"
555 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
559 (declare (type (array * (4 4)) x))
561 (setq x (make-array '(4 4)))
562 (adjust-array y '(3 5))
563 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
565 * (foo (make-array '(4 4) :adjustable t))
568 205: "environment issues in cross compiler"
569 (These bugs have no impact on user code, but should be fixed or
571 a. Macroexpanders introduced with MACROLET are defined in the null
573 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
574 the null lexical environment.
575 c. The cross-compiler cannot inline functions defined in a non-null
578 206: ":SB-FLUID feature broken"
579 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
580 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
583 207: "poorly distributed SXHASH results for compound data"
584 SBCL's SXHASH could probably try a little harder. ANSI: "the
585 intent is that an implementation should make a good-faith
586 effort to produce hash-codes that are well distributed
587 within the range of non-negative fixnums". But
588 (let ((hits (make-hash-table)))
591 (let* ((ij (cons i j))
592 (newlist (push ij (gethash (sxhash ij) hits))))
594 (format t "~&collision: ~S~%" newlist))))))
595 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
596 would be an obvious way of fix. Maybe it would be acceptably efficient
597 to redo MIX using a lookup into a 256-entry s-box containing
598 29-bit pseudorandom numbers?
600 211: "keywords processing"
601 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
602 number of keyword arguments.
605 (flet ((foo (&key y) (list y)))
606 (list (foo :y 1 :y 2)))
608 issues confusing message
613 ; caught STYLE-WARNING:
614 ; The variable #:G15 is defined but never used.
616 212: "Sequence functions and circular arguments"
617 COERCE, MERGE and CONCATENATE go into an infinite loop when given
618 circular arguments; it would be good for the user if they could be
619 given an error instead (ANSI 17.1.1 allows this behaviour on the part
620 of the implementation, as conforming code cannot give non-proper
621 sequences to these functions. MAP also has this problem (and
622 solution), though arguably the convenience of being able to do
623 (MAP 'LIST '+ FOO '#1=(1 . #1#))
624 might be classed as more important (though signalling an error when
625 all of the arguments are circular is probably desireable).
627 213: "Sequence functions and type checking"
628 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
629 check that it will return a sequence of the given type. Fixing
630 it along the same lines as the others (cf. work done around
631 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
632 entirely straightforward.
633 c. All of these functions will silently accept a type of the form
635 whether or not the return value is of this type. This is
636 probably permitted by ANSI (see "Exceptional Situations" under
637 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
638 know about this escape clause, so code of the form
639 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
640 can erroneously return T.
642 215: ":TEST-NOT handling by functions"
643 a. FIND and POSITION currently signal errors when given non-NIL for
644 both their :TEST and (deprecated) :TEST-NOT arguments, but by
645 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
646 means that the effect is "unpredictable but harmless". It's not
647 clear what that actually means; it may preclude conforming
648 implementations from signalling errors.
649 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
650 when conflict occurs. As a quality of implementation issue, it
651 might be preferable to treat :TEST and :TEST-NOT as being in some
652 sense the same &KEY, and effectively take the first test function in
654 c. Again, a quality of implementation issue: it would be good to issue a
655 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
656 WARNING for calls with both :TEST and :TEST-NOT; possibly this
657 latter should be WARNed about at execute-time too.
659 216: "debugger confused by frames with invalid number of arguments"
660 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
661 leaves the system confused, enough so that (QUIT) no longer works.
662 It's as though the process of working with the uninitialized slot in
663 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
664 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
666 This is probably the same bug as 162
668 217: "Bad type operations with FUNCTION types"
671 * (values-type-union (specifier-type '(function (base-char)))
672 (specifier-type '(function (integer))))
674 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
676 It causes insertion of wrong type assertions into generated
680 (let ((f (etypecase x
681 (character #'write-char)
682 (integer #'write-byte))))
685 (character (write-char x s))
686 (integer (write-byte x s)))))
688 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
690 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
691 produce invalid code, but type checking is not accurate.)
693 233: bugs in constraint propagation
695 (declaim (optimize (speed 2) (safety 3)))
697 (if (typep (prog1 x (setq x y)) 'double-float)
700 (foo 1d0 5) => segmentation violation
702 235: "type system and inline expansion"
704 (declaim (ftype (function (cons) number) acc))
705 (declaim (inline acc))
707 (the number (car c)))
710 (values (locally (declare (optimize (safety 0)))
712 (locally (declare (optimize (safety 3)))
715 (foo '(nil) '(t)) => NIL, T.
717 237: "Environment arguments to type functions"
718 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
719 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
720 argument, but they ignore it completely. This is almost
721 certainly not correct.
722 b. Also, the compiler's optimizers for TYPEP have not been informed
723 about the new argument; consequently, they will not transform
724 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
725 just as optimizeable as (TYPEP 1 'INTEGER).
727 238: "REPL compiler overenthusiasm for CLOS code"
729 * (defclass foo () ())
730 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
731 causes approximately 100 lines of code deletion notes. Some
732 discussion on this issue happened under the title 'Three "interesting"
733 bugs in PCL', resulting in a fix for this oververbosity from the
734 compiler proper; however, the problem persists in the interactor
735 because the notion of original source is not preserved: for the
736 compiler, the original source of the above expression is (DEFMETHOD
737 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
738 compiler gets its hands on the code needing compilation from the REPL,
739 it has been macroexpanded several times.
741 A symptom of the same underlying problem, reported by Tony Martinez:
743 (with-input-from-string (*query-io* " no")
745 (simple-type-error () 'error))
747 ; (SB-KERNEL:FLOAT-WAIT)
749 ; note: deleting unreachable code
750 ; compilation unit finished
753 242: "WRITE-SEQUENCE suboptimality"
754 (observed from clx performance)
755 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
756 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
757 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
758 rather than writing the sequence in one go, leading to severe
759 performance degradation.
761 243: "STYLE-WARNING overenthusiasm for unused variables"
762 (observed from clx compilation)
763 In sbcl-0.7.14, in the presence of the macros
764 (DEFMACRO FOO (X) `(BAR ,X))
765 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
766 somewhat surprising style warnings are emitted for
767 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
769 ; (LAMBDA (Y) (FOO Y))
771 ; caught STYLE-WARNING:
772 ; The variable Y is defined but never used.
774 245: bugs in disassembler
775 b. On X86 operand size prefix is not recognized.
778 (defun foo (&key (a :x))
782 does not cause a warning. (BTW: old SBCL issued a warning, but for a
783 function, which was never called!)
