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.
97 Currently INSPECT and DESCRIBE do show the values, but showing the
98 names of the bindings would be even nicer.
101 The compiler assumes that any time a function of declared FTYPE
102 doesn't signal an error, its arguments were of the declared type.
103 E.g. compiling and loading
104 (DECLAIM (OPTIMIZE (SAFETY 3)))
105 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
107 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
109 (COND ((> (FACTORIAL X) 1.0E6)
110 (FORMAT T "too big~%"))
112 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
114 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
117 will cause the INTEGERP case to be selected, giving bogus output a la
119 This violates the "declarations are assertions" principle.
120 According to the ANSI spec, in the section "System Class FUNCTION",
121 this is a case of "lying to the compiler", but the lying is done
122 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
123 not by the unexpectedly general definition of FACTORIAL. In any case,
124 "declarations are assertions" means that lying to the compiler should
125 cause an error to be signalled, and should not cause a bogus
126 result to be returned. Thus, the compiler should not assume
127 that arbitrary functions check their argument types. (It might
128 make sense to add another flag (CHECKED?) to DEFKNOWN to
129 identify functions which *do* check their argument types.)
130 (Also, verify that the compiler handles declared function
131 return types as assertions.)
134 The definitions of SIGCONTEXT-FLOAT-REGISTER and
135 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
136 supported on FreeBSD because the floating point state is not saved,
137 but at least as of FreeBSD 4.0, the floating point state *is* saved,
138 so they could be supported after all. Very likely
139 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
142 Compiling and loading
143 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
145 then requesting a BACKTRACE at the debugger prompt gives no information
146 about where in the user program the problem occurred.
148 (this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
149 while giving the backtrace the non-x86 systems complains about "unknown
150 source location: using block start", but apart from that the
151 backtrace seems reasonable. On x86 this is masked by bug 353. See
152 tests/debug.impure.lisp for a test case)
155 Using the pretty-printer from the command prompt gives funny
156 results, apparently because the pretty-printer doesn't know
157 about user's command input, including the user's carriage return
158 that the user, and therefore the pretty-printer thinks that
159 the new output block should start indented 2 or more characters
160 rightward of the correct location.
163 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
164 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
165 crashes SBCL. In general tracing anything which is used in the
166 implementation of TRACE is likely to have the same problem.
169 ANSI says in one place that type declarations can be abbreviated even
170 when the type name is not a symbol, e.g.
171 (DECLAIM ((VECTOR T) *FOOVECTOR*))
172 SBCL doesn't support this. But ANSI says in another place that this
173 isn't allowed. So it's not clear this is a bug after all. (See the
174 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
178 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
179 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
180 the mean of the distribution can be systematically O(0.1%) wrong.
181 Just increasing R-I-E-B is probably not a good solution, since
182 it would decrease efficiency more than is probably necessary. Perhaps
183 using some sort of accept/reject method would be better.
186 Internally the compiler sometimes evaluates
187 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
188 (I stumbled across this when I added an
189 (assert (not (eq type1 *wild-type*)))
190 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
191 in a type context should probably be translated to T, and so it's
192 probably wrong to ask whether it's equal to the T type and then (using
193 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
194 (I haven't tried to investigate this bug enough to guess whether
195 there might be any user-level symptoms.)
197 In fact, the type system is likely to depend on this inequality not
198 holding... * is not equivalent to T in many cases, such as
199 (VECTOR *) /= (VECTOR T).
202 The facility for dumping a running Lisp image to disk gets confused
203 when run without the PURIFY option, and creates an unnecessarily large
204 core file (apparently representing memory usage up to the previous
205 high-water mark). Moreover, when the file is loaded, it confuses the
206 GC, so that thereafter memory usage can never be reduced below that
209 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
210 The interaction between gencgc and the variables used by
211 save-lisp-and-die is still nonoptimal, though, so no respite from
215 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
216 CL), out-of-line structure slot setters are horribly inefficient
217 whenever the type of the slot is declared, because out-of-line
218 structure slot setters are implemented as closures to save space,
219 so the compiler doesn't compile the type test into code, but
220 instead just saves the type in a lexical closure and interprets it
222 To exercise the problem, compile and load
223 (cl:in-package :cl-user)
225 (bar (error "missing") :type bar))
228 (loop (setf (foo-bar *foo*) x)))
230 (defvar *bar* (make-bar))
231 (defvar *foo* (make-foo :bar *bar*))
232 (defvar *setf-foo-bar* #'(setf foo-bar))
234 (loop (funcall *setf-foo-bar* x *foo*)))
235 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
236 use BACKTRACE, to see it's spending all essentially all its time
237 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
238 One possible solution would be simply to give up on
239 representing structure slot accessors as functions, and represent
240 them as macroexpansions instead. This can be inconvenient for users,
241 but it's not clear that it's worse than trying to help by expanding
242 into a horribly inefficient implementation.
243 As a workaround for the problem, #'(SETF FOO) expressions
244 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
245 (defmacro efficient-setf-function (place-function-name)
246 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
247 ;; a workaround for the problem, encouraging the
248 ;; inline expansion of the structure accessor, so
249 ;; that the compiler can optimize its type test
250 (let ((new-value (gensym "NEW-VALUE-"))
251 (structure-value (gensym "STRUCTURE-VALUE-")))
252 `(lambda (,new-value ,structure-value)
253 (setf (,place-function-name ,structure-value)
255 ;; no problem, can just use the ordinary expansion
256 `(function (setf ,place-function-name))))
259 There's apparently a bug in CEILING optimization which caused
260 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
261 applied the patches to SBCL and they didn't seem to cause problems
262 (as reported sbcl-devel 2001-05-04). However, since the patches
263 modify nontrivial code which was apparently written incorrectly
264 the first time around, until regression tests are written I'm not
265 comfortable merging the patches in the CVS version of SBCL.
268 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
269 a clean, just-started SBCL system. And it seems to be right:
270 (ROOM T) can bring a small computer to its knees for a *long*
271 time trying to GC afterwards. Surely there's some more economical
272 way to implement (ROOM T).
274 Daniel Barlow doesn't know what fixed this, but observes that it
275 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
276 in a fresh SBCL causes
278 debugger invoked on a SB-INT:BUG in thread 5911:
279 failed AVER: "(SAP= CURRENT END)"
281 unless a GC has happened beforehand.
284 When the compiler inline expands functions, it may be that different
285 kinds of return values are generated from different code branches.
286 E.g. an inline expansion of POSITION generates integer results
287 from one branch, and NIL results from another. When that inline
288 expansion is used in a context where only one of those results
291 (aref *a1* (position x *a2*)))
292 and the compiler can't prove that the unacceptable branch is
293 never taken, then bogus type mismatch warnings can be generated.
294 If you need to suppress the type mismatch warnings, you can
295 suppress the inline expansion,
297 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
298 (aref *a1* (position x *a2*)))
299 or, sometimes, suppress them by declaring the result to be of an
302 (aref *a1* (the integer (position x *a2*))))
304 This is not a new compiler problem in 0.7.0, but the new compiler
305 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
306 more conspicuous. If you don't need performance from these functions,
307 and the bogus warnings are a nuisance for you, you can return to
308 your pre-0.7.0 state of grace with
309 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
314 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
315 the entire lexical environment at the point of MACROLET available
316 in the bodies of the macroexpander functions. In particular, it
317 allows the function bodies (which run at compile time) to try to
318 access lexical variables (which are only defined at runtime).
319 It doesn't even issue a warning, which is bad.
321 The SBCL behavior arguably conforms to the ANSI spec (since the
322 spec says that the behavior is undefined, ergo anything conforms).
