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 And as long as we're wishing, it would be awfully nice if INSPECT could
95 also report on closures, telling about the values of the bound variables.
98 The compiler assumes that any time a function of declared FTYPE
99 doesn't signal an error, its arguments were of the declared type.
100 E.g. compiling and loading
101 (DECLAIM (OPTIMIZE (SAFETY 3)))
102 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
104 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
106 (COND ((> (FACTORIAL X) 1.0E6)
107 (FORMAT T "too big~%"))
109 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
111 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
114 will cause the INTEGERP case to be selected, giving bogus output a la
116 This violates the "declarations are assertions" principle.
117 According to the ANSI spec, in the section "System Class FUNCTION",
118 this is a case of "lying to the compiler", but the lying is done
119 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
120 not by the unexpectedly general definition of FACTORIAL. In any case,
121 "declarations are assertions" means that lying to the compiler should
122 cause an error to be signalled, and should not cause a bogus
123 result to be returned. Thus, the compiler should not assume
124 that arbitrary functions check their argument types. (It might
125 make sense to add another flag (CHECKED?) to DEFKNOWN to
126 identify functions which *do* check their argument types.)
127 (Also, verify that the compiler handles declared function
128 return types as assertions.)
131 The definitions of SIGCONTEXT-FLOAT-REGISTER and
132 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
133 supported on FreeBSD because the floating point state is not saved,
134 but at least as of FreeBSD 4.0, the floating point state *is* saved,
135 so they could be supported after all. Very likely
136 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
139 Compiling and loading
140 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
142 then requesting a BACKTRACE at the debugger prompt gives no information
143 about where in the user program the problem occurred.
145 (this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
146 while giving the backtrace the non-x86 systems complains about "unknown
147 source location: using block start", but apart from that the
148 backtrace seems reasonable. On x86 this is masked by bug 353. See
149 tests/debug.impure.lisp for a test case)
152 Using the pretty-printer from the command prompt gives funny
153 results, apparently because the pretty-printer doesn't know
154 about user's command input, including the user's carriage return
155 that the user, and therefore the pretty-printer thinks that
156 the new output block should start indented 2 or more characters
157 rightward of the correct location.
160 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
161 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
162 crashes SBCL. In general tracing anything which is used in the
163 implementation of TRACE is likely to have the same problem.
166 ANSI says in one place that type declarations can be abbreviated even
167 when the type name is not a symbol, e.g.
168 (DECLAIM ((VECTOR T) *FOOVECTOR*))
169 SBCL doesn't support this. But ANSI says in another place that this
170 isn't allowed. So it's not clear this is a bug after all. (See the
171 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
175 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
176 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
177 the mean of the distribution can be systematically O(0.1%) wrong.
178 Just increasing R-I-E-B is probably not a good solution, since
179 it would decrease efficiency more than is probably necessary. Perhaps
180 using some sort of accept/reject method would be better.
183 Internally the compiler sometimes evaluates
184 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
185 (I stumbled across this when I added an
186 (assert (not (eq type1 *wild-type*)))
187 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
188 in a type context should probably be translated to T, and so it's
189 probably wrong to ask whether it's equal to the T type and then (using
190 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
191 (I haven't tried to investigate this bug enough to guess whether
192 there might be any user-level symptoms.)
194 In fact, the type system is likely to depend on this inequality not
195 holding... * is not equivalent to T in many cases, such as
196 (VECTOR *) /= (VECTOR T).
199 The facility for dumping a running Lisp image to disk gets confused
200 when run without the PURIFY option, and creates an unnecessarily large
201 core file (apparently representing memory usage up to the previous
202 high-water mark). Moreover, when the file is loaded, it confuses the
203 GC, so that thereafter memory usage can never be reduced below that
206 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
207 The interaction between gencgc and the variables used by
208 save-lisp-and-die is still nonoptimal, though, so no respite from
212 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
213 CL), out-of-line structure slot setters are horribly inefficient
214 whenever the type of the slot is declared, because out-of-line
215 structure slot setters are implemented as closures to save space,
216 so the compiler doesn't compile the type test into code, but
217 instead just saves the type in a lexical closure and interprets it
219 To exercise the problem, compile and load
220 (cl:in-package :cl-user)
222 (bar (error "missing") :type bar))
225 (loop (setf (foo-bar *foo*) x)))
227 (defvar *bar* (make-bar))
228 (defvar *foo* (make-foo :bar *bar*))
229 (defvar *setf-foo-bar* #'(setf foo-bar))
231 (loop (funcall *setf-foo-bar* x *foo*)))
232 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
233 use BACKTRACE, to see it's spending all essentially all its time
234 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
235 One possible solution would be simply to give up on
236 representing structure slot accessors as functions, and represent
237 them as macroexpansions instead. This can be inconvenient for users,
238 but it's not clear that it's worse than trying to help by expanding
239 into a horribly inefficient implementation.
240 As a workaround for the problem, #'(SETF FOO) expressions
241 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
242 (defmacro efficient-setf-function (place-function-name)
243 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
244 ;; a workaround for the problem, encouraging the
245 ;; inline expansion of the structure accessor, so
246 ;; that the compiler can optimize its type test
247 (let ((new-value (gensym "NEW-VALUE-"))
248 (structure-value (gensym "STRUCTURE-VALUE-")))
249 `(lambda (,new-value ,structure-value)
250 (setf (,place-function-name ,structure-value)
252 ;; no problem, can just use the ordinary expansion
253 `(function (setf ,place-function-name))))
256 There's apparently a bug in CEILING optimization which caused
257 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
258 applied the patches to SBCL and they didn't seem to cause problems
259 (as reported sbcl-devel 2001-05-04). However, since the patches
260 modify nontrivial code which was apparently written incorrectly
261 the first time around, until regression tests are written I'm not
262 comfortable merging the patches in the CVS version of SBCL.
265 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
266 a clean, just-started SBCL system. And it seems to be right:
267 (ROOM T) can bring a small computer to its knees for a *long*
268 time trying to GC afterwards. Surely there's some more economical
269 way to implement (ROOM T).
271 Daniel Barlow doesn't know what fixed this, but observes that it
272 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
273 in a fresh SBCL causes
275 debugger invoked on a SB-INT:BUG in thread 5911:
276 failed AVER: "(SAP= CURRENT END)"
278 unless a GC has happened beforehand.
281 When the compiler inline expands functions, it may be that different
282 kinds of return values are generated from different code branches.
283 E.g. an inline expansion of POSITION generates integer results
284 from one branch, and NIL results from another. When that inline
285 expansion is used in a context where only one of those results
288 (aref *a1* (position x *a2*)))
289 and the compiler can't prove that the unacceptable branch is
290 never taken, then bogus type mismatch warnings can be generated.
291 If you need to suppress the type mismatch warnings, you can
292 suppress the inline expansion,
294 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
295 (aref *a1* (position x *a2*)))
296 or, sometimes, suppress them by declaring the result to be of an
299 (aref *a1* (the integer (position x *a2*))))
301 This is not a new compiler problem in 0.7.0, but the new compiler
302 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
303 more conspicuous. If you don't need performance from these functions,
304 and the bogus warnings are a nuisance for you, you can return to
305 your pre-0.7.0 state of grace with
306 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
311 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
312 the entire lexical environment at the point of MACROLET available
313 in the bodies of the macroexpander functions. In particular, it
314 allows the function bodies (which run at compile time) to try to
315 access lexical variables (which are only defined at runtime).
316 It doesn't even issue a warning, which is bad.
318 The SBCL behavior arguably conforms to the ANSI spec (since the
319 spec says that the behavior is undefined, ergo anything conforms).
320 However, it would be better to issue a compile-time error.
321 Unfortunately I (WHN) don't see any simple way to detect this
322 condition in order to issue such an error, so for the meantime
323 SBCL just does this weird broken "conforming" thing.
325 The ANSI standard says, in the definition of the special operator
327 The macro-expansion functions defined by MACROLET are defined
328 in the lexical environment in which the MACROLET form appears.
329 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
330 the local macro definitions in a MACROLET, but the consequences
331 are undefined if the local macro definitions reference any
332 local variable or function bindings that are visible in that
334 Then it seems to contradict itself by giving the example
336 (macrolet ((fudge (z)
337 ;The parameters x and flag are not accessible
338 ; at this point; a reference to flag would be to
339 ; the global variable of that name.
340 ` (if flag (* ,z ,z) ,z)))
341 ;The parameters x and flag are accessible here.
345 The comment "a reference to flag would be to the global variable
346 of the same name" sounds like good behavior for the system to have.
347 but actual specification quoted above says that the actual behavior
350 (Since 0.7.8.23 macroexpanders are defined in a restricted version
351 of the lexical environment, containing no lexical variables and
352 functions, which seems to conform to ANSI and CLtL2, but signalling
353 a STYLE-WARNING for references to variables similar to locals might
357 Ideally, uninterning a symbol would allow it, and its associated
358 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
359 at least as of sbcl-0.7.0, this isn't the case. Information about
360 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
361 essentially in ordinary (non-weak) hash tables keyed by symbols.