786 Compiler does not emit warnings for
788 a. (lambda () (svref (make-array 8 :adjustable t) 1))
791 (list (let ((y (the real x)))
792 (unless (floatp y) (error ""))
797 (declare (optimize (debug 0)))
798 (declare (type vector x))
799 (list (fill-pointer x)
803 Complex array type does not have corresponding type specifier.
805 This is a problem because the compiler emits optimization notes when
806 you use a non-simple array, and without a type specifier for hairy
807 array types, there's no good way to tell it you're doing it
808 intentionally so that it should shut up and just compile the code.
810 Another problem is confusing error message "asserted type ARRAY
811 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
812 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
814 The last problem is that when type assertions are converted to type
815 checks, types are represented with type specifiers, so we could lose
816 complex attribute. (Now this is probably not important, because
817 currently checks for complex arrays seem to be performed by
821 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
822 warning. Analogous cases with the index and length being equal and
823 greater than 0 are warned for; the problem here seems to be that the
824 type required for an array reference of this type is (INTEGER 0 (0))
825 which is canonicalized to NIL.
830 (t1 (specifier-type s)))
831 (eval `(defstruct ,s))
832 (type= t1 (specifier-type s)))
837 b. The same for CSUBTYPEP.
839 262: "yet another bug in inline expansion of local functions"
840 During inline expansion of a local function Python can try to
841 reference optimized away objects (functions, variables, CTRANs from
842 tags and blocks), which later may lead to problems. Some of the
843 cases are worked around by forbidding expansion in such cases, but
844 the better way would be to reimplement inline expansion by copying
848 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
849 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
850 looks reasonable, if not obviously correct; however, it caused the
851 PPC/Linux port to segfault during warm-init while loading
852 src/pcl/std-class.fasl. A workaround patch was made, but it would
853 be nice to understand why the first patch caused problems, and to
854 fix the cause if possible.
856 268: "wrong free declaration scope"
857 The following code must signal type error:
859 (locally (declare (optimize (safety 3)))
860 (flet ((foo (x &optional (y (car x)))
861 (declare (optimize (safety 0)))
863 (funcall (eval #'foo) 1)))
866 In the following function constraint propagator optimizes nothing:
869 (declare (integer x))
870 (declare (optimize speed))
878 Compilation of the following two forms causes "X is unbound" error:
880 (symbol-macrolet ((x pi))
881 (macrolet ((foo (y) (+ x y)))
882 (declaim (inline bar))
888 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
891 CLHS says that type declaration of a symbol macro should not affect
892 its expansion, but in SBCL it does. (If you like magic and want to
893 fix it, don't forget to change all uses of MACROEXPAND to
897 The following code (taken from CLOCC) takes a lot of time to compile:
900 (declare (type (integer 0 #.large-constant) n))
903 (fixed in 0.8.2.51, but a test case would be good)
908 (declare (optimize speed))
909 (loop for i of-type (integer 0) from 0 by 2 below 10
912 uses generic arithmetic.
914 b. (fixed in 0.8.3.6)
916 279: type propagation error -- correctly inferred type goes astray?
917 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
918 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
919 &OPTIONAL), not a (INTEGER 1 536870911)
920 is emitted when compiling this file:
921 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
922 (integer #.most-negative-fixnum 0))
927 (let* (;; Uncomment this for a type mismatch warning indicating
928 ;; that the type of (FOO X) is correctly understood.
929 #+nil (fs-foo (float-sign (foo x)))
930 ;; Uncomment this for a type mismatch warning
931 ;; indicating that the type of (ABS (FOO X)) is
932 ;; correctly understood.
933 #+nil (fs-abs-foo (float-sign (abs (foo x))))
934 ;; something wrong with this one though
935 (abs-foo (abs (foo x))))
936 (declare (type (integer 1 100) abs-foo))
941 281: COMPUTE-EFFECTIVE-METHOD error signalling.
942 (slightly obscured by a non-0 default value for
943 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
944 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
945 when it finds a method with invalid qualifiers. However, it
946 shouldn't signal errors when any such methods are not applicable to
947 the particular call being evaluated, and certainly it shouldn't when
948 simply precomputing effective methods that may never be called.
949 (setf sb-pcl::*max-emf-precompute-methods* 0)
951 (:method-combination +)
952 (:method ((x symbol)) 1)
953 (:method + ((x number)) x))
954 (foo 1) -> ERROR, but should simply return 1
956 The issue seems to be that construction of a discriminating function
957 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
959 283: Thread safety: libc functions
960 There are places that we call unsafe-for-threading libc functions
961 that we should find alternatives for, or put locks around. Known or
962 strongly suspected problems, as of 0.8.3.10: please update this
963 bug instead of creating new ones
965 localtime() - called for timezone calculations in code/time.lisp
967 284: Thread safety: special variables
968 There are lots of special variables in SBCL, and I feel sure that at
969 least some of them are indicative of potentially thread-unsafe
970 parts of the system. See doc/internals/notes/threading-specials
972 286: "recursive known functions"
973 Self-call recognition conflicts with known function
974 recognition. Currently cross compiler and target COMPILE do not
975 recognize recursion, and in target compiler it can be disabled. We
976 can always disable it for known functions with RECURSIVE attribute,
977 but there remains a possibility of a function with a
978 (tail)-recursive simplification pass and transforms/VOPs for base
981 287: PPC/Linux miscompilation or corruption in first GC
982 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
983 segmentation fault is reported at the point of first triggered GC,
984 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
985 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
986 that this merely obscures, not solves, the underlying problem; as and when
987 underlying problems are fixed, it would be worth trying again to provoke
990 288: fundamental cross-compilation issues (from old UGLINESS file)
991 Using host floating point numbers to represent target floating point
992 numbers, or host characters to represent target characters, is
993 theoretically shaky. (The characters are OK as long as the characters
994 are in the ANSI-guaranteed character set, though, so they aren't a
995 real problem as long as the sources don't need anything but that;
996 the floats are a real problem.)
998 289: "type checking and source-transforms"
1000 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1003 Our policy is to check argument types at the moment of a call. It
1004 disagrees with ANSI, which says that type assertions are put
1005 immediately onto argument expressions, but is easier to implement in
1006 IR1 and is more compatible to type inference, inline expansion,
1007 etc. IR1-transforms automatically keep this policy, but source
1008 transforms for associative functions (such as +), being applied
1009 during IR1-convertion, do not. It may be tolerable for direct calls
1010 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1012 b. Another aspect of this problem is efficiency. [x y + z +]
1013 requires less registers than [x y z + +]. This transformation is
1014 currently performed with source transforms, but it would be good to
1015 also perform it in IR1 optimization phase.