323 However, it would be better to issue a compile-time error.
324 Unfortunately I (WHN) don't see any simple way to detect this
325 condition in order to issue such an error, so for the meantime
326 SBCL just does this weird broken "conforming" thing.
328 The ANSI standard says, in the definition of the special operator
330 The macro-expansion functions defined by MACROLET are defined
331 in the lexical environment in which the MACROLET form appears.
332 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
333 the local macro definitions in a MACROLET, but the consequences
334 are undefined if the local macro definitions reference any
335 local variable or function bindings that are visible in that
337 Then it seems to contradict itself by giving the example
339 (macrolet ((fudge (z)
340 ;The parameters x and flag are not accessible
341 ; at this point; a reference to flag would be to
342 ; the global variable of that name.
343 ` (if flag (* ,z ,z) ,z)))
344 ;The parameters x and flag are accessible here.
348 The comment "a reference to flag would be to the global variable
349 of the same name" sounds like good behavior for the system to have.
350 but actual specification quoted above says that the actual behavior
353 (Since 0.7.8.23 macroexpanders are defined in a restricted version
354 of the lexical environment, containing no lexical variables and
355 functions, which seems to conform to ANSI and CLtL2, but signalling
356 a STYLE-WARNING for references to variables similar to locals might
360 Ideally, uninterning a symbol would allow it, and its associated
361 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
362 at least as of sbcl-0.7.0, this isn't the case. Information about
363 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
364 essentially in ordinary (non-weak) hash tables keyed by symbols.
365 Thus, once a system has an entry in this system, it tends to live
366 forever, even when it is uninterned and all other references to it
370 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
371 prominent SourceForge web/db bug tracking system, which is
372 unfortunately not a reliable way to get a timely response from
373 the SBCL maintainers)
374 In the course of trying to build a test case for an
375 application error, I encountered this behavior:
376 If you start up sbcl, and then lay on CTRL-C for a
377 minute or two, the lisp process will eventually say:
378 %PRIMITIVE HALT called; the party is over.
379 and throw you into the monitor. If I start up lisp,
380 attach to the process with strace, and then do the same
381 (abusive) thing, I get instead:
382 access failure in heap page not marked as write-protected
383 and the monitor again. I don't know enough to have the
384 faintest idea of what is going on here.
385 This is with sbcl 6.12, uname -a reports:
386 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
387 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
388 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
389 it took more than two minutes (but less than five) for me.
393 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
394 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
395 COMPLEX implementation didn't deal with this, and hasn't been
396 upgraded to do so. (This doesn't seem to be a high priority
397 conformance problem, since seems hard to construct useful code
400 [ partially fixed by CSR in 0.8.17.17 because of a PFD ansi-tests
401 report that (COMPLEX RATIO) was failing; still failing on types of
402 the form (AND NUMBER (SATISFIES REALP) (SATISFIES ZEROP)). ]
404 b. (fixed in 0.8.3.43)
407 Floating point errors are reported poorly. E.g. on x86 OpenBSD
410 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
411 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
412 No traps are enabled? How can this be?
413 It should be possible to be much more specific (overflow, division
414 by zero, etc.) and of course the "How can this be?" should be fixable.
416 See also bugs #45.c and #183
419 (reported by Robert E. Brown 2002-04-16)
420 When a function is called with too few arguments, causing the
421 debugger to be entered, the uninitialized slots in the bad call frame
422 seem to cause GCish problems, being interpreted as tagged data even
423 though they're not. In particular, executing ROOM in the
424 debugger at that point causes AVER failures:
427 * (lisp-implementation-version)
433 failed AVER: "(SAP= CURRENT END)"
434 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
435 isn't too surprising since there are many differences in stack
436 implementation and GC conservatism between the X86 and other ports.)
438 This is probably the same bug as 216
441 The compiler sometimes tries to constant-fold expressions before
442 it checks to see whether they can be reached. This can lead to
443 bogus warnings about errors in the constant folding, e.g. in code
446 (WRITE-STRING (> X 0) "+" "0"))
447 compiled in a context where the compiler can prove that X is NIL,
448 and the compiler complains that (> X 0) causes a type error because
449 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
450 caused a full WARNING, which made the bug really annoying because then
451 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
452 code. Since then the warning has been downgraded to STYLE-WARNING,
453 so it's still a bug but at least it's a little less annoying.
455 183: "IEEE floating point issues"
456 Even where floating point handling is being dealt with relatively
457 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
458 accrued-exceptions and current-exceptions part of the fp control
459 word don't seem to bear much relation to reality. E.g. on
463 debugger invoked on condition of type DIVISION-BY-ZERO:
464 arithmetic error DIVISION-BY-ZERO signalled
465 0] (sb-vm::get-floating-point-modes)
467 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
468 :ROUNDING-MODE :NEAREST
469 :CURRENT-EXCEPTIONS NIL
470 :ACCRUED-EXCEPTIONS (:INEXACT)
473 * (sb-vm::get-floating-point-modes)
474 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
475 :ROUNDING-MODE :NEAREST
476 :CURRENT-EXCEPTIONS (:INEXACT)
477 :ACCRUED-EXCEPTIONS (:INEXACT)
480 188: "compiler performance fiasco involving type inference and UNION-TYPE"
484 (declare (optimize (safety 3)))
485 (declare (optimize (compilation-speed 2)))
486 (declare (optimize (speed 1) (debug 1) (space 1)))
488 (declare (type (integer 0) start))
489 (print (incf start 22))
490 (print (incf start 26))
491 (print (incf start 28)))
493 (declare (type (integer 0) start))
494 (print (incf start 22))
495 (print (incf start 26)))
497 (declare (type (integer 0) start))
498 (print (incf start 22))
499 (print (incf start 26))))))
501 This example could be solved with clever enough constraint
502 propagation or with SSA, but consider
507 The careful type of X is {2k} :-(. Is it really important to be
508 able to work with unions of many intervals?
510 191: "Miscellaneous PCL deficiencies"
511 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
512 a. DEFCLASS does not inform the compiler about generated
513 functions. Compiling a file with
517 (WITH-SLOTS (A-CLASS-X) A
519 results in a STYLE-WARNING:
521 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
523 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
524 Mai points out that the declamation of functions is in fact
525 incorrect in some cases (most notably for structure
526 classes). This means that at present erroneous attempts to use
527 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
528 won't get the corresponding STYLE-WARNING.
529 c. (fixed in 0.8.4.23)
531 201: "Incautious type inference from compound types"
532 a. (reported by APD sbcl-devel 2002-09-17)
534 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
536 (FORMAT NIL "~S IS ~S, Y = ~S"
543 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
547 (declare (type (array * (4 4)) x))
549 (setq x (make-array '(4 4)))
550 (adjust-array y '(3 5))
551 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
553 * (foo (make-array '(4 4) :adjustable t))
556 205: "environment issues in cross compiler"
557 (These bugs have no impact on user code, but should be fixed or
559 a. Macroexpanders introduced with MACROLET are defined in the null
561 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
562 the null lexical environment.
563 c. The cross-compiler cannot inline functions defined in a non-null
566 206: ":SB-FLUID feature broken"
567 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
568 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
571 207: "poorly distributed SXHASH results for compound data"
572 SBCL's SXHASH could probably try a little harder. ANSI: "the
573 intent is that an implementation should make a good-faith
574 effort to produce hash-codes that are well distributed
575 within the range of non-negative fixnums". But
576 (let ((hits (make-hash-table)))
579 (let* ((ij (cons i j))
580 (newlist (push ij (gethash (sxhash ij) hits))))
582 (format t "~&collision: ~S~%" newlist))))))
583 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
584 would be an obvious way of fix. Maybe it would be acceptably efficient
585 to redo MIX using a lookup into a 256-entry s-box containing
586 29-bit pseudorandom numbers?