362 Thus, once a system has an entry in this system, it tends to live
363 forever, even when it is uninterned and all other references to it
367 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
368 prominent SourceForge web/db bug tracking system, which is
369 unfortunately not a reliable way to get a timely response from
370 the SBCL maintainers)
371 In the course of trying to build a test case for an
372 application error, I encountered this behavior:
373 If you start up sbcl, and then lay on CTRL-C for a
374 minute or two, the lisp process will eventually say:
375 %PRIMITIVE HALT called; the party is over.
376 and throw you into the monitor. If I start up lisp,
377 attach to the process with strace, and then do the same
378 (abusive) thing, I get instead:
379 access failure in heap page not marked as write-protected
380 and the monitor again. I don't know enough to have the
381 faintest idea of what is going on here.
382 This is with sbcl 6.12, uname -a reports:
383 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
384 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
385 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
386 it took more than two minutes (but less than five) for me.
390 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
391 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
392 COMPLEX implementation didn't deal with this, and hasn't been
393 upgraded to do so. (This doesn't seem to be a high priority
394 conformance problem, since seems hard to construct useful code
397 [ partially fixed by CSR in 0.8.17.17 because of a PFD ansi-tests
398 report that (COMPLEX RATIO) was failing; still failing on types of
399 the form (AND NUMBER (SATISFIES REALP) (SATISFIES ZEROP)). ]
401 b. (fixed in 0.8.3.43)
404 Floating point errors are reported poorly. E.g. on x86 OpenBSD
407 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
408 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
409 No traps are enabled? How can this be?
410 It should be possible to be much more specific (overflow, division
411 by zero, etc.) and of course the "How can this be?" should be fixable.
413 See also bugs #45.c and #183
416 (reported by Robert E. Brown 2002-04-16)
417 When a function is called with too few arguments, causing the
418 debugger to be entered, the uninitialized slots in the bad call frame
419 seem to cause GCish problems, being interpreted as tagged data even
420 though they're not. In particular, executing ROOM in the
421 debugger at that point causes AVER failures:
424 * (lisp-implementation-version)
430 failed AVER: "(SAP= CURRENT END)"
431 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
432 isn't too surprising since there are many differences in stack
433 implementation and GC conservatism between the X86 and other ports.)
435 This is probably the same bug as 216
438 The compiler sometimes tries to constant-fold expressions before
439 it checks to see whether they can be reached. This can lead to
440 bogus warnings about errors in the constant folding, e.g. in code
443 (WRITE-STRING (> X 0) "+" "0"))
444 compiled in a context where the compiler can prove that X is NIL,
445 and the compiler complains that (> X 0) causes a type error because
446 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
447 caused a full WARNING, which made the bug really annoying because then
448 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
449 code. Since then the warning has been downgraded to STYLE-WARNING,
450 so it's still a bug but at least it's a little less annoying.
452 183: "IEEE floating point issues"
453 Even where floating point handling is being dealt with relatively
454 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
455 accrued-exceptions and current-exceptions part of the fp control
456 word don't seem to bear much relation to reality. E.g. on
460 debugger invoked on condition of type DIVISION-BY-ZERO:
461 arithmetic error DIVISION-BY-ZERO signalled
462 0] (sb-vm::get-floating-point-modes)
464 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
465 :ROUNDING-MODE :NEAREST
466 :CURRENT-EXCEPTIONS NIL
467 :ACCRUED-EXCEPTIONS (:INEXACT)
470 * (sb-vm::get-floating-point-modes)
471 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
472 :ROUNDING-MODE :NEAREST
473 :CURRENT-EXCEPTIONS (:INEXACT)
474 :ACCRUED-EXCEPTIONS (:INEXACT)
477 188: "compiler performance fiasco involving type inference and UNION-TYPE"
481 (declare (optimize (safety 3)))
482 (declare (optimize (compilation-speed 2)))
483 (declare (optimize (speed 1) (debug 1) (space 1)))
485 (declare (type (integer 0) start))
486 (print (incf start 22))
487 (print (incf start 26))
488 (print (incf start 28)))
490 (declare (type (integer 0) start))
491 (print (incf start 22))
492 (print (incf start 26)))
494 (declare (type (integer 0) start))
495 (print (incf start 22))
496 (print (incf start 26))))))
498 This example could be solved with clever enough constraint
499 propagation or with SSA, but consider
504 The careful type of X is {2k} :-(. Is it really important to be
505 able to work with unions of many intervals?
507 191: "Miscellaneous PCL deficiencies"
508 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
509 a. DEFCLASS does not inform the compiler about generated
510 functions. Compiling a file with
514 (WITH-SLOTS (A-CLASS-X) A
516 results in a STYLE-WARNING:
518 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
520 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
521 Mai points out that the declamation of functions is in fact
522 incorrect in some cases (most notably for structure
523 classes). This means that at present erroneous attempts to use
524 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
525 won't get the corresponding STYLE-WARNING.
526 c. (fixed in 0.8.4.23)
528 201: "Incautious type inference from compound types"
529 a. (reported by APD sbcl-devel 2002-09-17)
531 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
533 (FORMAT NIL "~S IS ~S, Y = ~S"
540 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
544 (declare (type (array * (4 4)) x))
546 (setq x (make-array '(4 4)))
547 (adjust-array y '(3 5))
548 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
550 * (foo (make-array '(4 4) :adjustable t))
553 205: "environment issues in cross compiler"
554 (These bugs have no impact on user code, but should be fixed or
556 a. Macroexpanders introduced with MACROLET are defined in the null
558 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
559 the null lexical environment.
560 c. The cross-compiler cannot inline functions defined in a non-null
563 206: ":SB-FLUID feature broken"
564 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
565 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
568 207: "poorly distributed SXHASH results for compound data"
569 SBCL's SXHASH could probably try a little harder. ANSI: "the
570 intent is that an implementation should make a good-faith
571 effort to produce hash-codes that are well distributed
572 within the range of non-negative fixnums". But
573 (let ((hits (make-hash-table)))
576 (let* ((ij (cons i j))
577 (newlist (push ij (gethash (sxhash ij) hits))))
579 (format t "~&collision: ~S~%" newlist))))))
580 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
581 would be an obvious way of fix. Maybe it would be acceptably efficient
582 to redo MIX using a lookup into a 256-entry s-box containing
583 29-bit pseudorandom numbers?
585 211: "keywords processing"
586 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
587 number of keyword arguments.
589 212: "Sequence functions and circular arguments"
590 COERCE, MERGE and CONCATENATE go into an infinite loop when given
591 circular arguments; it would be good for the user if they could be
592 given an error instead (ANSI 17.1.1 allows this behaviour on the part
593 of the implementation, as conforming code cannot give non-proper
594 sequences to these functions. MAP also has this problem (and
595 solution), though arguably the convenience of being able to do
596 (MAP 'LIST '+ FOO '#1=(1 . #1#))
597 might be classed as more important (though signalling an error when
598 all of the arguments are circular is probably desireable).
600 213: "Sequence functions and type checking"
601 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
602 check that it will return a sequence of the given type. Fixing
603 it along the same lines as the others (cf. work done around
604 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
605 entirely straightforward.
606 c. All of these functions will silently accept a type of the form
608 whether or not the return value is of this type. This is
609 probably permitted by ANSI (see "Exceptional Situations" under
610 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
611 know about this escape clause, so code of the form
612 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
613 can erroneously return T.
615 215: ":TEST-NOT handling by functions"
616 a. FIND and POSITION currently signal errors when given non-NIL for
617 both their :TEST and (deprecated) :TEST-NOT arguments, but by
618 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
619 means that the effect is "unpredictable but harmless". It's not
620 clear what that actually means; it may preclude conforming
621 implementations from signalling errors.
622 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
623 when conflict occurs. As a quality of implementation issue, it
624 might be preferable to treat :TEST and :TEST-NOT as being in some
625 sense the same &KEY, and effectively take the first test function in
627 c. Again, a quality of implementation issue: it would be good to issue a
628 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
629 WARNING for calls with both :TEST and :TEST-NOT; possibly this
630 latter should be WARNed about at execute-time too.
632 216: "debugger confused by frames with invalid number of arguments"
633 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
634 leaves the system confused, enough so that (QUIT) no longer works.
635 It's as though the process of working with the uninitialized slot in
636 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
637 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
639 This is probably the same bug as 162
641 217: "Bad type operations with FUNCTION types"
644 * (values-type-union (specifier-type '(function (base-char)))
645 (specifier-type '(function (integer))))
647 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
649 It causes insertion of wrong type assertions into generated
653 (let ((f (etypecase x
654 (character #'write-char)
655 (integer #'write-byte))))
658 (character (write-char x s))
659 (integer (write-byte x s)))))
661 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
663 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
664 produce invalid code, but type checking is not accurate.)