1017 290: Alpha floating point and denormalized traps
1018 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1019 hardware or kernel deficiency: the status of the enable/disable
1020 denormalized-float traps bit seems to be ambiguous; by the time we
1021 get to os_restore_fp_control after a trap, denormalized traps seem
1022 to be enabled. Since we don't want a trap every time someone uses a
1023 denormalized float, in general, we mask out that bit when we restore
1024 the control word; however, this clobbers any change the user might
1028 (reported by Adam Warner, sbcl-devel 2003-09-23)
1030 The --load toplevel argument does not perform any sanitization of its
1031 argument. As a result, files with Lisp pathname pattern characters
1032 (#\* or #\?, for instance) or quotation marks can cause the system
1033 to perform arbitrary behaviour.
1036 LOOP with non-constant arithmetic step clauses suffers from overzealous
1037 type constraint: code of the form
1038 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1039 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1040 (0))). However, an integral value of X should be legal, because
1041 successive adds of integers to double-floats produces double-floats,
1042 so none of the type restrictions in the code is violated.
1044 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1045 argument type only after having read a character. This is caused
1046 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1047 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1048 less error prone to have EXPLICIT-CHECK be a local declaration,
1049 being put into the definition, instead of an attribute being kept in
1050 a separate file; maybe also put it into SB-EXT?
1052 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1053 in NOTE-ASSUMED-TYPES
1054 In sbcl-0.8.7.32, compiling the file
1056 (declare (type integer x))
1057 (declare (type (vector (or hash-table bit)) y))
1060 (declare (type integer x))
1061 (declare (type (simple-array base (2)) y))
1064 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1066 303: "nonlinear LVARs" (aka MISC.293)
1068 (multiple-value-call #'list
1070 (multiple-value-prog1
1071 (eval '(values :a :b :c))
1077 (throw 'bar (values 3 4)))))))))))
1079 (BUU 1) returns garbage.
1081 The problem is that both EVALs sequentially write to the same LVAR.
1084 (Reported by Dave Roberts.)
1085 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1088 (declare (ftype (function () (integer 0 10)) fee)
1092 uses generic arithmetic with INLINE and fixnum without.
1094 306: "Imprecise unions of array types"
1096 (declare (optimize speed)
1097 (type (or (array cons) (array vector)) x))
1099 (foo #((0))) => TYPE-ERROR
1106 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1107 collect `(array ,(sb-vm:saetp-specifier x)))))
1108 => NIL, T (when it should be T, T)
1110 309: "Dubious values for implementation limits"
1111 (reported by Bruno Haible sbcl-devel "Incorrect value of
1112 multiple-values-limit" 2004-04-19)
1113 (values-list (make-list 1000000)), on x86/linux, signals a stack
1114 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1115 significantly larger than 1000000. There are probably similar
1116 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1117 around the same time regarding a call to LIST on sparc with 1000
1118 arguments) and other implementation limit constants.
1120 311: "Tokeniser not thread-safe"
1121 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1123 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1124 spurious errors should two threads attempt to tokenise at the same
1127 314: "LOOP :INITIALLY clauses and scope of initializers"
1128 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1129 test suite, originally by Thomas F. Burdick.
1130 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1131 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1132 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1133 ;; initially clauses are evaluated in the loop prologue, which precedes all
1134 ;; loop code except for the initial settings provided by with, for, or as.
1135 (loop :for x = 0 :then (1+ x)
1136 :for y = (1+ x) :then (ash y 1)
1137 :for z :across #(1 3 9 27 81 243)
1139 :initially (assert (zerop x)) :initially (assert (= 2 w))
1140 :until (>= w 100) :collect w)
1141 Expected: (2 6 15 38)
1144 318: "stack overflow in compiler warning with redefined class"
1145 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1148 (setf (find-class 'foo) nil)
1149 (defstruct foo slot-1)
1150 This used to give a stack overflow from within the printer, which has
1151 been fixed as of 0.8.16.11. Current result:
1153 ; can't compile TYPEP of anonymous or undefined class:
1154 ; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
1156 debugger invoked on a TYPE-ERROR in thread 19973:
1157 The value NIL is not of type FUNCTION.
1159 CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
1160 meant to be enough to delete structure classes from the system?
1162 319: "backquote with comma inside array"
1163 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1165 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1167 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1168 which probably isn't intentional.
1170 323: "REPLACE, BIT-BASH and large strings"
1171 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1172 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1174 (declare (optimize speed (safety 1)))
1175 (let ((x (make-string 140000000))
1176 (y (make-string 140000000)))
1177 (length (replace x y))))
1180 debugger invoked on a TYPE-ERROR in thread 2412:
1181 The value 1120000000 is not of type (MOD 536870911).
1182 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1184 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1185 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1186 for all positive integral <x>. At present, it only works for <x> up
1187 to about 1024 (and similarly for signed-byte), so
1188 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1189 gives an error in sbcl-0.8.10.
1191 325: "CLOSE :ABORT T on supeseding streams"
1192 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1193 file on disk, even if one existed before opening.
1195 The illegality of this is not crystal clear, as the ANSI dictionary
1196 entry for CLOSE says that when :ABORT is T superseded files are not
1197 superseded (ie. the original should be restored), whereas the OPEN
1198 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1199 implementation should not destroy the old file until the new stream
1200 is closed." -- implying that even though undesirable, early deletion
1201 is legal. Restoring the original would none the less be the polite
1204 326: "*PRINT-CIRCLE* crosstalk between streams"
1205 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1206 mixed between streams when output operations are intermingled closely
1207 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1208 test case below), so that e.g. the references #2# appears on a stream
1209 with no preceding #2= on that stream to define it (because the #2= was
1210 sent to another stream).
1211 (cl:in-package :cl-user)
1212 (defstruct foo index)
1213 (defparameter *foo* (make-foo :index 4))
1215 (defparameter *bar* (make-bar))
1216 (defparameter *tangle* (list *foo* *bar* *foo*))
1217 (defmethod print-object ((foo foo) stream)
1218 (let ((index (foo-index foo)))
1219 (format *trace-output*
1220 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1222 (format stream "[FOO ~D]" index))
1224 (let ((tsos (make-string-output-stream))
1225 (ssos (make-string-output-stream)))
1226 (let ((*print-circle* t)
1227 (*trace-output* tsos)
1228 (*standard-output* ssos))
1229 (prin1 *tangle* *standard-output*))
1230 (let ((string (get-output-stream-string ssos)))
1231 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1232 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1233 (error "oops: ~S" string)))))
1234 It might be straightforward to fix this by turning the
1235 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1236 per-stream slots, but (1) it would probably be sort of messy faking
1237 up the special variable binding semantics using UNWIND-PROTECT and
1238 (2) it might be sort of a pain to test that no other bugs had been
1241 328: "Profiling generic functions", transplanted from #241
1242 (from tonyms on #lisp IRC 2003-02-25)
1243 In sbcl-0.7.12.55, typing
1244 (defclass foo () ((bar :accessor foo-bar)))
1247 (defclass foo () ((bar :accessor foo-bar)))
1248 gives the error message
1249 "#:FOO-BAR already names an ordinary function or a macro."