588 211: "keywords processing"
589 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
590 number of keyword arguments.
592 212: "Sequence functions and circular arguments"
593 COERCE, MERGE and CONCATENATE go into an infinite loop when given
594 circular arguments; it would be good for the user if they could be
595 given an error instead (ANSI 17.1.1 allows this behaviour on the part
596 of the implementation, as conforming code cannot give non-proper
597 sequences to these functions. MAP also has this problem (and
598 solution), though arguably the convenience of being able to do
599 (MAP 'LIST '+ FOO '#1=(1 . #1#))
600 might be classed as more important (though signalling an error when
601 all of the arguments are circular is probably desireable).
603 213: "Sequence functions and type checking"
604 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
605 check that it will return a sequence of the given type. Fixing
606 it along the same lines as the others (cf. work done around
607 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
608 entirely straightforward.
609 c. All of these functions will silently accept a type of the form
611 whether or not the return value is of this type. This is
612 probably permitted by ANSI (see "Exceptional Situations" under
613 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
614 know about this escape clause, so code of the form
615 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
616 can erroneously return T.
618 215: ":TEST-NOT handling by functions"
619 a. FIND and POSITION currently signal errors when given non-NIL for
620 both their :TEST and (deprecated) :TEST-NOT arguments, but by
621 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
622 means that the effect is "unpredictable but harmless". It's not
623 clear what that actually means; it may preclude conforming
624 implementations from signalling errors.
625 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
626 when conflict occurs. As a quality of implementation issue, it
627 might be preferable to treat :TEST and :TEST-NOT as being in some
628 sense the same &KEY, and effectively take the first test function in
630 c. Again, a quality of implementation issue: it would be good to issue a
631 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
632 WARNING for calls with both :TEST and :TEST-NOT; possibly this
633 latter should be WARNed about at execute-time too.
635 216: "debugger confused by frames with invalid number of arguments"
636 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
637 leaves the system confused, enough so that (QUIT) no longer works.
638 It's as though the process of working with the uninitialized slot in
639 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
640 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
642 This is probably the same bug as 162
644 217: "Bad type operations with FUNCTION types"
647 * (values-type-union (specifier-type '(function (base-char)))
648 (specifier-type '(function (integer))))
650 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
652 It causes insertion of wrong type assertions into generated
656 (let ((f (etypecase x
657 (character #'write-char)
658 (integer #'write-byte))))
661 (character (write-char x s))
662 (integer (write-byte x s)))))
664 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
666 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
667 produce invalid code, but type checking is not accurate.)
669 233: bugs in constraint propagation
671 (declaim (optimize (speed 2) (safety 3)))
673 (if (typep (prog1 x (setq x y)) 'double-float)
676 (foo 1d0 5) => segmentation violation
678 235: "type system and inline expansion"
680 (declaim (ftype (function (cons) number) acc))
681 (declaim (inline acc))
683 (the number (car c)))
686 (values (locally (declare (optimize (safety 0)))
688 (locally (declare (optimize (safety 3)))
691 (foo '(nil) '(t)) => NIL, T.
693 237: "Environment arguments to type functions"
694 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
695 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
696 argument, but they ignore it completely. This is almost
697 certainly not correct.
698 b. Also, the compiler's optimizers for TYPEP have not been informed
699 about the new argument; consequently, they will not transform
700 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
701 just as optimizeable as (TYPEP 1 'INTEGER).
703 238: "REPL compiler overenthusiasm for CLOS code"
705 * (defclass foo () ())
706 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
707 causes approximately 100 lines of code deletion notes. Some
708 discussion on this issue happened under the title 'Three "interesting"
709 bugs in PCL', resulting in a fix for this oververbosity from the
710 compiler proper; however, the problem persists in the interactor
711 because the notion of original source is not preserved: for the
712 compiler, the original source of the above expression is (DEFMETHOD
713 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
714 compiler gets its hands on the code needing compilation from the REPL,
715 it has been macroexpanded several times.
717 A symptom of the same underlying problem, reported by Tony Martinez:
719 (with-input-from-string (*query-io* " no")
721 (simple-type-error () 'error))
723 ; (SB-KERNEL:FLOAT-WAIT)
725 ; note: deleting unreachable code
726 ; compilation unit finished
729 242: "WRITE-SEQUENCE suboptimality"
730 (observed from clx performance)
731 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
732 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
733 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
734 rather than writing the sequence in one go, leading to severe
735 performance degradation.
736 As of sbcl-0.9.0.36, this is solved for fd-streams, so is less of a
737 problem in practice. (Fully fixing this would require adding a
738 ansi-stream-n-bout slot and associated methods to write a byte
739 sequence to ansi-stream, similar to the existing ansi-stream-sout
742 243: "STYLE-WARNING overenthusiasm for unused variables"
743 (observed from clx compilation)
744 In sbcl-0.7.14, in the presence of the macros
745 (DEFMACRO FOO (X) `(BAR ,X))
746 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
747 somewhat surprising style warnings are emitted for
748 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
750 ; (LAMBDA (Y) (FOO Y))
752 ; caught STYLE-WARNING:
753 ; The variable Y is defined but never used.
755 245: bugs in disassembler
756 b. On X86 operand size prefix is not recognized.
759 (defun foo (&key (a :x))
763 does not cause a warning. (BTW: old SBCL issued a warning, but for a
764 function, which was never called!)
767 Compiler does not emit warnings for
769 a. (lambda () (svref (make-array 8 :adjustable t) 1))
772 (list (let ((y (the real x)))
773 (unless (floatp y) (error ""))
778 (declare (optimize (debug 0)))
779 (declare (type vector x))
780 (list (fill-pointer x)
784 Complex array type does not have corresponding type specifier.
786 This is a problem because the compiler emits optimization notes when
787 you use a non-simple array, and without a type specifier for hairy
788 array types, there's no good way to tell it you're doing it
789 intentionally so that it should shut up and just compile the code.
791 Another problem is confusing error message "asserted type ARRAY
792 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
793 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
795 The last problem is that when type assertions are converted to type
796 checks, types are represented with type specifiers, so we could lose
797 complex attribute. (Now this is probably not important, because
798 currently checks for complex arrays seem to be performed by
802 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
803 warning. Analogous cases with the index and length being equal and
804 greater than 0 are warned for; the problem here seems to be that the
805 type required for an array reference of this type is (INTEGER 0 (0))
806 which is canonicalized to NIL.
811 (t1 (specifier-type s)))
812 (eval `(defstruct ,s))
813 (type= t1 (specifier-type s)))
818 b. The same for CSUBTYPEP.
820 262: "yet another bug in inline expansion of local functions"
821 During inline expansion of a local function Python can try to
822 reference optimized away objects (functions, variables, CTRANs from
823 tags and blocks), which later may lead to problems. Some of the
824 cases are worked around by forbidding expansion in such cases, but
825 the better way would be to reimplement inline expansion by copying
829 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
830 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
831 looks reasonable, if not obviously correct; however, it caused the
832 PPC/Linux port to segfault during warm-init while loading
833 src/pcl/std-class.fasl. A workaround patch was made, but it would
834 be nice to understand why the first patch caused problems, and to
835 fix the cause if possible.