666 233: bugs in constraint propagation
668 (declaim (optimize (speed 2) (safety 3)))
670 (if (typep (prog1 x (setq x y)) 'double-float)
673 (foo 1d0 5) => segmentation violation
675 235: "type system and inline expansion"
677 (declaim (ftype (function (cons) number) acc))
678 (declaim (inline acc))
680 (the number (car c)))
683 (values (locally (declare (optimize (safety 0)))
685 (locally (declare (optimize (safety 3)))
688 (foo '(nil) '(t)) => NIL, T.
690 237: "Environment arguments to type functions"
691 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
692 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
693 argument, but they ignore it completely. This is almost
694 certainly not correct.
695 b. Also, the compiler's optimizers for TYPEP have not been informed
696 about the new argument; consequently, they will not transform
697 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
698 just as optimizeable as (TYPEP 1 'INTEGER).
700 238: "REPL compiler overenthusiasm for CLOS code"
702 * (defclass foo () ())
703 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
704 causes approximately 100 lines of code deletion notes. Some
705 discussion on this issue happened under the title 'Three "interesting"
706 bugs in PCL', resulting in a fix for this oververbosity from the
707 compiler proper; however, the problem persists in the interactor
708 because the notion of original source is not preserved: for the
709 compiler, the original source of the above expression is (DEFMETHOD
710 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
711 compiler gets its hands on the code needing compilation from the REPL,
712 it has been macroexpanded several times.
714 A symptom of the same underlying problem, reported by Tony Martinez:
716 (with-input-from-string (*query-io* " no")
718 (simple-type-error () 'error))
720 ; (SB-KERNEL:FLOAT-WAIT)
722 ; note: deleting unreachable code
723 ; compilation unit finished
726 242: "WRITE-SEQUENCE suboptimality"
727 (observed from clx performance)
728 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
729 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
730 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
731 rather than writing the sequence in one go, leading to severe
732 performance degradation.
733 As of sbcl-0.9.0.36, this is solved for fd-streams, so is less of a
734 problem in practice. (Fully fixing this would require adding a
735 ansi-stream-n-bout slot and associated methods to write a byte
736 sequence to ansi-stream, similar to the existing ansi-stream-sout
739 243: "STYLE-WARNING overenthusiasm for unused variables"
740 (observed from clx compilation)
741 In sbcl-0.7.14, in the presence of the macros
742 (DEFMACRO FOO (X) `(BAR ,X))
743 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
744 somewhat surprising style warnings are emitted for
745 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
747 ; (LAMBDA (Y) (FOO Y))
749 ; caught STYLE-WARNING:
750 ; The variable Y is defined but never used.
752 245: bugs in disassembler
753 b. On X86 operand size prefix is not recognized.
756 (defun foo (&key (a :x))
760 does not cause a warning. (BTW: old SBCL issued a warning, but for a
761 function, which was never called!)
764 Compiler does not emit warnings for
766 a. (lambda () (svref (make-array 8 :adjustable t) 1))
769 (list (let ((y (the real x)))
770 (unless (floatp y) (error ""))
775 (declare (optimize (debug 0)))
776 (declare (type vector x))
777 (list (fill-pointer x)
781 Complex array type does not have corresponding type specifier.
783 This is a problem because the compiler emits optimization notes when
784 you use a non-simple array, and without a type specifier for hairy
785 array types, there's no good way to tell it you're doing it
786 intentionally so that it should shut up and just compile the code.
788 Another problem is confusing error message "asserted type ARRAY
789 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
790 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
792 The last problem is that when type assertions are converted to type
793 checks, types are represented with type specifiers, so we could lose
794 complex attribute. (Now this is probably not important, because
795 currently checks for complex arrays seem to be performed by
799 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
800 warning. Analogous cases with the index and length being equal and
801 greater than 0 are warned for; the problem here seems to be that the
802 type required for an array reference of this type is (INTEGER 0 (0))
803 which is canonicalized to NIL.
808 (t1 (specifier-type s)))
809 (eval `(defstruct ,s))
810 (type= t1 (specifier-type s)))
815 b. The same for CSUBTYPEP.
817 262: "yet another bug in inline expansion of local functions"
818 During inline expansion of a local function Python can try to
819 reference optimized away objects (functions, variables, CTRANs from
820 tags and blocks), which later may lead to problems. Some of the
821 cases are worked around by forbidding expansion in such cases, but
822 the better way would be to reimplement inline expansion by copying
826 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
827 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
828 looks reasonable, if not obviously correct; however, it caused the
829 PPC/Linux port to segfault during warm-init while loading
830 src/pcl/std-class.fasl. A workaround patch was made, but it would
831 be nice to understand why the first patch caused problems, and to
832 fix the cause if possible.
834 268: "wrong free declaration scope"
835 The following code must signal type error:
837 (locally (declare (optimize (safety 3)))
838 (flet ((foo (x &optional (y (car x)))
839 (declare (optimize (safety 0)))
841 (funcall (eval #'foo) 1)))
844 In the following function constraint propagator optimizes nothing:
847 (declare (integer x))
848 (declare (optimize speed))
856 Compilation of the following two forms causes "X is unbound" error:
858 (symbol-macrolet ((x pi))
859 (macrolet ((foo (y) (+ x y)))
860 (declaim (inline bar))
866 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
869 CLHS says that type declaration of a symbol macro should not affect
870 its expansion, but in SBCL it does. (If you like magic and want to
871 fix it, don't forget to change all uses of MACROEXPAND to
875 The following code (taken from CLOCC) takes a lot of time to compile:
878 (declare (type (integer 0 #.large-constant) n))
881 (fixed in 0.8.2.51, but a test case would be good)
884 b. The same as in a., but using MULTIPLE-VALUE-SETQ instead of SETQ.
886 (defmethod faa ((*faa* double-float))
887 (set '*faa* (when (< *faa* 0) (- *faa*)))
889 (faa 1d0) => type error
894 (declare (optimize speed))
895 (loop for i of-type (integer 0) from 0 by 2 below 10
898 uses generic arithmetic.
900 b. (fixed in 0.8.3.6)
902 279: type propagation error -- correctly inferred type goes astray?
903 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
904 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
905 &OPTIONAL), not a (INTEGER 1 536870911)
906 is emitted when compiling this file:
907 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
908 (integer #.most-negative-fixnum 0))
913 (let* (;; Uncomment this for a type mismatch warning indicating
914 ;; that the type of (FOO X) is correctly understood.
915 #+nil (fs-foo (float-sign (foo x)))
916 ;; Uncomment this for a type mismatch warning
917 ;; indicating that the type of (ABS (FOO X)) is
918 ;; correctly understood.
919 #+nil (fs-abs-foo (float-sign (abs (foo x))))
920 ;; something wrong with this one though
921 (abs-foo (abs (foo x))))
922 (declare (type (integer 1 100) abs-foo))
927 283: Thread safety: libc functions
928 There are places that we call unsafe-for-threading libc functions
929 that we should find alternatives for, or put locks around. Known or
930 strongly suspected problems, as of 0.8.3.10: please update this
931 bug instead of creating new ones
933 localtime() - called for timezone calculations in code/time.lisp
935 284: Thread safety: special variables
936 There are lots of special variables in SBCL, and I feel sure that at
937 least some of them are indicative of potentially thread-unsafe
938 parts of the system. See doc/internals/notes/threading-specials
940 286: "recursive known functions"
941 Self-call recognition conflicts with known function
942 recognition. Currently cross compiler and target COMPILE do not
943 recognize recursion, and in target compiler it can be disabled. We
944 can always disable it for known functions with RECURSIVE attribute,
945 but there remains a possibility of a function with a
946 (tail)-recursive simplification pass and transforms/VOPs for base
949 287: PPC/Linux miscompilation or corruption in first GC
950 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
951 segmentation fault is reported at the point of first triggered GC,
952 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
953 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
954 that this merely obscures, not solves, the underlying problem; as and when
955 underlying problems are fixed, it would be worth trying again to provoke
958 288: fundamental cross-compilation issues (from old UGLINESS file)
959 Using host floating point numbers to represent target floating point
960 numbers, or host characters to represent target characters, is
961 theoretically shaky. (The characters are OK as long as the characters
962 are in the ANSI-guaranteed character set, though, so they aren't a
963 real problem as long as the sources don't need anything but that;
964 the floats are a real problem.)
966 289: "type checking and source-transforms"
968 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
971 Our policy is to check argument types at the moment of a call. It
972 disagrees with ANSI, which says that type assertions are put
973 immediately onto argument expressions, but is easier to implement in
974 IR1 and is more compatible to type inference, inline expansion,
975 etc. IR1-transforms automatically keep this policy, but source
976 transforms for associative functions (such as +), being applied
977 during IR1-convertion, do not. It may be tolerable for direct calls
978 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
980 b. Another aspect of this problem is efficiency. [x y + z +]
981 requires less registers than [x y z + +]. This transformation is
982 currently performed with source transforms, but it would be good to
983 also perform it in IR1 optimization phase.