1251 Problem: when a generic function is profiled, it appears as an ordinary
1252 function to PCL. (Remembering the uninterned accessor is OK, as the
1253 redefinition must be able to remove old accessors from their generic
1256 329: "Sequential class redefinition"
1257 reported by Bruno Haible:
1258 (defclass reactor () ((max-temp :initform 10000000)))
1259 (defvar *r1* (make-instance 'reactor))
1260 (defvar *r2* (make-instance 'reactor))
1261 (slot-value *r1* 'max-temp)
1262 (slot-value *r2* 'max-temp)
1263 (defclass reactor () ((uptime :initform 0)))
1264 (slot-value *r1* 'uptime)
1265 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1266 (slot-value *r1* 'max-temp) ; => 10000
1267 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1270 The method effective when the wrapper is obsoleted can be saved
1271 in the wrapper, and then to update the instance just run through
1272 all the old wrappers in order from oldest to newest.
1274 332: "fasl stack inconsistency in structure redefinition"
1275 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1276 Even though structure redefinition is undefined by the standard, the
1277 following behaviour is suboptimal: running
1278 (defun stimulate-sbcl ()
1279 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1280 ;;create a file which redefines a structure incompatibly
1281 (with-open-file (f filename :direction :output :if-exists :supersede)
1282 (print '(defstruct astruct foo) f)
1283 (print '(defstruct astruct foo bar) f))
1284 ;;compile and load the file, then invoke the continue restart on
1285 ;;the structure redefinition error
1286 (handler-bind ((error (lambda (c) (continue c))))
1287 (load (compile-file filename)))))
1289 and choosing the CONTINUE restart yields the message
1290 debugger invoked on a SB-INT:BUG in thread 27726:
1291 fasl stack not empty when it should be
1293 336: "slot-definitions must retain the generic functions of accessors"
1294 reported by Tony Martinez:
1295 (defclass foo () ((bar :reader foo-bar)))
1296 (defun foo-bar (x) x)
1297 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1299 Note: just punting the accessor removal if the fdefinition
1300 is not a generic function is not enough:
1302 (defclass foo () ((bar :reader foo-bar)))
1303 (defvar *reader* #'foo-bar)
1304 (defun foo-bar (x) x)
1305 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1306 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1307 ; the method must be removed
1308 ; by the class redefinition
1310 Fixing this should also fix a subset of #328 -- update the
1311 description with a new test-case then.
1313 337: MAKE-METHOD and user-defined method classes
1314 (reported by Bruno Haible sbcl-devel 2004-06-11)
1318 (defclass user-method (standard-method) (myslot))
1319 (defmacro def-user-method (name &rest rest)
1320 (let* ((lambdalist-position (position-if #'listp rest))
1321 (qualifiers (subseq rest 0 lambdalist-position))
1322 (lambdalist (elt rest lambdalist-position))
1323 (body (subseq rest (+ lambdalist-position 1)))
1325 (subseq lambdalist 0 (or
1327 (lambda (x) (member x lambda-list-keywords))
1329 (length lambdalist))))
1330 (specializers (mapcar #'find-class
1331 (mapcar (lambda (x) (if (consp x) (second x) t))
1333 (unspecialized-required-part
1334 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1335 (unspecialized-lambdalist
1336 (append unspecialized-required-part
1337 (subseq lambdalist (length required-part)))))
1340 (MAKE-INSTANCE 'USER-METHOD
1341 :QUALIFIERS ',qualifiers
1342 :LAMBDA-LIST ',unspecialized-lambdalist
1343 :SPECIALIZERS ',specializers
1345 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1346 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1347 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1348 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1349 (IF (NULL NEXT-METHODS-LIST)
1350 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1351 (FUNCALL (SB-PCL:METHOD-FUNCTION
1352 (FIRST NEXT-METHODS-LIST))
1353 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1354 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1358 (defgeneric test-um03 (x))
1359 (defmethod test-um03 ((x integer))
1360 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1361 (def-user-method test-um03 ((x rational))
1362 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1363 (defmethod test-um03 ((x real))
1364 (list 'real x (not (null (next-method-p)))))
1369 (defgeneric test-um10 (x))
1370 (defmethod test-um10 ((x integer))
1371 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1372 (defmethod test-um10 ((x rational))
1373 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1374 (defmethod test-um10 ((x real))
1375 (list 'real x (not (null (next-method-p)))))
1376 (defmethod test-um10 :after ((x real)))
1377 (def-user-method test-um10 :around ((x integer))
1378 (list* 'around-integer x
1379 (not (null (next-method-p))) (call-next-method)))
1380 (defmethod test-um10 :around ((x rational))
1381 (list* 'around-rational x
1382 (not (null (next-method-p))) (call-next-method)))
1383 (defmethod test-um10 :around ((x real))
1384 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1386 fails with a type error, and
1389 (defgeneric test-um12 (x))
1390 (defmethod test-um12 ((x integer))
1391 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1392 (defmethod test-um12 ((x rational))
1393 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1394 (defmethod test-um12 ((x real))
1395 (list 'real x (not (null (next-method-p)))))
1396 (defmethod test-um12 :after ((x real)))
1397 (defmethod test-um12 :around ((x integer))
1398 (list* 'around-integer x
1399 (not (null (next-method-p))) (call-next-method)))
1400 (defmethod test-um12 :around ((x rational))
1401 (list* 'around-rational x
1402 (not (null (next-method-p))) (call-next-method)))
1403 (def-user-method test-um12 :around ((x real))
1404 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1406 fails with NO-APPLICABLE-METHOD.
1408 339: "DEFINE-METHOD-COMBINATION bugs"
1409 (reported by Bruno Haible via the clisp test suite)
1411 a. Syntax checking laxity (should produce errors):
1412 i. (define-method-combination foo :documentation :operator)
1413 ii. (define-method-combination foo :documentation nil)
1414 iii. (define-method-combination foo nil)
1415 iv. (define-method-combination foo nil nil
1416 (:arguments order &aux &key))
1417 v. (define-method-combination foo nil nil (:arguments &whole))
1418 vi. (define-method-combination foo nil nil (:generic-function))
1419 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1420 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1421 ix. (define-method-combination foo nil ((3)))
1422 x. (define-method-combination foo nil ((a)))
1424 b. define-method-combination arguments lambda list badness
1425 i. &aux args are currently unsupported;
1426 ii. default values of &optional and &key arguments are ignored;
1427 iii. supplied-p variables for &optional and &key arguments are not
1430 c. qualifier matching incorrect
1432 (define-method-combination mc27 ()
1434 (ignored (:ignore :unused)))
1436 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1437 (defgeneric test-mc27 (x)
1438 (:method-combination mc27)
1439 (:method :ignore ((x number)) (/ 0)))
1442 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1443 method is applicable, and yet matches neither of the method group
1446 341: PPRINT-LOGICAL-BLOCK / PPRINT-FILL / PPRINT-LINEAR sharing detection.