837 268: "wrong free declaration scope"
838 The following code must signal type error:
840 (locally (declare (optimize (safety 3)))
841 (flet ((foo (x &optional (y (car x)))
842 (declare (optimize (safety 0)))
844 (funcall (eval #'foo) 1)))
847 In the following function constraint propagator optimizes nothing:
850 (declare (integer x))
851 (declare (optimize speed))
859 Compilation of the following two forms causes "X is unbound" error:
861 (symbol-macrolet ((x pi))
862 (macrolet ((foo (y) (+ x y)))
863 (declaim (inline bar))
869 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
872 CLHS says that type declaration of a symbol macro should not affect
873 its expansion, but in SBCL it does. (If you like magic and want to
874 fix it, don't forget to change all uses of MACROEXPAND to
878 The following code (taken from CLOCC) takes a lot of time to compile:
881 (declare (type (integer 0 #.large-constant) n))
884 (fixed in 0.8.2.51, but a test case would be good)
887 b. The same as in a., but using MULTIPLE-VALUE-SETQ instead of SETQ.
889 (defmethod faa ((*faa* double-float))
890 (set '*faa* (when (< *faa* 0) (- *faa*)))
892 (faa 1d0) => type error
897 (declare (optimize speed))
898 (loop for i of-type (integer 0) from 0 by 2 below 10
901 uses generic arithmetic.
903 b. (fixed in 0.8.3.6)
905 279: type propagation error -- correctly inferred type goes astray?
906 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
907 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
908 &OPTIONAL), not a (INTEGER 1 536870911)
909 is emitted when compiling this file:
910 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
911 (integer #.most-negative-fixnum 0))
916 (let* (;; Uncomment this for a type mismatch warning indicating
917 ;; that the type of (FOO X) is correctly understood.
918 #+nil (fs-foo (float-sign (foo x)))
919 ;; Uncomment this for a type mismatch warning
920 ;; indicating that the type of (ABS (FOO X)) is
921 ;; correctly understood.
922 #+nil (fs-abs-foo (float-sign (abs (foo x))))
923 ;; something wrong with this one though
924 (abs-foo (abs (foo x))))
925 (declare (type (integer 1 100) abs-foo))
930 283: Thread safety: libc functions
931 There are places that we call unsafe-for-threading libc functions
932 that we should find alternatives for, or put locks around. Known or
933 strongly suspected problems, as of 0.8.3.10: please update this
934 bug instead of creating new ones
936 localtime() - called for timezone calculations in code/time.lisp
938 284: Thread safety: special variables
939 There are lots of special variables in SBCL, and I feel sure that at
940 least some of them are indicative of potentially thread-unsafe
941 parts of the system. See doc/internals/notes/threading-specials
943 286: "recursive known functions"
944 Self-call recognition conflicts with known function
945 recognition. Currently cross compiler and target COMPILE do not
946 recognize recursion, and in target compiler it can be disabled. We
947 can always disable it for known functions with RECURSIVE attribute,
948 but there remains a possibility of a function with a
949 (tail)-recursive simplification pass and transforms/VOPs for base
952 287: PPC/Linux miscompilation or corruption in first GC
953 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
954 segmentation fault is reported at the point of first triggered GC,
955 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
956 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
957 that this merely obscures, not solves, the underlying problem; as and when
958 underlying problems are fixed, it would be worth trying again to provoke
961 288: fundamental cross-compilation issues (from old UGLINESS file)
962 Using host floating point numbers to represent target floating point
963 numbers, or host characters to represent target characters, is
964 theoretically shaky. (The characters are OK as long as the characters
965 are in the ANSI-guaranteed character set, though, so they aren't a
966 real problem as long as the sources don't need anything but that;
967 the floats are a real problem.)
969 289: "type checking and source-transforms"
971 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
974 Our policy is to check argument types at the moment of a call. It
975 disagrees with ANSI, which says that type assertions are put
976 immediately onto argument expressions, but is easier to implement in
977 IR1 and is more compatible to type inference, inline expansion,
978 etc. IR1-transforms automatically keep this policy, but source
979 transforms for associative functions (such as +), being applied
980 during IR1-convertion, do not. It may be tolerable for direct calls
981 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
983 b. Another aspect of this problem is efficiency. [x y + z +]
984 requires less registers than [x y z + +]. This transformation is
985 currently performed with source transforms, but it would be good to
986 also perform it in IR1 optimization phase.
988 290: Alpha floating point and denormalized traps
989 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
990 hardware or kernel deficiency: the status of the enable/disable
991 denormalized-float traps bit seems to be ambiguous; by the time we
992 get to os_restore_fp_control after a trap, denormalized traps seem
993 to be enabled. Since we don't want a trap every time someone uses a
994 denormalized float, in general, we mask out that bit when we restore
995 the control word; however, this clobbers any change the user might
999 (reported by Adam Warner, sbcl-devel 2003-09-23)
1001 The --load toplevel argument does not perform any sanitization of its
1002 argument. As a result, files with Lisp pathname pattern characters
1003 (#\* or #\?, for instance) or quotation marks can cause the system
1004 to perform arbitrary behaviour.
1007 LOOP with non-constant arithmetic step clauses suffers from overzealous
1008 type constraint: code of the form
1009 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1010 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1011 (0))). However, an integral value of X should be legal, because
1012 successive adds of integers to double-floats produces double-floats,
1013 so none of the type restrictions in the code is violated.
1015 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1016 argument type only after having read a character. This is caused
1017 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1018 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1019 less error prone to have EXPLICIT-CHECK be a local declaration,
1020 being put into the definition, instead of an attribute being kept in
1021 a separate file; maybe also put it into SB-EXT?
1023 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1024 in NOTE-ASSUMED-TYPES
1025 In sbcl-0.8.7.32, compiling the file
1027 (declare (type integer x))
1028 (declare (type (vector (or hash-table bit)) y))
1031 (declare (type integer x))
1032 (declare (type (simple-array base (2)) y))
1035 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1037 303: "nonlinear LVARs" (aka MISC.293)
1039 (multiple-value-call #'list
1041 (multiple-value-prog1
1042 (eval '(values :a :b :c))
1048 (throw 'bar (values 3 4)))))))))))
1050 (BUU 1) returns garbage.
1052 The problem is that both EVALs sequentially write to the same LVAR.
1054 306: "Imprecise unions of array types"
1056 (declare (optimize speed)
1057 (type (or (array cons) (array vector)) x))
1059 (foo #((0))) => TYPE-ERROR
1066 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1067 collect `(array ,(sb-vm:saetp-specifier x)))))
1068 => NIL, T (when it should be T, T)
1070 309: "Dubious values for implementation limits"
1071 (reported by Bruno Haible sbcl-devel "Incorrect value of
1072 multiple-values-limit" 2004-04-19)
1073 (values-list (make-list 1000000)), on x86/linux, signals a stack
1074 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1075 significantly larger than 1000000. There are probably similar
1076 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1077 around the same time regarding a call to LIST on sparc with 1000
1078 arguments) and other implementation limit constants.
1080 314: "LOOP :INITIALLY clauses and scope of initializers"
1081 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1082 test suite, originally by Thomas F. Burdick.
1083 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1084 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1085 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1086 ;; initially clauses are evaluated in the loop prologue, which precedes all
1087 ;; loop code except for the initial settings provided by with, for, or as.
1088 (loop :for x = 0 :then (1+ x)
1089 :for y = (1+ x) :then (ash y 1)
1090 :for z :across #(1 3 9 27 81 243)
1092 :initially (assert (zerop x)) :initially (assert (= 2 w))
1093 :until (>= w 100) :collect w)
1094 Expected: (2 6 15 38)
1097 318: "stack overflow in compiler warning with redefined class"
1098 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1101 (setf (find-class 'foo) nil)
1102 (defstruct foo slot-1)
1103 This used to give a stack overflow from within the printer, which has
1104 been fixed as of 0.8.16.11. Current result:
1106 ; can't compile TYPEP of anonymous or undefined class:
1107 ; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
1109 debugger invoked on a TYPE-ERROR in thread 19973:
1110 The value NIL is not of type FUNCTION.
1112 CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
1113 meant to be enough to delete structure classes from the system?