985 290: Alpha floating point and denormalized traps
986 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
987 hardware or kernel deficiency: the status of the enable/disable
988 denormalized-float traps bit seems to be ambiguous; by the time we
989 get to os_restore_fp_control after a trap, denormalized traps seem
990 to be enabled. Since we don't want a trap every time someone uses a
991 denormalized float, in general, we mask out that bit when we restore
992 the control word; however, this clobbers any change the user might
996 (reported by Adam Warner, sbcl-devel 2003-09-23)
998 The --load toplevel argument does not perform any sanitization of its
999 argument. As a result, files with Lisp pathname pattern characters
1000 (#\* or #\?, for instance) or quotation marks can cause the system
1001 to perform arbitrary behaviour.
1004 LOOP with non-constant arithmetic step clauses suffers from overzealous
1005 type constraint: code of the form
1006 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1007 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1008 (0))). However, an integral value of X should be legal, because
1009 successive adds of integers to double-floats produces double-floats,
1010 so none of the type restrictions in the code is violated.
1012 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1013 argument type only after having read a character. This is caused
1014 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1015 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1016 less error prone to have EXPLICIT-CHECK be a local declaration,
1017 being put into the definition, instead of an attribute being kept in
1018 a separate file; maybe also put it into SB-EXT?
1020 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1021 in NOTE-ASSUMED-TYPES
1022 In sbcl-0.8.7.32, compiling the file
1024 (declare (type integer x))
1025 (declare (type (vector (or hash-table bit)) y))
1028 (declare (type integer x))
1029 (declare (type (simple-array base (2)) y))
1032 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1034 303: "nonlinear LVARs" (aka MISC.293)
1036 (multiple-value-call #'list
1038 (multiple-value-prog1
1039 (eval '(values :a :b :c))
1045 (throw 'bar (values 3 4)))))))))))
1047 (BUU 1) returns garbage.
1049 The problem is that both EVALs sequentially write to the same LVAR.
1051 306: "Imprecise unions of array types"
1053 (declare (optimize speed)
1054 (type (or (array cons) (array vector)) x))
1056 (foo #((0))) => TYPE-ERROR
1063 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1064 collect `(array ,(sb-vm:saetp-specifier x)))))
1065 => NIL, T (when it should be T, T)
1067 309: "Dubious values for implementation limits"
1068 (reported by Bruno Haible sbcl-devel "Incorrect value of
1069 multiple-values-limit" 2004-04-19)
1070 (values-list (make-list 1000000)), on x86/linux, signals a stack
1071 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1072 significantly larger than 1000000. There are probably similar
1073 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1074 around the same time regarding a call to LIST on sparc with 1000
1075 arguments) and other implementation limit constants.
1077 314: "LOOP :INITIALLY clauses and scope of initializers"
1078 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1079 test suite, originally by Thomas F. Burdick.
1080 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1081 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1082 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1083 ;; initially clauses are evaluated in the loop prologue, which precedes all
1084 ;; loop code except for the initial settings provided by with, for, or as.
1085 (loop :for x = 0 :then (1+ x)
1086 :for y = (1+ x) :then (ash y 1)
1087 :for z :across #(1 3 9 27 81 243)
1089 :initially (assert (zerop x)) :initially (assert (= 2 w))
1090 :until (>= w 100) :collect w)
1091 Expected: (2 6 15 38)
1094 318: "stack overflow in compiler warning with redefined class"
1095 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1098 (setf (find-class 'foo) nil)
1099 (defstruct foo slot-1)
1100 This used to give a stack overflow from within the printer, which has
1101 been fixed as of 0.8.16.11. Current result:
1103 ; can't compile TYPEP of anonymous or undefined class:
1104 ; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
1106 debugger invoked on a TYPE-ERROR in thread 19973:
1107 The value NIL is not of type FUNCTION.
1109 CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
1110 meant to be enough to delete structure classes from the system?
1112 319: "backquote with comma inside array"
1113 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1115 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1117 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1118 which probably isn't intentional.
1120 323: "REPLACE, BIT-BASH and large strings"
1121 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1122 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1124 (declare (optimize speed (safety 1)))
1125 (let ((x (make-string 140000000))
1126 (y (make-string 140000000)))
1127 (length (replace x y))))
1130 debugger invoked on a TYPE-ERROR in thread 2412:
1131 The value 1120000000 is not of type (MOD 536870911).
1132 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1134 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1135 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1136 for all positive integral <x>. At present, it only works for <x> up
1137 to about 1024 (and similarly for signed-byte), so
1138 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1139 gives an error in sbcl-0.8.10.
1141 325: "CLOSE :ABORT T on supeseding streams"
1142 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1143 file on disk, even if one existed before opening.
1145 The illegality of this is not crystal clear, as the ANSI dictionary
1146 entry for CLOSE says that when :ABORT is T superseded files are not
1147 superseded (ie. the original should be restored), whereas the OPEN
1148 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1149 implementation should not destroy the old file until the new stream
1150 is closed." -- implying that even though undesirable, early deletion
1151 is legal. Restoring the original would none the less be the polite
1154 326: "*PRINT-CIRCLE* crosstalk between streams"
1155 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1156 mixed between streams when output operations are intermingled closely
1157 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1158 test case below), so that e.g. the references #2# appears on a stream
1159 with no preceding #2= on that stream to define it (because the #2= was
1160 sent to another stream).
1161 (cl:in-package :cl-user)
1162 (defstruct foo index)
1163 (defparameter *foo* (make-foo :index 4))
1165 (defparameter *bar* (make-bar))
1166 (defparameter *tangle* (list *foo* *bar* *foo*))
1167 (defmethod print-object ((foo foo) stream)
1168 (let ((index (foo-index foo)))
1169 (format *trace-output*
1170 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1172 (format stream "[FOO ~D]" index))
1174 (let ((tsos (make-string-output-stream))
1175 (ssos (make-string-output-stream)))
1176 (let ((*print-circle* t)
1177 (*trace-output* tsos)
1178 (*standard-output* ssos))
1179 (prin1 *tangle* *standard-output*))
1180 (let ((string (get-output-stream-string ssos)))
1181 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1182 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1183 (error "oops: ~S" string)))))
1184 It might be straightforward to fix this by turning the
1185 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1186 per-stream slots, but (1) it would probably be sort of messy faking
1187 up the special variable binding semantics using UNWIND-PROTECT and
1188 (2) it might be sort of a pain to test that no other bugs had been
1191 328: "Profiling generic functions", transplanted from #241
1192 (from tonyms on #lisp IRC 2003-02-25)
1193 In sbcl-0.7.12.55, typing
1194 (defclass foo () ((bar :accessor foo-bar)))
1197 (defclass foo () ((bar :accessor foo-bar)))
1198 gives the error message
1199 "#:FOO-BAR already names an ordinary function or a macro."
1201 Problem: when a generic function is profiled, it appears as an ordinary
1202 function to PCL. (Remembering the uninterned accessor is OK, as the
1203 redefinition must be able to remove old accessors from their generic
1206 329: "Sequential class redefinition"
1207 reported by Bruno Haible:
1208 (defclass reactor () ((max-temp :initform 10000000)))
1209 (defvar *r1* (make-instance 'reactor))
1210 (defvar *r2* (make-instance 'reactor))
1211 (slot-value *r1* 'max-temp)
1212 (slot-value *r2* 'max-temp)
1213 (defclass reactor () ((uptime :initform 0)))
1214 (slot-value *r1* 'uptime)
1215 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1216 (slot-value *r1* 'max-temp) ; => 10000
1217 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1220 The method effective when the wrapper is obsoleted can be saved
1221 in the wrapper, and then to update the instance just run through
1222 all the old wrappers in order from oldest to newest.
1224 332: "fasl stack inconsistency in structure redefinition"
1225 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1226 Even though structure redefinition is undefined by the standard, the
1227 following behaviour is suboptimal: running
1228 (defun stimulate-sbcl ()
1229 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1230 ;;create a file which redefines a structure incompatibly
1231 (with-open-file (f filename :direction :output :if-exists :supersede)
1232 (print '(defstruct astruct foo) f)
1233 (print '(defstruct astruct foo bar) f))
1234 ;;compile and load the file, then invoke the continue restart on
1235 ;;the structure redefinition error
1236 (handler-bind ((error (lambda (c) (continue c))))
1237 (load (compile-file filename)))))
1239 and choosing the CONTINUE restart yields the message
1240 debugger invoked on a SB-INT:BUG in thread 27726:
1241 fasl stack not empty when it should be
1243 336: "slot-definitions must retain the generic functions of accessors"
1244 reported by Tony Martinez:
1245 (defclass foo () ((bar :reader foo-bar)))
1246 (defun foo-bar (x) x)
1247 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1249 Note: just punting the accessor removal if the fdefinition
1250 is not a generic function is not enough:
1252 (defclass foo () ((bar :reader foo-bar)))
1253 (defvar *reader* #'foo-bar)
1254 (defun foo-bar (x) x)
1255 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1256 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1257 ; the method must be removed
1258 ; by the class redefinition
1260 Fixing this should also fix a subset of #328 -- update the
1261 description with a new test-case then.