1447 (from Paul Dietz' test suite)
1449 CLHS on PPRINT-LINEAR and PPRINT-FILL (and PPRINT-TABULAR, though
1450 that's slightly different) states that these functions perform
1451 circular and shared structure detection on their object. Therefore,
1453 a.(let ((*print-circle* t))
1454 (pprint-linear *standard-output* (let ((x '(a))) (list x x))))
1455 should print "(#1=(A) #1#)"
1457 b.(let ((*print-circle* t))
1458 (pprint-linear *standard-output*
1459 (let ((x (cons nil nil))) (setf (cdr x) x) x)))
1460 should print "#1=(NIL . #1#)"
1462 (it is likely that the fault lies in PPRINT-LOGICAL-BLOCK, as
1463 suggested by the suggested implementation of PPRINT-TABULAR)
1465 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1466 Even the simplest possible overriding of
1467 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1468 as "canonical", does not work:
1469 (defclass my-generic-function (standard-generic-function) ()
1470 (:metaclass funcallable-standard-class))
1471 (defmethod compute-discriminating-function ((gf my-generic-function))
1472 (let ((dfun (call-next-method)))
1473 (lambda (&rest args)
1474 (apply dfun args))))
1476 (:generic-function-class my-generic-function))
1477 (defmethod foo (x) (+ x x))
1479 signals an error. This error is the same even if the LAMBDA is
1480 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1481 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1482 src/code/target-defstruct.lisp is broken? This seems to be broken
1483 in CMUCL 18e, so it's not caused by a recent change.
1485 344: more (?) ROOM T problems (possibly part of bug 108)
1486 In sbcl-0.8.12.51, and off and on leading up to it, the
1487 SB!VM:MEMORY-USAGE operations in ROOM T caused
1488 unhandled condition (of type SB-INT:BUG):
1489 failed AVER: "(SAP= CURRENT END)"
1490 Several clever people have taken a shot at this without fixing
1491 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1492 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1493 out how to make them work reliably with the rest of the GC.
1495 (Note: there's at least one dubious thing in room.lisp: see the
1496 comment in VALID-OBJ)
1498 346: alpha backtrace
1499 In sbcl-0.8.13, all backtraces from errors caused by internal errors
1500 on the alpha seem to have a "bogus stack frame".
1502 349: PPRINT-INDENT rounding implementation decisions
1503 At present, pprint-indent (and indeed the whole pretty printer)
1504 more-or-less assumes that it's using a monospace font. That's
1505 probably not too silly an assumption, but one piece of information
1506 the current implementation loses is from requests to indent by a
1507 non-integral amount. As of sbcl-0.8.15.9, the system silently
1508 truncates the indentation to an integer at the point of request, but
1509 maybe the non-integral value should be propagated through the
1510 pprinter and only truncated at output? (So that indenting by 1/2
1511 then 3/2 would indent by two spaces, not one?)
1513 352: forward-referenced-class trouble
1514 reported by Bruno Haible on sbcl-devel
1516 (setf (class-name (find-class 'a)) 'b)
1520 Expected: an instance of c, with a slot named x
1521 Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
1522 While computing the class precedence list of the class named C.
1523 The class named B is a forward referenced class.
1524 The class named B is a direct superclass of the class named C.
1526 353: debugger suboptimalities on x86
1527 On x86 backtraces for undefined functions start with a bogus stack
1528 frame, and backtraces for throws to unknown catch tags with a "no
1529 debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
1530 (used on non-x86 platforms) being a more complete solution then what
1533 More generally, the debugger internals suffer from excessive x86/non-x86
1534 conditionalization and OAOOMization: refactoring the common parts would
1537 354: XEPs in backtraces
1538 Under default compilation policy
1542 Has the XEP for TEST in the backtrace, not the TEST frame itself.
1543 (sparc and x86 at least)
1545 355: change-class of generic-function
1546 (reported by Bruno Haible)
1547 The MOP doesn't support change-class on a generic-function. However, SBCL
1548 apparently supports it, since it doesn't give an error or warning when doing
1549 so so. Then, however, it produces wrong results for calls to this generic
1551 ;;; The effective-methods cache:
1553 (defgeneric testgf35 (x))
1554 (defmethod testgf35 ((x integer))
1555 (cons 'integer (if (next-method-p) (call-next-method))))
1556 (defmethod testgf35 ((x real))
1557 (cons 'real (if (next-method-p) (call-next-method))))
1558 (defclass customized5-generic-function (standard-generic-function)
1560 (:metaclass sb-pcl:funcallable-standard-class))
1561 (defmethod sb-pcl:compute-effective-method ((gf customized5-generic-function) method-combination methods)
1562 `(REVERSE ,(call-next-method)))
1566 (change-class #'testgf35 'customized5-generic-function)
1568 Expected: ((INTEGER REAL) (REAL INTEGER))
1569 Got: ((INTEGER REAL) (INTEGER REAL))
1570 ;;; The discriminating-function cache:
1572 (defgeneric testgf36 (x))
1573 (defmethod testgf36 ((x integer))
1574 (cons 'integer (if (next-method-p) (call-next-method))))
1575 (defmethod testgf36 ((x real))
1576 (cons 'real (if (next-method-p) (call-next-method))))
1577 (defclass customized6-generic-function (standard-generic-function)
1579 (:metaclass sb-pcl:funcallable-standard-class))
1580 (defmethod sb-pcl:compute-discriminating-function ((gf customized6-generic-function))
1581 (let ((orig-df (call-next-method)))
1582 #'(lambda (&rest arguments)
1583 (reverse (apply orig-df arguments)))))
1587 (change-class #'testgf36 'customized6-generic-function)
1589 Expected: ((INTEGER REAL) (REAL INTEGER))
1590 Got: ((INTEGER REAL) (INTEGER REAL))
1593 (reported by Bruno Haible)
1594 After the "layout depth conflict" error, the CLOS is left in a state where
1595 it's not possible to define new standard-class subclasses any more.