1115 319: "backquote with comma inside array"
1116 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1118 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1120 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1121 which probably isn't intentional.
1123 323: "REPLACE, BIT-BASH and large strings"
1124 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1125 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1127 (declare (optimize speed (safety 1)))
1128 (let ((x (make-string 140000000))
1129 (y (make-string 140000000)))
1130 (length (replace x y))))
1133 debugger invoked on a TYPE-ERROR in thread 2412:
1134 The value 1120000000 is not of type (MOD 536870911).
1135 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1137 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1138 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1139 for all positive integral <x>. At present, it only works for <x> up
1140 to about 1024 (and similarly for signed-byte), so
1141 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1142 gives an error in sbcl-0.8.10.
1144 325: "CLOSE :ABORT T on supeseding streams"
1145 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1146 file on disk, even if one existed before opening.
1148 The illegality of this is not crystal clear, as the ANSI dictionary
1149 entry for CLOSE says that when :ABORT is T superseded files are not
1150 superseded (ie. the original should be restored), whereas the OPEN
1151 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1152 implementation should not destroy the old file until the new stream
1153 is closed." -- implying that even though undesirable, early deletion
1154 is legal. Restoring the original would none the less be the polite
1157 326: "*PRINT-CIRCLE* crosstalk between streams"
1158 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1159 mixed between streams when output operations are intermingled closely
1160 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1161 test case below), so that e.g. the references #2# appears on a stream
1162 with no preceding #2= on that stream to define it (because the #2= was
1163 sent to another stream).
1164 (cl:in-package :cl-user)
1165 (defstruct foo index)
1166 (defparameter *foo* (make-foo :index 4))
1168 (defparameter *bar* (make-bar))
1169 (defparameter *tangle* (list *foo* *bar* *foo*))
1170 (defmethod print-object ((foo foo) stream)
1171 (let ((index (foo-index foo)))
1172 (format *trace-output*
1173 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1175 (format stream "[FOO ~D]" index))
1177 (let ((tsos (make-string-output-stream))
1178 (ssos (make-string-output-stream)))
1179 (let ((*print-circle* t)
1180 (*trace-output* tsos)
1181 (*standard-output* ssos))
1182 (prin1 *tangle* *standard-output*))
1183 (let ((string (get-output-stream-string ssos)))
1184 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1185 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1186 (error "oops: ~S" string)))))
1187 It might be straightforward to fix this by turning the
1188 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1189 per-stream slots, but (1) it would probably be sort of messy faking
1190 up the special variable binding semantics using UNWIND-PROTECT and
1191 (2) it might be sort of a pain to test that no other bugs had been
1194 328: "Profiling generic functions", transplanted from #241
1195 (from tonyms on #lisp IRC 2003-02-25)
1196 In sbcl-0.7.12.55, typing
1197 (defclass foo () ((bar :accessor foo-bar)))
1200 (defclass foo () ((bar :accessor foo-bar)))
1201 gives the error message
1202 "#:FOO-BAR already names an ordinary function or a macro."
1204 Problem: when a generic function is profiled, it appears as an ordinary
1205 function to PCL. (Remembering the uninterned accessor is OK, as the
1206 redefinition must be able to remove old accessors from their generic
1209 329: "Sequential class redefinition"
1210 reported by Bruno Haible:
1211 (defclass reactor () ((max-temp :initform 10000000)))
1212 (defvar *r1* (make-instance 'reactor))
1213 (defvar *r2* (make-instance 'reactor))
1214 (slot-value *r1* 'max-temp)
1215 (slot-value *r2* 'max-temp)
1216 (defclass reactor () ((uptime :initform 0)))
1217 (slot-value *r1* 'uptime)
1218 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1219 (slot-value *r1* 'max-temp) ; => 10000
1220 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1223 The method effective when the wrapper is obsoleted can be saved
1224 in the wrapper, and then to update the instance just run through
1225 all the old wrappers in order from oldest to newest.
1227 332: "fasl stack inconsistency in structure redefinition"
1228 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1229 Even though structure redefinition is undefined by the standard, the
1230 following behaviour is suboptimal: running
1231 (defun stimulate-sbcl ()
1232 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1233 ;;create a file which redefines a structure incompatibly
1234 (with-open-file (f filename :direction :output :if-exists :supersede)
1235 (print '(defstruct astruct foo) f)
1236 (print '(defstruct astruct foo bar) f))
1237 ;;compile and load the file, then invoke the continue restart on
1238 ;;the structure redefinition error
1239 (handler-bind ((error (lambda (c) (continue c))))
1240 (load (compile-file filename)))))
1242 and choosing the CONTINUE restart yields the message
1243 debugger invoked on a SB-INT:BUG in thread 27726:
1244 fasl stack not empty when it should be
1246 336: "slot-definitions must retain the generic functions of accessors"
1247 reported by Tony Martinez:
1248 (defclass foo () ((bar :reader foo-bar)))
1249 (defun foo-bar (x) x)
1250 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1252 Note: just punting the accessor removal if the fdefinition
1253 is not a generic function is not enough:
1255 (defclass foo () ((bar :reader foo-bar)))
1256 (defvar *reader* #'foo-bar)
1257 (defun foo-bar (x) x)
1258 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1259 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1260 ; the method must be removed
1261 ; by the class redefinition
1263 Fixing this should also fix a subset of #328 -- update the
1264 description with a new test-case then.
1266 337: MAKE-METHOD and user-defined method classes
1267 (reported by Bruno Haible sbcl-devel 2004-06-11)
1271 (defclass user-method (standard-method) (myslot))
1272 (defmacro def-user-method (name &rest rest)
1273 (let* ((lambdalist-position (position-if #'listp rest))
1274 (qualifiers (subseq rest 0 lambdalist-position))
1275 (lambdalist (elt rest lambdalist-position))
1276 (body (subseq rest (+ lambdalist-position 1)))
1278 (subseq lambdalist 0 (or
1280 (lambda (x) (member x lambda-list-keywords))
1282 (length lambdalist))))
1283 (specializers (mapcar #'find-class
1284 (mapcar (lambda (x) (if (consp x) (second x) t))
1286 (unspecialized-required-part
1287 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1288 (unspecialized-lambdalist
1289 (append unspecialized-required-part
1290 (subseq lambdalist (length required-part)))))
1293 (MAKE-INSTANCE 'USER-METHOD
1294 :QUALIFIERS ',qualifiers
1295 :LAMBDA-LIST ',unspecialized-lambdalist
1296 :SPECIALIZERS ',specializers
1298 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1299 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1300 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1301 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1302 (IF (NULL NEXT-METHODS-LIST)
1303 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1304 (FUNCALL (SB-PCL:METHOD-FUNCTION
1305 (FIRST NEXT-METHODS-LIST))
1306 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1307 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1311 (defgeneric test-um03 (x))
1312 (defmethod test-um03 ((x integer))
1313 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1314 (def-user-method test-um03 ((x rational))
1315 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1316 (defmethod test-um03 ((x real))
1317 (list 'real x (not (null (next-method-p)))))
1322 (defgeneric test-um10 (x))
1323 (defmethod test-um10 ((x integer))
1324 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1325 (defmethod test-um10 ((x rational))
1326 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1327 (defmethod test-um10 ((x real))
1328 (list 'real x (not (null (next-method-p)))))
1329 (defmethod test-um10 :after ((x real)))
1330 (def-user-method test-um10 :around ((x integer))
1331 (list* 'around-integer x
1332 (not (null (next-method-p))) (call-next-method)))
1333 (defmethod test-um10 :around ((x rational))
1334 (list* 'around-rational x
1335 (not (null (next-method-p))) (call-next-method)))
1336 (defmethod test-um10 :around ((x real))
1337 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1339 fails with a type error, and
1342 (defgeneric test-um12 (x))
1343 (defmethod test-um12 ((x integer))
1344 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1345 (defmethod test-um12 ((x rational))
1346 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1347 (defmethod test-um12 ((x real))
1348 (list 'real x (not (null (next-method-p)))))
1349 (defmethod test-um12 :after ((x real)))
1350 (defmethod test-um12 :around ((x integer))
1351 (list* 'around-integer x
1352 (not (null (next-method-p))) (call-next-method)))
1353 (defmethod test-um12 :around ((x rational))
1354 (list* 'around-rational x
1355 (not (null (next-method-p))) (call-next-method)))
1356 (def-user-method test-um12 :around ((x real))
1357 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1359 fails with NO-APPLICABLE-METHOD.