1263 337: MAKE-METHOD and user-defined method classes
1264 (reported by Bruno Haible sbcl-devel 2004-06-11)
1268 (defclass user-method (standard-method) (myslot))
1269 (defmacro def-user-method (name &rest rest)
1270 (let* ((lambdalist-position (position-if #'listp rest))
1271 (qualifiers (subseq rest 0 lambdalist-position))
1272 (lambdalist (elt rest lambdalist-position))
1273 (body (subseq rest (+ lambdalist-position 1)))
1275 (subseq lambdalist 0 (or
1277 (lambda (x) (member x lambda-list-keywords))
1279 (length lambdalist))))
1280 (specializers (mapcar #'find-class
1281 (mapcar (lambda (x) (if (consp x) (second x) t))
1283 (unspecialized-required-part
1284 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1285 (unspecialized-lambdalist
1286 (append unspecialized-required-part
1287 (subseq lambdalist (length required-part)))))
1290 (MAKE-INSTANCE 'USER-METHOD
1291 :QUALIFIERS ',qualifiers
1292 :LAMBDA-LIST ',unspecialized-lambdalist
1293 :SPECIALIZERS ',specializers
1295 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1296 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1297 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1298 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1299 (IF (NULL NEXT-METHODS-LIST)
1300 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1301 (FUNCALL (SB-PCL:METHOD-FUNCTION
1302 (FIRST NEXT-METHODS-LIST))
1303 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1304 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1308 (defgeneric test-um03 (x))
1309 (defmethod test-um03 ((x integer))
1310 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1311 (def-user-method test-um03 ((x rational))
1312 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1313 (defmethod test-um03 ((x real))
1314 (list 'real x (not (null (next-method-p)))))
1319 (defgeneric test-um10 (x))
1320 (defmethod test-um10 ((x integer))
1321 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1322 (defmethod test-um10 ((x rational))
1323 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1324 (defmethod test-um10 ((x real))
1325 (list 'real x (not (null (next-method-p)))))
1326 (defmethod test-um10 :after ((x real)))
1327 (def-user-method test-um10 :around ((x integer))
1328 (list* 'around-integer x
1329 (not (null (next-method-p))) (call-next-method)))
1330 (defmethod test-um10 :around ((x rational))
1331 (list* 'around-rational x
1332 (not (null (next-method-p))) (call-next-method)))
1333 (defmethod test-um10 :around ((x real))
1334 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1336 fails with a type error, and
1339 (defgeneric test-um12 (x))
1340 (defmethod test-um12 ((x integer))
1341 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1342 (defmethod test-um12 ((x rational))
1343 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1344 (defmethod test-um12 ((x real))
1345 (list 'real x (not (null (next-method-p)))))
1346 (defmethod test-um12 :after ((x real)))
1347 (defmethod test-um12 :around ((x integer))
1348 (list* 'around-integer x
1349 (not (null (next-method-p))) (call-next-method)))
1350 (defmethod test-um12 :around ((x rational))
1351 (list* 'around-rational x
1352 (not (null (next-method-p))) (call-next-method)))
1353 (def-user-method test-um12 :around ((x real))
1354 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1356 fails with NO-APPLICABLE-METHOD.
1358 339: "DEFINE-METHOD-COMBINATION bugs"
1359 (reported by Bruno Haible via the clisp test suite)
1361 a. Syntax checking laxity (should produce errors):
1362 i. (define-method-combination foo :documentation :operator)
1363 ii. (define-method-combination foo :documentation nil)
1364 iii. (define-method-combination foo nil)
1365 iv. (define-method-combination foo nil nil
1366 (:arguments order &aux &key))
1367 v. (define-method-combination foo nil nil (:arguments &whole))
1368 vi. (define-method-combination foo nil nil (:generic-function))
1369 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1370 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1371 ix. (define-method-combination foo nil ((3)))
1372 x. (define-method-combination foo nil ((a)))
1374 b. define-method-combination arguments lambda list badness
1375 i. &aux args are currently unsupported;
1376 ii. default values of &optional and &key arguments are ignored;
1377 iii. supplied-p variables for &optional and &key arguments are not
1380 c. qualifier matching incorrect
1382 (define-method-combination mc27 ()
1384 (ignored (:ignore :unused)))
1386 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1387 (defgeneric test-mc27 (x)
1388 (:method-combination mc27)
1389 (:method :ignore ((x number)) (/ 0)))
1392 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1393 method is applicable, and yet matches neither of the method group
1396 341: PPRINT-LOGICAL-BLOCK / PPRINT-FILL / PPRINT-LINEAR sharing detection.
1397 (from Paul Dietz' test suite)
1399 CLHS on PPRINT-LINEAR and PPRINT-FILL (and PPRINT-TABULAR, though
1400 that's slightly different) states that these functions perform
1401 circular and shared structure detection on their object. Therefore,
1403 a.(let ((*print-circle* t))
1404 (pprint-linear *standard-output* (let ((x '(a))) (list x x))))
1405 should print "(#1=(A) #1#)"
1407 b.(let ((*print-circle* t))
1408 (pprint-linear *standard-output*
1409 (let ((x (cons nil nil))) (setf (cdr x) x) x)))
1410 should print "#1=(NIL . #1#)"
1412 (it is likely that the fault lies in PPRINT-LOGICAL-BLOCK, as
1413 suggested by the suggested implementation of PPRINT-TABULAR)
1415 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1416 Even the simplest possible overriding of
1417 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1418 as "canonical", does not work:
1419 (defclass my-generic-function (standard-generic-function) ()
1420 (:metaclass funcallable-standard-class))
1421 (defmethod compute-discriminating-function ((gf my-generic-function))
1422 (let ((dfun (call-next-method)))
1423 (lambda (&rest args)
1424 (apply dfun args))))
1426 (:generic-function-class my-generic-function))
1427 (defmethod foo (x) (+ x x))
1429 signals an error. This error is the same even if the LAMBDA is
1430 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1431 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1432 src/code/target-defstruct.lisp is broken? This seems to be broken
1433 in CMUCL 18e, so it's not caused by a recent change.
1435 344: more (?) ROOM T problems (possibly part of bug 108)
1436 In sbcl-0.8.12.51, and off and on leading up to it, the
1437 SB!VM:MEMORY-USAGE operations in ROOM T caused
1438 unhandled condition (of type SB-INT:BUG):
1439 failed AVER: "(SAP= CURRENT END)"
1440 Several clever people have taken a shot at this without fixing
1441 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1442 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1443 out how to make them work reliably with the rest of the GC.
1445 (Note: there's at least one dubious thing in room.lisp: see the
1446 comment in VALID-OBJ)
1448 346: alpha backtrace
1449 In sbcl-0.8.13, all backtraces from errors caused by internal errors
1450 on the alpha seem to have a "bogus stack frame".
1452 349: PPRINT-INDENT rounding implementation decisions
1453 At present, pprint-indent (and indeed the whole pretty printer)
1454 more-or-less assumes that it's using a monospace font. That's
1455 probably not too silly an assumption, but one piece of information
1456 the current implementation loses is from requests to indent by a
1457 non-integral amount. As of sbcl-0.8.15.9, the system silently
1458 truncates the indentation to an integer at the point of request, but
1459 maybe the non-integral value should be propagated through the
1460 pprinter and only truncated at output? (So that indenting by 1/2
1461 then 3/2 would indent by two spaces, not one?)
1463 352: forward-referenced-class trouble
1464 reported by Bruno Haible on sbcl-devel
1466 (setf (class-name (find-class 'a)) 'b)
1470 Expected: an instance of c, with a slot named x
1471 Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
1472 While computing the class precedence list of the class named C.
1473 The class named B is a forward referenced class.
1474 The class named B is a direct superclass of the class named C.
1476 353: debugger suboptimalities on x86
1477 On x86 backtraces for undefined functions start with a bogus stack
1478 frame, and backtraces for throws to unknown catch tags with a "no
1479 debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
1480 (used on non-x86 platforms) being a more complete solution then what
1483 On x86/linux large portions of tests/debug.impure.lisp have been commented
1484 out as failures. The probable culprit for these problems is in x86-call-context
1485 (things work fine on x86/freebsd).