1597 (defclass prioritized-dispatcher ()
1598 ((dependents :type list :initform nil)))
1599 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1600 (c2 (eql (find-class 'prioritized-dispatcher))))
1602 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1604 (:metaclass sb-pcl:funcallable-standard-class))
1605 ;; ERROR, Quit the debugger with ABORT
1606 (defclass typechecking-reader-class (standard-class)
1608 Expected: #<STANDARD-CLASS TYPECHECKING-READER-CLASS>
1609 Got: ERROR "The assertion SB-PCL::WRAPPERS failed."
1611 357: defstruct inheritance of initforms
1612 (reported by Bruno Haible)
1613 When defstruct and defclass (with :metaclass structure-class) are mixed,
1614 1. some slot initforms are ignored by the DEFSTRUCT generated constructor
1616 2. all slot initforms are ignored by MAKE-INSTANCE. (This can be arguably
1617 OK for initforms that were given in a DEFSTRUCT form, but for those
1618 given in a DEFCLASS form, I think it qualifies as a bug.)
1620 (defstruct structure02a
1624 (defclass structure02b (structure02a)
1625 ((slot4 :initform -44)
1628 (slot7 :initform (floor (* pi pi)))
1629 (slot8 :initform 88))
1630 (:metaclass structure-class))
1631 (defstruct (structure02c (:include structure02b (slot8 -88)))
1634 (slot11 (floor (exp 3))))
1636 (let ((a (make-structure02c)))
1637 (list (structure02c-slot4 a)
1638 (structure02c-slot5 a)
1639 (structure02c-slot6 a)
1640 (structure02c-slot7 a)))
1641 Expected: (-44 nil t 9)
1642 Got: (SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..
1643 SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..)
1645 (let ((b (make-instance 'structure02c)))
1646 (list (structure02c-slot2 b)
1647 (structure02c-slot3 b)
1648 (structure02c-slot4 b)
1649 (structure02c-slot6 b)
1650 (structure02c-slot7 b)
1651 (structure02c-slot8 b)
1652 (structure02c-slot10 b)
1653 (structure02c-slot11 b)))
1654 Expected: (t 3 -44 t 9 -88 t 20)
1655 Got: (0 0 0 0 0 0 0 0)
1657 358: :DECLARE argument to ENSURE-GENERIC-FUNCTION
1658 (reported by Bruno Haible)
1659 According to ANSI CL, ensure-generic-function must accept a :DECLARE
1660 keyword argument. In SBCL 0.8.16 it does not.
1663 (ensure-generic-function 'foo113 :declare '((optimize (speed 3))))
1664 (sb-pcl:generic-function-declarations #'foo113))
1665 Expected: ((OPTIMIZE (SPEED 3)))
1667 Invalid initialization argument:
1669 in call for class #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>.
1671 The ANSI Standard, Section 7.1.2
1673 Bruno notes: The MOP specifies that ensure-generic-function accepts :DECLARATIONS.
1674 The easiest way to be compliant to both specs is to accept both (exclusively
1677 359: wrong default value for ensure-generic-function's :generic-function-class argument
1678 (reported by Bruno Haible)
1679 ANSI CL is silent on this, but the MOP's specification of ENSURE-GENERIC-FUNCTION says:
1680 "The remaining arguments are the complete set of keyword arguments
1681 received by ENSURE-GENERIC-FUNCTION."
1682 and the spec of ENSURE-GENERIC-FUNCTION-USING-CLASS:
1683 ":GENERIC-FUNCTION-CLASS - a class metaobject or a class name. If it is not
1684 supplied, it defaults to the class named STANDARD-GENERIC-FUNCTION."
1685 This is not the case in SBCL. Test case:
1686 (defclass my-generic-function (standard-generic-function)
1688 (:metaclass sb-pcl:funcallable-standard-class))
1689 (setf (fdefinition 'foo1)
1690 (make-instance 'my-generic-function :name 'foo1))
1691 (ensure-generic-function 'foo1
1692 :generic-function-class (find-class 'standard-generic-function))
1694 ; => #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1695 (setf (fdefinition 'foo2)
1696 (make-instance 'my-generic-function :name 'foo2))
1697 (ensure-generic-function 'foo2)
1699 Expected: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1700 Got: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS MY-GENERIC-FUNCTION>
1702 360: CALL-METHOD not recognized in method-combination body
1703 (reported by Bruno Haible)
1704 This method combination, which adds 'redo' and 'return' restarts for each
1705 method invocation to standard method combination, gives an error in SBCL.
1706 (defun prompt-for-new-values ()
1707 (format *debug-io* "~&New values: ")
1708 (list (read *debug-io*)))
1709 (defun add-method-restarts (form method)
1710 (let ((block (gensym))
1718 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Try calling ~S again." ,method))
1721 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Specify return values for ~S call." ,method))
1722 :INTERACTIVE (LAMBDA () (PROMPT-FOR-NEW-VALUES))
1723 (RETURN-FROM ,block (VALUES-LIST L)))))))))
1724 (defun convert-effective-method (efm)
1726 (if (eq (car efm) 'CALL-METHOD)
1727 (let ((method-list (third efm)))
1728 (if (or (typep (first method-list) 'method) (rest method-list))
1729 ; Reduce the case of multiple methods to a single one.
1730 ; Make the call to the next-method explicit.
1731 (convert-effective-method
1732 `(CALL-METHOD ,(second efm)
1734 (CALL-METHOD ,(first method-list) ,(rest method-list))))))
1735 ; Now the case of at most one method.
1736 (if (typep (second efm) 'method)
1737 ; Wrap the method call in a RESTART-CASE.
1738 (add-method-restarts
1739 (cons (convert-effective-method (car efm))
1740 (convert-effective-method (cdr efm)))
1742 ; Normal recursive processing.