1361 339: "DEFINE-METHOD-COMBINATION bugs"
1362 (reported by Bruno Haible via the clisp test suite)
1364 a. Syntax checking laxity (should produce errors):
1365 i. (define-method-combination foo :documentation :operator)
1366 ii. (define-method-combination foo :documentation nil)
1367 iii. (define-method-combination foo nil)
1368 iv. (define-method-combination foo nil nil
1369 (:arguments order &aux &key))
1370 v. (define-method-combination foo nil nil (:arguments &whole))
1371 vi. (define-method-combination foo nil nil (:generic-function))
1372 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1373 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1374 ix. (define-method-combination foo nil ((3)))
1375 x. (define-method-combination foo nil ((a)))
1377 b. define-method-combination arguments lambda list badness
1378 i. &aux args are currently unsupported;
1379 ii. default values of &optional and &key arguments are ignored;
1380 iii. supplied-p variables for &optional and &key arguments are not
1383 c. qualifier matching incorrect
1385 (define-method-combination mc27 ()
1387 (ignored (:ignore :unused)))
1389 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1390 (defgeneric test-mc27 (x)
1391 (:method-combination mc27)
1392 (:method :ignore ((x number)) (/ 0)))
1395 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1396 method is applicable, and yet matches neither of the method group
1399 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1400 Even the simplest possible overriding of
1401 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1402 as "canonical", does not work:
1403 (defclass my-generic-function (standard-generic-function) ()
1404 (:metaclass funcallable-standard-class))
1405 (defmethod compute-discriminating-function ((gf my-generic-function))
1406 (let ((dfun (call-next-method)))
1407 (lambda (&rest args)
1408 (apply dfun args))))
1410 (:generic-function-class my-generic-function))
1411 (defmethod foo (x) (+ x x))
1413 signals an error. This error is the same even if the LAMBDA is
1414 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1415 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1416 src/code/target-defstruct.lisp is broken? This seems to be broken
1417 in CMUCL 18e, so it's not caused by a recent change.
1419 344: more (?) ROOM T problems (possibly part of bug 108)
1420 In sbcl-0.8.12.51, and off and on leading up to it, the
1421 SB!VM:MEMORY-USAGE operations in ROOM T caused
1422 unhandled condition (of type SB-INT:BUG):
1423 failed AVER: "(SAP= CURRENT END)"
1424 Several clever people have taken a shot at this without fixing
1425 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1426 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1427 out how to make them work reliably with the rest of the GC.
1429 (Note: there's at least one dubious thing in room.lisp: see the
1430 comment in VALID-OBJ)
1432 346: alpha backtrace
1433 In sbcl-0.8.13, all backtraces from errors caused by internal errors
1434 on the alpha seem to have a "bogus stack frame".
1436 349: PPRINT-INDENT rounding implementation decisions
1437 At present, pprint-indent (and indeed the whole pretty printer)
1438 more-or-less assumes that it's using a monospace font. That's
1439 probably not too silly an assumption, but one piece of information
1440 the current implementation loses is from requests to indent by a
1441 non-integral amount. As of sbcl-0.8.15.9, the system silently
1442 truncates the indentation to an integer at the point of request, but
1443 maybe the non-integral value should be propagated through the
1444 pprinter and only truncated at output? (So that indenting by 1/2
1445 then 3/2 would indent by two spaces, not one?)
1447 352: forward-referenced-class trouble
1448 reported by Bruno Haible on sbcl-devel
1450 (setf (class-name (find-class 'a)) 'b)
1454 Expected: an instance of c, with a slot named x
1455 Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
1456 While computing the class precedence list of the class named C.
1457 The class named B is a forward referenced class.
1458 The class named B is a direct superclass of the class named C.
1460 353: debugger suboptimalities on x86
1461 On x86 backtraces for undefined functions start with a bogus stack
1462 frame, and backtraces for throws to unknown catch tags with a "no
1463 debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
1464 (used on non-x86 platforms) being a more complete solution then what
1467 On x86/linux large portions of tests/debug.impure.lisp have been commented
1468 out as failures. The probable culprit for these problems is in x86-call-context
1469 (things work fine on x86/freebsd).
1471 More generally, the debugger internals suffer from excessive x86/non-x86
1472 conditionalization and OAOOMization: refactoring the common parts would
1475 354: XEPs in backtraces
1476 Under default compilation policy
1480 Has the XEP for TEST in the backtrace, not the TEST frame itself.
1481 (sparc and x86 at least)
1483 Since SBCL 0.8.20.1 this is hidden unless *SHOW-ENTRY-POINT-DETAILS*
1484 is true (instead there appear two TEST frames at least on ppc). The
1485 underlying cause seems to be that SB-C::TAIL-ANNOTATE will not merge
1486 the tail-call for the XEP, since Python has by that time proved that
1487 the function can never return; same happens if the function holds an
1488 unconditional call to ERROR.
1490 355: change-class of generic-function
1491 (reported by Bruno Haible)
1492 The MOP doesn't support change-class on a generic-function. However, SBCL
1493 apparently supports it, since it doesn't give an error or warning when doing
1494 so so. Then, however, it produces wrong results for calls to this generic
1496 ;;; The effective-methods cache:
1498 (defgeneric testgf35 (x))
1499 (defmethod testgf35 ((x integer))
1500 (cons 'integer (if (next-method-p) (call-next-method))))
1501 (defmethod testgf35 ((x real))
1502 (cons 'real (if (next-method-p) (call-next-method))))
1503 (defclass customized5-generic-function (standard-generic-function)
1505 (:metaclass sb-pcl:funcallable-standard-class))
1506 (defmethod sb-pcl:compute-effective-method ((gf customized5-generic-function) method-combination methods)
1507 `(REVERSE ,(call-next-method)))
1511 (change-class #'testgf35 'customized5-generic-function)
1513 Expected: ((INTEGER REAL) (REAL INTEGER))
1514 Got: ((INTEGER REAL) (INTEGER REAL))
1515 ;;; The discriminating-function cache:
1517 (defgeneric testgf36 (x))
1518 (defmethod testgf36 ((x integer))
1519 (cons 'integer (if (next-method-p) (call-next-method))))
1520 (defmethod testgf36 ((x real))
1521 (cons 'real (if (next-method-p) (call-next-method))))
1522 (defclass customized6-generic-function (standard-generic-function)
1524 (:metaclass sb-pcl:funcallable-standard-class))
1525 (defmethod sb-pcl:compute-discriminating-function ((gf customized6-generic-function))
1526 (let ((orig-df (call-next-method)))
1527 #'(lambda (&rest arguments)
1528 (reverse (apply orig-df arguments)))))
1532 (change-class #'testgf36 'customized6-generic-function)
1534 Expected: ((INTEGER REAL) (REAL INTEGER))
1535 Got: ((INTEGER REAL) (INTEGER REAL))
1538 (reported by Bruno Haible)
1539 After the "layout depth conflict" error, the CLOS is left in a state where
1540 it's not possible to define new standard-class subclasses any more.