1487 More generally, the debugger internals suffer from excessive x86/non-x86
1488 conditionalization and OAOOMization: refactoring the common parts would
1491 354: XEPs in backtraces
1492 Under default compilation policy
1496 Has the XEP for TEST in the backtrace, not the TEST frame itself.
1497 (sparc and x86 at least)
1499 Since SBCL 0.8.20.1 this is hidden unless *SHOW-ENTRY-POINT-DETAILS*
1500 is true (instead there appear two TEST frames at least on ppc). The
1501 underlying cause seems to be that SB-C::TAIL-ANNOTATE will not merge
1502 the tail-call for the XEP, since Python has by that time proved that
1503 the function can never return; same happens if the function holds an
1504 unconditional call to ERROR.
1506 355: change-class of generic-function
1507 (reported by Bruno Haible)
1508 The MOP doesn't support change-class on a generic-function. However, SBCL
1509 apparently supports it, since it doesn't give an error or warning when doing
1510 so so. Then, however, it produces wrong results for calls to this generic
1512 ;;; The effective-methods cache:
1514 (defgeneric testgf35 (x))
1515 (defmethod testgf35 ((x integer))
1516 (cons 'integer (if (next-method-p) (call-next-method))))
1517 (defmethod testgf35 ((x real))
1518 (cons 'real (if (next-method-p) (call-next-method))))
1519 (defclass customized5-generic-function (standard-generic-function)
1521 (:metaclass sb-pcl:funcallable-standard-class))
1522 (defmethod sb-pcl:compute-effective-method ((gf customized5-generic-function) method-combination methods)
1523 `(REVERSE ,(call-next-method)))
1527 (change-class #'testgf35 'customized5-generic-function)
1529 Expected: ((INTEGER REAL) (REAL INTEGER))
1530 Got: ((INTEGER REAL) (INTEGER REAL))
1531 ;;; The discriminating-function cache:
1533 (defgeneric testgf36 (x))
1534 (defmethod testgf36 ((x integer))
1535 (cons 'integer (if (next-method-p) (call-next-method))))
1536 (defmethod testgf36 ((x real))
1537 (cons 'real (if (next-method-p) (call-next-method))))
1538 (defclass customized6-generic-function (standard-generic-function)
1540 (:metaclass sb-pcl:funcallable-standard-class))
1541 (defmethod sb-pcl:compute-discriminating-function ((gf customized6-generic-function))
1542 (let ((orig-df (call-next-method)))
1543 #'(lambda (&rest arguments)
1544 (reverse (apply orig-df arguments)))))
1548 (change-class #'testgf36 'customized6-generic-function)
1550 Expected: ((INTEGER REAL) (REAL INTEGER))
1551 Got: ((INTEGER REAL) (INTEGER REAL))
1554 (reported by Bruno Haible)
1555 After the "layout depth conflict" error, the CLOS is left in a state where
1556 it's not possible to define new standard-class subclasses any more.
1558 (defclass prioritized-dispatcher ()
1559 ((dependents :type list :initform nil)))
1560 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1561 (c2 (eql (find-class 'prioritized-dispatcher))))
1563 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1565 (:metaclass sb-pcl:funcallable-standard-class))
1566 ;; ERROR, Quit the debugger with ABORT
1567 (defclass typechecking-reader-class (standard-class)
1569 Expected: #<STANDARD-CLASS TYPECHECKING-READER-CLASS>
1570 Got: ERROR "The assertion SB-PCL::WRAPPERS failed."
1572 357: defstruct inheritance of initforms
1573 (reported by Bruno Haible)
1574 When defstruct and defclass (with :metaclass structure-class) are mixed,
1575 1. some slot initforms are ignored by the DEFSTRUCT generated constructor
1577 2. all slot initforms are ignored by MAKE-INSTANCE. (This can be arguably
1578 OK for initforms that were given in a DEFSTRUCT form, but for those
1579 given in a DEFCLASS form, I think it qualifies as a bug.)
1581 (defstruct structure02a
1585 (defclass structure02b (structure02a)
1586 ((slot4 :initform -44)
1589 (slot7 :initform (floor (* pi pi)))
1590 (slot8 :initform 88))
1591 (:metaclass structure-class))
1592 (defstruct (structure02c (:include structure02b (slot8 -88)))
1595 (slot11 (floor (exp 3))))
1597 (let ((a (make-structure02c)))
1598 (list (structure02c-slot4 a)
1599 (structure02c-slot5 a)
1600 (structure02c-slot6 a)
1601 (structure02c-slot7 a)))
1602 Expected: (-44 nil t 9)
1603 Got: (SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..
1604 SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..)
1606 (let ((b (make-instance 'structure02c)))
1607 (list (structure02c-slot2 b)
1608 (structure02c-slot3 b)
1609 (structure02c-slot4 b)
1610 (structure02c-slot6 b)
1611 (structure02c-slot7 b)
1612 (structure02c-slot8 b)
1613 (structure02c-slot10 b)
1614 (structure02c-slot11 b)))
1615 Expected: (t 3 -44 t 9 -88 t 20)
1616 Got: (0 0 0 0 0 0 0 0)
1618 358: :DECLARE argument to ENSURE-GENERIC-FUNCTION
1619 (reported by Bruno Haible)
1620 According to ANSI CL, ensure-generic-function must accept a :DECLARE
1621 keyword argument. In SBCL 0.8.16 it does not.
1624 (ensure-generic-function 'foo113 :declare '((optimize (speed 3))))
1625 (sb-pcl:generic-function-declarations #'foo113))
1626 Expected: ((OPTIMIZE (SPEED 3)))
1628 Invalid initialization argument:
1630 in call for class #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>.
1632 The ANSI Standard, Section 7.1.2
1634 Bruno notes: The MOP specifies that ensure-generic-function accepts :DECLARATIONS.
1635 The easiest way to be compliant to both specs is to accept both (exclusively
1638 359: wrong default value for ensure-generic-function's :generic-function-class argument
1639 (reported by Bruno Haible)
1640 ANSI CL is silent on this, but the MOP's specification of ENSURE-GENERIC-FUNCTION says:
1641 "The remaining arguments are the complete set of keyword arguments
1642 received by ENSURE-GENERIC-FUNCTION."
1643 and the spec of ENSURE-GENERIC-FUNCTION-USING-CLASS:
1644 ":GENERIC-FUNCTION-CLASS - a class metaobject or a class name. If it is not
1645 supplied, it defaults to the class named STANDARD-GENERIC-FUNCTION."
1646 This is not the case in SBCL. Test case:
1647 (defclass my-generic-function (standard-generic-function)
1649 (:metaclass sb-pcl:funcallable-standard-class))
1650 (setf (fdefinition 'foo1)
1651 (make-instance 'my-generic-function :name 'foo1))
1652 (ensure-generic-function 'foo1
1653 :generic-function-class (find-class 'standard-generic-function))
1655 ; => #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1656 (setf (fdefinition 'foo2)
1657 (make-instance 'my-generic-function :name 'foo2))
1658 (ensure-generic-function 'foo2)
1660 Expected: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1661 Got: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS MY-GENERIC-FUNCTION>
1663 360: CALL-METHOD not recognized in method-combination body
1664 (reported by Bruno Haible)
1665 This method combination, which adds 'redo' and 'return' restarts for each
1666 method invocation to standard method combination, gives an error in SBCL.
1667 (defun prompt-for-new-values ()
1668 (format *debug-io* "~&New values: ")
1669 (list (read *debug-io*)))
1670 (defun add-method-restarts (form method)
1671 (let ((block (gensym))
1679 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Try calling ~S again." ,method))
1682 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Specify return values for ~S call." ,method))
1683 :INTERACTIVE (LAMBDA () (PROMPT-FOR-NEW-VALUES))
1684 (RETURN-FROM ,block (VALUES-LIST L)))))))))
1685 (defun convert-effective-method (efm)
1687 (if (eq (car efm) 'CALL-METHOD)
1688 (let ((method-list (third efm)))
1689 (if (or (typep (first method-list) 'method) (rest method-list))
1690 ; Reduce the case of multiple methods to a single one.
1691 ; Make the call to the next-method explicit.
1692 (convert-effective-method
1693 `(CALL-METHOD ,(second efm)
1695 (CALL-METHOD ,(first method-list) ,(rest method-list))))))
1696 ; Now the case of at most one method.
1697 (if (typep (second efm) 'method)
1698 ; Wrap the method call in a RESTART-CASE.
1699 (add-method-restarts
1700 (cons (convert-effective-method (car efm))
1701 (convert-effective-method (cdr efm)))
1703 ; Normal recursive processing.