1743 (cons (convert-effective-method (car efm))
1744 (convert-effective-method (cdr efm))))))
1745 (cons (convert-effective-method (car efm))
1746 (convert-effective-method (cdr efm))))
1748 (define-method-combination standard-with-restarts ()
1751 (primary () :required t)
1753 (flet ((call-methods-sequentially (methods)
1754 (mapcar #'(lambda (method)
1755 `(CALL-METHOD ,method))
1757 (let ((form (if (or before after (rest primary))
1758 `(MULTIPLE-VALUE-PROG1
1760 ,@(call-methods-sequentially before)
1761 (CALL-METHOD ,(first primary) ,(rest primary)))
1762 ,@(call-methods-sequentially (reverse after)))
1763 `(CALL-METHOD ,(first primary)))))
1766 `(CALL-METHOD ,(first around)
1767 (,@(rest around) (MAKE-METHOD ,form)))))
1768 (convert-effective-method form))))
1769 (defgeneric testgf16 (x) (:method-combination standard-with-restarts))
1770 (defclass testclass16a () ())
1771 (defclass testclass16b (testclass16a) ())
1772 (defclass testclass16c (testclass16a) ())
1773 (defclass testclass16d (testclass16b testclass16c) ())
1774 (defmethod testgf16 ((x testclass16a))
1776 (not (null (find-restart 'method-redo)))
1777 (not (null (find-restart 'method-return)))))
1778 (defmethod testgf16 ((x testclass16b))
1779 (cons 'b (call-next-method)))
1780 (defmethod testgf16 ((x testclass16c))
1781 (cons 'c (call-next-method)))
1782 (defmethod testgf16 ((x testclass16d))
1783 (cons 'd (call-next-method)))
1784 (testgf16 (make-instance 'testclass16d))
1786 Expected: (D B C A T T)
1787 Got: ERROR CALL-METHOD outside of a effective method form
1789 This is a bug because ANSI CL HyperSpec/Body/locmac_call-m__make-method
1791 "The macro call-method invokes the specified method, supplying it with
1792 arguments and with definitions for call-next-method and for next-method-p.
1793 If the invocation of call-method is lexically inside of a make-method,
1794 the arguments are those that were supplied to that method. Otherwise
1795 the arguments are those that were supplied to the generic function."
1796 and the example uses nothing more than these two cases (as you can see by
1797 doing (trace convert-effective-method)).
1799 361: initialize-instance of standard-reader-method ignores :function argument
1800 (reported by Bruno Haible)
1801 Pass a custom :function argument to initialize-instance of a
1802 standard-reader-method instance, but it has no effect.
1803 ;; Check that it's possible to define reader methods that do typechecking.
1805 (defclass typechecking-reader-method (sb-pcl:standard-reader-method)
1807 (defmethod initialize-instance ((method typechecking-reader-method) &rest initargs
1808 &key slot-definition)
1809 (let ((name (sb-pcl:slot-definition-name slot-definition))
1810 (type (sb-pcl:slot-definition-type slot-definition)))
1811 (apply #'call-next-method method
1812 :function #'(lambda (args next-methods)
1813 (declare (ignore next-methods))
1814 (apply #'(lambda (instance)
1815 (let ((value (slot-value instance name)))
1816 (unless (typep value type)
1817 (error "Slot ~S of ~S is not of type ~S: ~S"
1818 name instance type value))
1822 (defclass typechecking-reader-class (standard-class)
1824 (defmethod sb-pcl:validate-superclass ((c1 typechecking-reader-class) (c2 standard-class))
1826 (defmethod reader-method-class ((class typechecking-reader-class) direct-slot &rest args)
1827 (find-class 'typechecking-reader-method))
1828 (defclass testclass25 ()
1829 ((pair :type (cons symbol (cons symbol null)) :initarg :pair :accessor testclass25-pair))
1830 (:metaclass typechecking-reader-class))
1831 (macrolet ((succeeds (form)
1832 `(not (nth-value 1 (ignore-errors ,form)))))
1833 (let ((p (list 'abc 'def))
1834 (x (make-instance 'testclass25)))
1835 (list (succeeds (make-instance 'testclass25 :pair '(seventeen 17)))
1836 (succeeds (setf (testclass25-pair x) p))
1837 (succeeds (setf (second p) 456))
1838 (succeeds (testclass25-pair x))
1839 (succeeds (slot-value x 'pair))))))
1840 Expected: (t t t nil t)
1843 (inspect (first (sb-pcl:generic-function-methods #'testclass25-pair)))
1844 shows that the method was created with a FAST-FUNCTION slot but with a
1845 FUNCTION slot of NIL.
1847 362: missing error when a slot-definition is created without a name
1848 (reported by Bruno Haible)
1849 The MOP says about slot-definition initialization:
1850 "The :NAME argument is a slot name. An ERROR is SIGNALled if this argument
1851 is not a symbol which can be used as a variable name. An ERROR is SIGNALled
1852 if this argument is not supplied."
1854 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition))
1856 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION NIL>
1858 363: missing error when a slot-definition is created with a wrong documentation object
1859 (reported by Bruno Haible)
1860 The MOP says about slot-definition initialization:
1861 "The :DOCUMENTATION argument is a STRING or NIL. An ERROR is SIGNALled
1862 if it is not. This argument default to NIL during initialization."
1864 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition)
1866 :documentation 'not-a-string)
1868 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION FOO>
1870 364: does not support class objects as specializer names
1871 (reported by Bruno Haible)
1872 According to ANSI CL 7.6.2, class objects are valid specializer names,
1873 and "Parameter specializer names are used in macros intended as the
1874 user-level interface (defmethod)". DEFMETHOD's syntax section doesn't
1875 mention this possibility in the BNF for parameter-specializer-name;
1876 however, this appears to be an editorial omission, since the CLHS
1877 mentions issue CLASS-OBJECT-SPECIALIZER:AFFIRM as being approved
1878 by X3J13. SBCL doesn't support it:
1879 (defclass foo () ())
1880 (defmethod goo ((x #.(find-class 'foo))) x)
1881 Expected: #<STANDARD-METHOD GOO (#<STANDARD-CLASS FOO>)>
1882 Got: ERROR "#<STANDARD-CLASS FOO> is not a legal class name."
1884 365: mixin on generic-function subclass
1885 (reported by Bruno Haible)
1887 (defclass prioritized-dispatcher ()
1888 ((dependents :type list :initform nil)))
1889 on a generic-function subclass:
1890 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1892 (:metaclass sb-pcl:funcallable-standard-class))
1893 SBCL gives an error on this, telling to define a method on SB-MOP:VALIDATE-SUPERCLASS. If done,
1894 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1895 (c2 (eql (find-class 'prioritized-dispatcher))))
1898 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1900 (:metaclass sb-pcl:funcallable-standard-class))
1901 => debugger invoked on a SIMPLE-ERROR in thread 6687:
1902 layout depth conflict: #(#<SB-KERNEL:LAYOUT for T {500E1E9}> ...)