1542 (defclass prioritized-dispatcher ()
1543 ((dependents :type list :initform nil)))
1544 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1545 (c2 (eql (find-class 'prioritized-dispatcher))))
1547 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1549 (:metaclass sb-pcl:funcallable-standard-class))
1550 ;; ERROR, Quit the debugger with ABORT
1551 (defclass typechecking-reader-class (standard-class)
1553 Expected: #<STANDARD-CLASS TYPECHECKING-READER-CLASS>
1554 Got: ERROR "The assertion SB-PCL::WRAPPERS failed."
1556 357: defstruct inheritance of initforms
1557 (reported by Bruno Haible)
1558 When defstruct and defclass (with :metaclass structure-class) are mixed,
1559 1. some slot initforms are ignored by the DEFSTRUCT generated constructor
1561 2. all slot initforms are ignored by MAKE-INSTANCE. (This can be arguably
1562 OK for initforms that were given in a DEFSTRUCT form, but for those
1563 given in a DEFCLASS form, I think it qualifies as a bug.)
1565 (defstruct structure02a
1569 (defclass structure02b (structure02a)
1570 ((slot4 :initform -44)
1573 (slot7 :initform (floor (* pi pi)))
1574 (slot8 :initform 88))
1575 (:metaclass structure-class))
1576 (defstruct (structure02c (:include structure02b (slot8 -88)))
1579 (slot11 (floor (exp 3))))
1581 (let ((a (make-structure02c)))
1582 (list (structure02c-slot4 a)
1583 (structure02c-slot5 a)
1584 (structure02c-slot6 a)
1585 (structure02c-slot7 a)))
1586 Expected: (-44 nil t 9)
1587 Got: (SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..
1588 SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..)
1590 (let ((b (make-instance 'structure02c)))
1591 (list (structure02c-slot2 b)
1592 (structure02c-slot3 b)
1593 (structure02c-slot4 b)
1594 (structure02c-slot6 b)
1595 (structure02c-slot7 b)
1596 (structure02c-slot8 b)
1597 (structure02c-slot10 b)
1598 (structure02c-slot11 b)))
1599 Expected: (t 3 -44 t 9 -88 t 20)
1600 Got: (0 0 0 0 0 0 0 0)
1602 358: :DECLARE argument to ENSURE-GENERIC-FUNCTION
1603 (reported by Bruno Haible)
1604 According to ANSI CL, ensure-generic-function must accept a :DECLARE
1605 keyword argument. In SBCL 0.8.16 it does not.
1608 (ensure-generic-function 'foo113 :declare '((optimize (speed 3))))
1609 (sb-pcl:generic-function-declarations #'foo113))
1610 Expected: ((OPTIMIZE (SPEED 3)))
1612 Invalid initialization argument:
1614 in call for class #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>.
1616 The ANSI Standard, Section 7.1.2
1618 Bruno notes: The MOP specifies that ensure-generic-function accepts :DECLARATIONS.
1619 The easiest way to be compliant to both specs is to accept both (exclusively
1622 359: wrong default value for ensure-generic-function's :generic-function-class argument
1623 (reported by Bruno Haible)
1624 ANSI CL is silent on this, but the MOP's specification of ENSURE-GENERIC-FUNCTION says:
1625 "The remaining arguments are the complete set of keyword arguments
1626 received by ENSURE-GENERIC-FUNCTION."
1627 and the spec of ENSURE-GENERIC-FUNCTION-USING-CLASS:
1628 ":GENERIC-FUNCTION-CLASS - a class metaobject or a class name. If it is not
1629 supplied, it defaults to the class named STANDARD-GENERIC-FUNCTION."
1630 This is not the case in SBCL. Test case:
1631 (defclass my-generic-function (standard-generic-function)
1633 (:metaclass sb-pcl:funcallable-standard-class))
1634 (setf (fdefinition 'foo1)
1635 (make-instance 'my-generic-function :name 'foo1))
1636 (ensure-generic-function 'foo1
1637 :generic-function-class (find-class 'standard-generic-function))
1639 ; => #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1640 (setf (fdefinition 'foo2)
1641 (make-instance 'my-generic-function :name 'foo2))
1642 (ensure-generic-function 'foo2)
1644 Expected: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
1645 Got: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS MY-GENERIC-FUNCTION>
1647 360: CALL-METHOD not recognized in method-combination body
1648 (reported by Bruno Haible)
1649 This method combination, which adds 'redo' and 'return' restarts for each
1650 method invocation to standard method combination, gives an error in SBCL.
1651 (defun prompt-for-new-values ()
1652 (format *debug-io* "~&New values: ")
1653 (list (read *debug-io*)))
1654 (defun add-method-restarts (form method)
1655 (let ((block (gensym))
1663 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Try calling ~S again." ,method))
1666 :REPORT (LAMBDA (STREAM) (FORMAT STREAM "Specify return values for ~S call." ,method))
1667 :INTERACTIVE (LAMBDA () (PROMPT-FOR-NEW-VALUES))
1668 (RETURN-FROM ,block (VALUES-LIST L)))))))))
1669 (defun convert-effective-method (efm)
1671 (if (eq (car efm) 'CALL-METHOD)
1672 (let ((method-list (third efm)))
1673 (if (or (typep (first method-list) 'method) (rest method-list))
1674 ; Reduce the case of multiple methods to a single one.
1675 ; Make the call to the next-method explicit.
1676 (convert-effective-method
1677 `(CALL-METHOD ,(second efm)
1679 (CALL-METHOD ,(first method-list) ,(rest method-list))))))
1680 ; Now the case of at most one method.
1681 (if (typep (second efm) 'method)
1682 ; Wrap the method call in a RESTART-CASE.
1683 (add-method-restarts
1684 (cons (convert-effective-method (car efm))
1685 (convert-effective-method (cdr efm)))
1687 ; Normal recursive processing.
1688 (cons (convert-effective-method (car efm))
1689 (convert-effective-method (cdr efm))))))
1690 (cons (convert-effective-method (car efm))
1691 (convert-effective-method (cdr efm))))
1693 (define-method-combination standard-with-restarts ()
1696 (primary () :required t)
1698 (flet ((call-methods-sequentially (methods)
1699 (mapcar #'(lambda (method)
1700 `(CALL-METHOD ,method))
1702 (let ((form (if (or before after (rest primary))
1703 `(MULTIPLE-VALUE-PROG1
1705 ,@(call-methods-sequentially before)
1706 (CALL-METHOD ,(first primary) ,(rest primary)))
1707 ,@(call-methods-sequentially (reverse after)))
1708 `(CALL-METHOD ,(first primary)))))
1711 `(CALL-METHOD ,(first around)
1712 (,@(rest around) (MAKE-METHOD ,form)))))
1713 (convert-effective-method form))))
1714 (defgeneric testgf16 (x) (:method-combination standard-with-restarts))
1715 (defclass testclass16a () ())
1716 (defclass testclass16b (testclass16a) ())
1717 (defclass testclass16c (testclass16a) ())
1718 (defclass testclass16d (testclass16b testclass16c) ())
1719 (defmethod testgf16 ((x testclass16a))
1721 (not (null (find-restart 'method-redo)))
1722 (not (null (find-restart 'method-return)))))
1723 (defmethod testgf16 ((x testclass16b))
1724 (cons 'b (call-next-method)))
1725 (defmethod testgf16 ((x testclass16c))
1726 (cons 'c (call-next-method)))
1727 (defmethod testgf16 ((x testclass16d))
1728 (cons 'd (call-next-method)))
1729 (testgf16 (make-instance 'testclass16d))
1731 Expected: (D B C A T T)
1732 Got: ERROR CALL-METHOD outside of a effective method form
1734 This is a bug because ANSI CL HyperSpec/Body/locmac_call-m__make-method
1736 "The macro call-method invokes the specified method, supplying it with
1737 arguments and with definitions for call-next-method and for next-method-p.