1704 (cons (convert-effective-method (car efm))
1705 (convert-effective-method (cdr efm))))))
1706 (cons (convert-effective-method (car efm))
1707 (convert-effective-method (cdr efm))))
1709 (define-method-combination standard-with-restarts ()
1712 (primary () :required t)
1714 (flet ((call-methods-sequentially (methods)
1715 (mapcar #'(lambda (method)
1716 `(CALL-METHOD ,method))
1718 (let ((form (if (or before after (rest primary))
1719 `(MULTIPLE-VALUE-PROG1
1721 ,@(call-methods-sequentially before)
1722 (CALL-METHOD ,(first primary) ,(rest primary)))
1723 ,@(call-methods-sequentially (reverse after)))
1724 `(CALL-METHOD ,(first primary)))))
1727 `(CALL-METHOD ,(first around)
1728 (,@(rest around) (MAKE-METHOD ,form)))))
1729 (convert-effective-method form))))
1730 (defgeneric testgf16 (x) (:method-combination standard-with-restarts))
1731 (defclass testclass16a () ())
1732 (defclass testclass16b (testclass16a) ())
1733 (defclass testclass16c (testclass16a) ())
1734 (defclass testclass16d (testclass16b testclass16c) ())
1735 (defmethod testgf16 ((x testclass16a))
1737 (not (null (find-restart 'method-redo)))
1738 (not (null (find-restart 'method-return)))))
1739 (defmethod testgf16 ((x testclass16b))
1740 (cons 'b (call-next-method)))
1741 (defmethod testgf16 ((x testclass16c))
1742 (cons 'c (call-next-method)))
1743 (defmethod testgf16 ((x testclass16d))
1744 (cons 'd (call-next-method)))
1745 (testgf16 (make-instance 'testclass16d))
1747 Expected: (D B C A T T)
1748 Got: ERROR CALL-METHOD outside of a effective method form
1750 This is a bug because ANSI CL HyperSpec/Body/locmac_call-m__make-method
1752 "The macro call-method invokes the specified method, supplying it with
1753 arguments and with definitions for call-next-method and for next-method-p.
1754 If the invocation of call-method is lexically inside of a make-method,
1755 the arguments are those that were supplied to that method. Otherwise
1756 the arguments are those that were supplied to the generic function."
1757 and the example uses nothing more than these two cases (as you can see by
1758 doing (trace convert-effective-method)).
1760 361: initialize-instance of standard-reader-method ignores :function argument
1761 (reported by Bruno Haible)
1762 Pass a custom :function argument to initialize-instance of a
1763 standard-reader-method instance, but it has no effect.
1764 ;; Check that it's possible to define reader methods that do typechecking.
1766 (defclass typechecking-reader-method (sb-pcl:standard-reader-method)
1768 (defmethod initialize-instance ((method typechecking-reader-method) &rest initargs
1769 &key slot-definition)
1770 (let ((name (sb-pcl:slot-definition-name slot-definition))
1771 (type (sb-pcl:slot-definition-type slot-definition)))
1772 (apply #'call-next-method method
1773 :function #'(lambda (args next-methods)
1774 (declare (ignore next-methods))
1775 (apply #'(lambda (instance)
1776 (let ((value (slot-value instance name)))
1777 (unless (typep value type)
1778 (error "Slot ~S of ~S is not of type ~S: ~S"
1779 name instance type value))
1783 (defclass typechecking-reader-class (standard-class)
1785 (defmethod sb-pcl:validate-superclass ((c1 typechecking-reader-class) (c2 standard-class))
1787 (defmethod reader-method-class ((class typechecking-reader-class) direct-slot &rest args)
1788 (find-class 'typechecking-reader-method))
1789 (defclass testclass25 ()
1790 ((pair :type (cons symbol (cons symbol null)) :initarg :pair :accessor testclass25-pair))
1791 (:metaclass typechecking-reader-class))
1792 (macrolet ((succeeds (form)
1793 `(not (nth-value 1 (ignore-errors ,form)))))
1794 (let ((p (list 'abc 'def))
1795 (x (make-instance 'testclass25)))
1796 (list (succeeds (make-instance 'testclass25 :pair '(seventeen 17)))
1797 (succeeds (setf (testclass25-pair x) p))
1798 (succeeds (setf (second p) 456))
1799 (succeeds (testclass25-pair x))
1800 (succeeds (slot-value x 'pair))))))
1801 Expected: (t t t nil t)
1804 (inspect (first (sb-pcl:generic-function-methods #'testclass25-pair)))
1805 shows that the method was created with a FAST-FUNCTION slot but with a
1806 FUNCTION slot of NIL.
1808 362: missing error when a slot-definition is created without a name
1809 (reported by Bruno Haible)
1810 The MOP says about slot-definition initialization:
1811 "The :NAME argument is a slot name. An ERROR is SIGNALled if this argument
1812 is not a symbol which can be used as a variable name. An ERROR is SIGNALled
1813 if this argument is not supplied."
1815 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition))
1817 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION NIL>
1819 363: missing error when a slot-definition is created with a wrong documentation object
1820 (reported by Bruno Haible)
1821 The MOP says about slot-definition initialization:
1822 "The :DOCUMENTATION argument is a STRING or NIL. An ERROR is SIGNALled
1823 if it is not. This argument default to NIL during initialization."
1825 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition)
1827 :documentation 'not-a-string)
1829 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION FOO>
1831 364: does not support class objects as specializer names
1832 (reported by Bruno Haible)
1833 According to ANSI CL 7.6.2, class objects are valid specializer names,
1834 and "Parameter specializer names are used in macros intended as the
1835 user-level interface (defmethod)". DEFMETHOD's syntax section doesn't
1836 mention this possibility in the BNF for parameter-specializer-name;
1837 however, this appears to be an editorial omission, since the CLHS
1838 mentions issue CLASS-OBJECT-SPECIALIZER:AFFIRM as being approved
1839 by X3J13. SBCL doesn't support it:
1840 (defclass foo () ())
1841 (defmethod goo ((x #.(find-class 'foo))) x)
1842 Expected: #<STANDARD-METHOD GOO (#<STANDARD-CLASS FOO>)>
1843 Got: ERROR "#<STANDARD-CLASS FOO> is not a legal class name."
1845 365: mixin on generic-function subclass
1846 (reported by Bruno Haible)
1848 (defclass prioritized-dispatcher ()
1849 ((dependents :type list :initform nil)))
1850 on a generic-function subclass:
1851 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1853 (:metaclass sb-pcl:funcallable-standard-class))
1854 SBCL gives an error on this, telling to define a method on SB-MOP:VALIDATE-SUPERCLASS. If done,
1855 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1856 (c2 (eql (find-class 'prioritized-dispatcher))))
1859 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1861 (:metaclass sb-pcl:funcallable-standard-class))
1862 => debugger invoked on a SIMPLE-ERROR in thread 6687:
1863 layout depth conflict: #(#<SB-KERNEL:LAYOUT for T {500E1E9}> ...)
1865 Further discussion on this: http://thread.gmane.org/gmane.lisp.steel-bank.general/491
1867 366: cannot define two generic functions with user-defined class
1868 (reported by Bruno Haible)
1869 it is possible to define one generic function class and an instance
1870 of it. But attempting to do the same thing again, in the same session,
1871 leads to a "Control stack exhausted" error. Test case:
1872 (defclass my-generic-function-1 (standard-generic-function)
1874 (:metaclass sb-pcl:funcallable-standard-class))
1875 (defgeneric testgf-1 (x) (:generic-function-class my-generic-function-1)
1876 (:method ((x integer)) (cons 'integer nil)))
1877 (defclass my-generic-function-2 (standard-generic-function)
1879 (:metaclass sb-pcl:funcallable-standard-class))
1880 (defgeneric testgf-2 (x) (:generic-function-class my-generic-function-2)
1881 (:method ((x integer)) (cons 'integer nil)))
1882 => SB-KERNEL::CONTROL-STACK-EXHAUSTED
1884 367: TYPE-ERROR at compile time, undetected TYPE-ERROR at runtime
1886 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1890 (i367s (make-array 0 :fill-pointer t) :type (or (vector i367) null)))
1892 (g367 (error "missing :G367") :type g367 :read-only t))
1893 ;;; In sbcl-0.8.18, commenting out this (DECLAIM (FTYPE ... R367))
1894 ;;; gives an internal error at compile time:
1895 ;;; The value #<SB-KERNEL:NAMED-TYPE NIL> is not of
1896 ;;; type SB-KERNEL:VALUES-TYPE.
1897 (declaim (ftype (function ((vector i367) e367) (or s367 null)) r367))
1898 (declaim (ftype (function ((vector e367)) (values)) h367))
1900 (let ((x (g367-i367s (make-g367))))
1901 (let* ((y (or (r367 x w)
1904 (format t "~&Y=~S Z=~S~%" y z)
1906 (defun r367 (x y) (declare (ignore x y)) nil)
1907 (defun h367 (x) (declare (ignore x)) (values))
1908 ;;; In sbcl-0.8.18, executing this form causes an low-level error
1909 ;;; segmentation violation at #X9B0E1F4
1910 ;;; (instead of the TYPE-ERROR that one might like).
1911 (frob 0 (make-e367))
1912 can be made to cause two different problems, as noted in the comments:
1913 bug 367a: Compile and load the file. No TYPE-ERROR is signalled at
1914 run time (in the (S367-G367 Y) form of FROB, when Y is NIL
1915 instead of an instance of S367). Instead (on x86/Linux at least)
1916 we end up with a segfault.