1904 Further discussion on this: http://thread.gmane.org/gmane.lisp.steel-bank.general/491
1906 366: cannot define two generic functions with user-defined class
1907 (reported by Bruno Haible)
1908 it is possible to define one generic function class and an instance
1909 of it. But attempting to do the same thing again, in the same session,
1910 leads to a "Control stack exhausted" error. Test case:
1911 (defclass my-generic-function-1 (standard-generic-function)
1913 (:metaclass sb-pcl:funcallable-standard-class))
1914 (defgeneric testgf-1 (x) (:generic-function-class my-generic-function-1)
1915 (:method ((x integer)) (cons 'integer nil)))
1916 (defclass my-generic-function-2 (standard-generic-function)
1918 (:metaclass sb-pcl:funcallable-standard-class))
1919 (defgeneric testgf-2 (x) (:generic-function-class my-generic-function-2)
1920 (:method ((x integer)) (cons 'integer nil)))
1921 => SB-KERNEL::CONTROL-STACK-EXHAUSTED
1923 367: TYPE-ERROR at compile time, undetected TYPE-ERROR at runtime
1925 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1929 (i367s (make-array 0 :fill-pointer t) :type (or (vector i367) null)))
1931 (g367 (error "missing :G367") :type g367 :read-only t))
1932 ;;; In sbcl-0.8.18, commenting out this (DECLAIM (FTYPE ... R367))
1933 ;;; gives an internal error at compile time:
1934 ;;; The value #<SB-KERNEL:NAMED-TYPE NIL> is not of
1935 ;;; type SB-KERNEL:VALUES-TYPE.
1936 (declaim (ftype (function ((vector i367) e367) (or s367 null)) r367))
1937 (declaim (ftype (function ((vector e367)) (values)) h367))
1939 (let ((x (g367-i367s (make-g367))))
1940 (let* ((y (or (r367 x w)
1943 (format t "~&Y=~S Z=~S~%" y z)
1945 (defun r367 (x y) (declare (ignore x y)) nil)
1946 (defun h367 (x) (declare (ignore x)) (values))
1947 ;;; In sbcl-0.8.18, executing this form causes an low-level error
1948 ;;; segmentation violation at #X9B0E1F4
1949 ;;; (instead of the TYPE-ERROR that one might like).
1950 (frob 0 (make-e367))
1951 can be made to cause two different problems, as noted in the comments:
1952 bug 367a: Compile and load the file. No TYPE-ERROR is signalled at
1953 run time (in the (S367-G367 Y) form of FROB, when Y is NIL
1954 instead of an instance of S367). Instead (on x86/Linux at least)
1955 we end up with a segfault.
1956 bug 367b: Comment out the (DECLAIM (FTYPE ... R367)), and compile
1957 the file. The compiler fails with TYPE-ERROR at compile time.
1959 368: miscompiled OR (perhaps related to bug 367)
1960 Trying to relax type declarations to find a workaround for bug 367,
1961 it turns out that even when the return type isn't declared (or
1962 declared to be T, anyway) the system remains confused about type
1963 inference in code similar to that for bug 367:
1964 (in-package :cl-user)
1965 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1969 (i368s (make-array 0 :fill-pointer t) :type (or (vector i368) null)))
1971 (g368 (error "missing :G368") :type g368 :read-only t))
1972 (declaim (ftype (function (fixnum (vector i368) e368) t) r368))
1973 (declaim (ftype (function (fixnum (vector e368)) t) h368))
1974 (defparameter *h368-was-called-p* nil)
1975 (defun nsu (vertices e368)
1976 (let ((i368s (g368-i368s (make-g368))))
1977 (let ((fuis (r368 0 i368s e368)))
1978 (format t "~&FUIS=~S~%" fuis)
1979 (or fuis (h368 0 i368s)))))
1981 (declare (ignore w x y))
1984 (declare (ignore w x))
1985 (setf *h368-was-called-p* t)
1986 (make-s368 :g368 (make-g368)))
1988 (format t "~&calling NSU~%")
1989 (let ((nsu (nsu #() (make-e368))))
1990 (format t "~&NSU returned ~S~%" nsu)
1991 (format t "~&*H368-WAS-CALLED-P*=~S~%" *h368-was-called-p*)
1992 (assert (s368-p nsu))
1993 (assert *h368-was-called-p*))
1994 In sbcl-0.8.18, both ASSERTs fail, and (DISASSEMBLE 'NSU) shows
1995 that no call to H368 is compiled.
1997 369: unlike-an-intersection behavior of VALUES-TYPE-INTERSECTION
1998 In sbcl-0.8.18.2, the identity $(x \cap y \cap y)=(x \cap y)$
1999 does not hold for VALUES-TYPE-INTERSECTION, even for types which
2000 can be intersected exactly, so that ASSERTs fail in this test case:
2001 (in-package :cl-user)
2002 (let ((types (mapcar #'sb-c::values-specifier-type
2003 '((values (vector package) &optional)
2004 (values (vector package) &rest t)
2005 (values (vector hash-table) &rest t)
2006 (values (vector hash-table) &optional)
2007 (values t &optional)
2009 (values nil &optional)
2010 (values nil &rest t)
2011 (values sequence &optional)
2012 (values sequence &rest t)
2013 (values list &optional)
2014 (values list &rest t)))))
2017 (let ((i (sb-c::values-type-intersection x y)))
2018 (assert (sb-c::type= i (sb-c::values-type-intersection i x)))
2019 (assert (sb-c::type= i (sb-c::values-type-intersection i y)))))))
2021 370: reader misbehaviour on large-exponent floats
2022 (read-from-string "1.0s1000000000000000000000000000000000000000")
2023 causes the reader to attempt to create a very large bignum (which it
2024 will then attempt to coerce to a rational). While this isn't
2025 completely wrong, it is probably not ideal -- checking the floating
2026 point control word state and then returning the relevant float
2027 (most-positive-short-float or short-float-infinity) or signalling an
2028 error immediately would seem to make more sense.
2030 372: floating-point overflow not signalled on ppc/darwin
2031 The following assertions in float.pure.lisp fail on ppc/darwin
2032 (Mac OS X version 10.3.7):
2033 (assert (raises-error? (scale-float 1.0 most-positive-fixnum)
2034 floating-point-overflow))
2035 (assert (raises-error? (scale-float 1.0d0 (1+ most-positive-fixnum))
2036 floating-point-overflow)))
2037 as the SCALE-FLOAT just returns
2038 #.SB-EXT:SINGLE/DOUBLE-FLOAT-POSITIVE-INFINITY. These tests have been
2039 disabled on Darwin for now.
2041 373: profiling issues on ppc/darwin
2042 The following bit from smoke.impure.lisp fails on ppc/darwin:
2044 (defun profiled-fun ()
2046 (profile profiled-fun)
2047 (loop repeat 100000 do (profiled-fun))
2049 dropping into the debugger with a TYPE-ERROR:
2050 The value -1073741382 is not of type UNSIGNED-BYTE.
2051 The test has been disabled on Darwin till the bug is fixed.
2053 374: BIT-AND problem on ppc/darwin:
2054 The BIT-AND test in bit-vector.impure-cload.lisp results in
2055 fatal error encountered in SBCL pid 8356:
2056 GC invariant lost, file "gc-common.c", line 605
2057 on ppc/darwin. Test disabled for the duration.