1917 bug 367b: Comment out the (DECLAIM (FTYPE ... R367)), and compile
1918 the file. The compiler fails with TYPE-ERROR at compile time.
1920 368: miscompiled OR (perhaps related to bug 367)
1921 Trying to relax type declarations to find a workaround for bug 367,
1922 it turns out that even when the return type isn't declared (or
1923 declared to be T, anyway) the system remains confused about type
1924 inference in code similar to that for bug 367:
1925 (in-package :cl-user)
1926 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1930 (i368s (make-array 0 :fill-pointer t) :type (or (vector i368) null)))
1932 (g368 (error "missing :G368") :type g368 :read-only t))
1933 (declaim (ftype (function (fixnum (vector i368) e368) t) r368))
1934 (declaim (ftype (function (fixnum (vector e368)) t) h368))
1935 (defparameter *h368-was-called-p* nil)
1936 (defun nsu (vertices e368)
1937 (let ((i368s (g368-i368s (make-g368))))
1938 (let ((fuis (r368 0 i368s e368)))
1939 (format t "~&FUIS=~S~%" fuis)
1940 (or fuis (h368 0 i368s)))))
1942 (declare (ignore w x y))
1945 (declare (ignore w x))
1946 (setf *h368-was-called-p* t)
1947 (make-s368 :g368 (make-g368)))
1949 (format t "~&calling NSU~%")
1950 (let ((nsu (nsu #() (make-e368))))
1951 (format t "~&NSU returned ~S~%" nsu)
1952 (format t "~&*H368-WAS-CALLED-P*=~S~%" *h368-was-called-p*)
1953 (assert (s368-p nsu))
1954 (assert *h368-was-called-p*))
1955 In sbcl-0.8.18, both ASSERTs fail, and (DISASSEMBLE 'NSU) shows
1956 that no call to H368 is compiled.
1958 369: unlike-an-intersection behavior of VALUES-TYPE-INTERSECTION
1959 In sbcl-0.8.18.2, the identity $(x \cap y \cap y)=(x \cap y)$
1960 does not hold for VALUES-TYPE-INTERSECTION, even for types which
1961 can be intersected exactly, so that ASSERTs fail in this test case:
1962 (in-package :cl-user)
1963 (let ((types (mapcar #'sb-c::values-specifier-type
1964 '((values (vector package) &optional)
1965 (values (vector package) &rest t)
1966 (values (vector hash-table) &rest t)
1967 (values (vector hash-table) &optional)
1968 (values t &optional)
1970 (values nil &optional)
1971 (values nil &rest t)
1972 (values sequence &optional)
1973 (values sequence &rest t)
1974 (values list &optional)
1975 (values list &rest t)))))
1978 (let ((i (sb-c::values-type-intersection x y)))
1979 (assert (sb-c::type= i (sb-c::values-type-intersection i x)))
1980 (assert (sb-c::type= i (sb-c::values-type-intersection i y)))))))
1982 370: reader misbehaviour on large-exponent floats
1983 (read-from-string "1.0s1000000000000000000000000000000000000000")
1984 causes the reader to attempt to create a very large bignum (which it
1985 will then attempt to coerce to a rational). While this isn't
1986 completely wrong, it is probably not ideal -- checking the floating
1987 point control word state and then returning the relevant float
1988 (most-positive-short-float or short-float-infinity) or signalling an
1989 error immediately would seem to make more sense.
1991 372: floating-point overflow not signalled on ppc/darwin
1992 The following assertions in float.pure.lisp fail on ppc/darwin
1993 (Mac OS X version 10.3.7):
1994 (assert (raises-error? (scale-float 1.0 most-positive-fixnum)
1995 floating-point-overflow))
1996 (assert (raises-error? (scale-float 1.0d0 (1+ most-positive-fixnum))
1997 floating-point-overflow)))
1998 as the SCALE-FLOAT just returns
1999 #.SB-EXT:SINGLE/DOUBLE-FLOAT-POSITIVE-INFINITY. These tests have been
2000 disabled on Darwin for now.
2002 374: BIT-AND problem on ppc/darwin:
2003 The BIT-AND test in bit-vector.impure-cload.lisp results in
2004 fatal error encountered in SBCL pid 8356:
2005 GC invariant lost, file "gc-common.c", line 605
2006 on ppc/darwin. Test disabled for the duration.
2009 (compile nil '(lambda (p1)
2010 (declare (optimize (speed 1) (safety 2) (debug 2) (space 0))
2014 fails on hairy type check in IR2.
2016 1. KEYWORDP is MAYBE-INLINE expanded (before TYPEP-like
2017 transformation could eliminate it).
2019 2. From the only call of KEYWORDP the type of its argument is
2020 derived to be KEYWORD.
2022 2. Type check for P1 is generated; it uses KEYWORDP to perform the
2023 check, and so references the local function; from the KEYWORDP
2024 argument type new CAST to KEYWORD is generated. The compiler
2027 377: Memory fault error reporting
2028 On those architectures where :C-STACK-IS-CONTROL-STACK is in
2029 *FEATURES*, we handle SIG_MEMORY_FAULT (SEGV or BUS) on an altstack,
2030 so we cannot handle the signal directly (as in interrupt_handle_now())
2031 in the case when the signal comes from some external agent (the user
2032 using kill(1), or a fault in some foreign code, for instance). As
2033 of sbcl-0.8.20.20, this is fixed by calling
2034 arrange_return_to_lisp_function() to a new error-signalling
2035 function, but as a result the error reporting is poor: we cannot
2036 even tell the user at which address the fault occurred. We should
2037 arrange such that arguments can be passed to the function called from
2038 arrange_return_to_lisp_function(), but this looked hard to do in
2039 general without suffering from memory leaks.
2041 379: TRACE :ENCAPSULATE NIL broken on ppc/darwin
2042 See commented-out test-case in debug.impure.lisp.
2044 380: Accessor redefinition fails because of old accessor name
2045 When redefining an accessor, SB-PCL::FIX-SLOT-ACCESSORS may try to
2046 find the generic function named by the old accessor name using
2047 ENSURE-GENERIC-FUNCTION and then remove the old accessor's method in
2048 the GF. If the old name does not name a function, or if the old name
2049 does not name a generic function, no attempt to find the GF or remove
2050 any methods is made.
2052 However, if an unrelated GF with an incompatible lambda list exists,
2053 the class redefinition will fail when SB-PCL::REMOVE-READER-METHOD
2054 tries to find and remove a method with an incompatible lambda list
2055 from the unrelated generic function.
2057 381: incautious calls to EQUAL in fasl dumping
2059 (frob #(#1=(a #1#)))
2060 (frob #(#1=(b #1#)))
2061 (frob #(#1=(a #1#)))
2062 in sbcl-0.9.0 causes CONTROL-STACK-EXHAUSTED. My (WHN) impression
2063 is that this follows from the use of (MAKE-HASH-TABLE :TEST 'EQUAL)
2064 to detect sharing, in which case fixing it might require either
2065 getting less ambitious about detecting shared list structure, or
2066 implementing the moral equivalent of EQUAL hash tables in a
2069 382: externalization unexpectedly changes array simplicity
2070 COMPILE-FILE and LOAD
2072 (let ((x #.(make-array 4 :fill-pointer 0)))
2073 (values (eval `(typep ',x 'simple-array))
2074 (typep x 'simple-array))))
2075 then (FOO) => T, NIL.
2077 Similar problems exist with SIMPLE-ARRAY-P, ARRAY-HEADER accessors
2078 and all array dimension functions.
2080 383: ASH'ing non-constant zeros
2083 (declare (type (integer -2 14) b))
2084 (declare (ignorable b))
2085 (ash (imagpart b) 57))
2086 on PPC (and other platforms, presumably) gives an error during the
2087 emission of FASH-ASH-LEFT/FIXNUM=>FIXNUM as the assembler attempts to
2088 stuff a too-large constant into the immediate field of a PPC
2089 instruction. Either the VOP should be fixed or the compiler should be
2090 taught how to transform this case away, paying particular attention
2091 to side-effects that might occur in the arguments to ASH.
2093 384: Compiler runaway on very large character types
2095 (compile nil '(lambda (x)
2096 (declare (type (member #\a 1) x))
2097 (the (member 1 nil) x)))
2099 The types apparently normalize into a very large type, and the compiler
2100 gets lost in REMOVE-DUPLICATES. Perhaps the latter should use
2101 a better algorithm (one based on hash tables, say) on very long lists
2102 when :TEST has its default value?
2106 (compile nil '(lambda (x) (the (not (eql #\a)) x)))
2108 (partially fixed in 0.9.3.1, but a better representation for these
2112 (format nil "~4,1F" 0.001) => "0.00" (should be " 0.0");
2113 (format nil "~4,1@F" 0.001) => "+.00" (should be "+0.0").