1738 If the invocation of call-method is lexically inside of a make-method,
1739 the arguments are those that were supplied to that method. Otherwise
1740 the arguments are those that were supplied to the generic function."
1741 and the example uses nothing more than these two cases (as you can see by
1742 doing (trace convert-effective-method)).
1744 361: initialize-instance of standard-reader-method ignores :function argument
1745 (reported by Bruno Haible)
1746 Pass a custom :function argument to initialize-instance of a
1747 standard-reader-method instance, but it has no effect.
1748 ;; Check that it's possible to define reader methods that do typechecking.
1750 (defclass typechecking-reader-method (sb-pcl:standard-reader-method)
1752 (defmethod initialize-instance ((method typechecking-reader-method) &rest initargs
1753 &key slot-definition)
1754 (let ((name (sb-pcl:slot-definition-name slot-definition))
1755 (type (sb-pcl:slot-definition-type slot-definition)))
1756 (apply #'call-next-method method
1757 :function #'(lambda (args next-methods)
1758 (declare (ignore next-methods))
1759 (apply #'(lambda (instance)
1760 (let ((value (slot-value instance name)))
1761 (unless (typep value type)
1762 (error "Slot ~S of ~S is not of type ~S: ~S"
1763 name instance type value))
1767 (defclass typechecking-reader-class (standard-class)
1769 (defmethod sb-pcl:validate-superclass ((c1 typechecking-reader-class) (c2 standard-class))
1771 (defmethod reader-method-class ((class typechecking-reader-class) direct-slot &rest args)
1772 (find-class 'typechecking-reader-method))
1773 (defclass testclass25 ()
1774 ((pair :type (cons symbol (cons symbol null)) :initarg :pair :accessor testclass25-pair))
1775 (:metaclass typechecking-reader-class))
1776 (macrolet ((succeeds (form)
1777 `(not (nth-value 1 (ignore-errors ,form)))))
1778 (let ((p (list 'abc 'def))
1779 (x (make-instance 'testclass25)))
1780 (list (succeeds (make-instance 'testclass25 :pair '(seventeen 17)))
1781 (succeeds (setf (testclass25-pair x) p))
1782 (succeeds (setf (second p) 456))
1783 (succeeds (testclass25-pair x))
1784 (succeeds (slot-value x 'pair))))))
1785 Expected: (t t t nil t)
1788 (inspect (first (sb-pcl:generic-function-methods #'testclass25-pair)))
1789 shows that the method was created with a FAST-FUNCTION slot but with a
1790 FUNCTION slot of NIL.
1792 362: missing error when a slot-definition is created without a name
1793 (reported by Bruno Haible)
1794 The MOP says about slot-definition initialization:
1795 "The :NAME argument is a slot name. An ERROR is SIGNALled if this argument
1796 is not a symbol which can be used as a variable name. An ERROR is SIGNALled
1797 if this argument is not supplied."
1799 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition))
1801 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION NIL>
1803 363: missing error when a slot-definition is created with a wrong documentation object
1804 (reported by Bruno Haible)
1805 The MOP says about slot-definition initialization:
1806 "The :DOCUMENTATION argument is a STRING or NIL. An ERROR is SIGNALled
1807 if it is not. This argument default to NIL during initialization."
1809 (make-instance (find-class 'sb-pcl:standard-direct-slot-definition)
1811 :documentation 'not-a-string)
1813 Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION FOO>
1815 364: does not support class objects as specializer names
1816 (reported by Bruno Haible)
1817 According to ANSI CL 7.6.2, class objects are valid specializer names,
1818 and "Parameter specializer names are used in macros intended as the
1819 user-level interface (defmethod)". DEFMETHOD's syntax section doesn't
1820 mention this possibility in the BNF for parameter-specializer-name;
1821 however, this appears to be an editorial omission, since the CLHS
1822 mentions issue CLASS-OBJECT-SPECIALIZER:AFFIRM as being approved
1823 by X3J13. SBCL doesn't support it:
1824 (defclass foo () ())
1825 (defmethod goo ((x #.(find-class 'foo))) x)
1826 Expected: #<STANDARD-METHOD GOO (#<STANDARD-CLASS FOO>)>
1827 Got: ERROR "#<STANDARD-CLASS FOO> is not a legal class name."
1829 365: mixin on generic-function subclass
1830 (reported by Bruno Haible)
1832 (defclass prioritized-dispatcher ()
1833 ((dependents :type list :initform nil)))
1834 on a generic-function subclass:
1835 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1837 (:metaclass sb-pcl:funcallable-standard-class))
1838 SBCL gives an error on this, telling to define a method on SB-MOP:VALIDATE-SUPERCLASS. If done,
1839 (defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
1840 (c2 (eql (find-class 'prioritized-dispatcher))))
1843 (defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
1845 (:metaclass sb-pcl:funcallable-standard-class))
1846 => debugger invoked on a SIMPLE-ERROR in thread 6687:
1847 layout depth conflict: #(#<SB-KERNEL:LAYOUT for T {500E1E9}> ...)
1849 Further discussion on this: http://thread.gmane.org/gmane.lisp.steel-bank.general/491
1851 366: cannot define two generic functions with user-defined class
1852 (reported by Bruno Haible)
1853 it is possible to define one generic function class and an instance
1854 of it. But attempting to do the same thing again, in the same session,
1855 leads to a "Control stack exhausted" error. Test case:
1856 (defclass my-generic-function-1 (standard-generic-function)
1858 (:metaclass sb-pcl:funcallable-standard-class))
1859 (defgeneric testgf-1 (x) (:generic-function-class my-generic-function-1)
1860 (:method ((x integer)) (cons 'integer nil)))
1861 (defclass my-generic-function-2 (standard-generic-function)
1863 (:metaclass sb-pcl:funcallable-standard-class))
1864 (defgeneric testgf-2 (x) (:generic-function-class my-generic-function-2)
1865 (:method ((x integer)) (cons 'integer nil)))
1866 => SB-KERNEL::CONTROL-STACK-EXHAUSTED
1868 367: TYPE-ERROR at compile time, undetected TYPE-ERROR at runtime
1870 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1874 (i367s (make-array 0 :fill-pointer t) :type (or (vector i367) null)))
1876 (g367 (error "missing :G367") :type g367 :read-only t))
1877 ;;; In sbcl-0.8.18, commenting out this (DECLAIM (FTYPE ... R367))
1878 ;;; gives an internal error at compile time:
1879 ;;; The value #<SB-KERNEL:NAMED-TYPE NIL> is not of
1880 ;;; type SB-KERNEL:VALUES-TYPE.
1881 (declaim (ftype (function ((vector i367) e367) (or s367 null)) r367))
1882 (declaim (ftype (function ((vector e367)) (values)) h367))
1884 (let ((x (g367-i367s (make-g367))))
1885 (let* ((y (or (r367 x w)
1888 (format t "~&Y=~S Z=~S~%" y z)
1890 (defun r367 (x y) (declare (ignore x y)) nil)
1891 (defun h367 (x) (declare (ignore x)) (values))
1892 ;;; In sbcl-0.8.18, executing this form causes an low-level error
1893 ;;; segmentation violation at #X9B0E1F4
1894 ;;; (instead of the TYPE-ERROR that one might like).
1895 (frob 0 (make-e367))
1896 can be made to cause two different problems, as noted in the comments:
1897 bug 367a: Compile and load the file. No TYPE-ERROR is signalled at
1898 run time (in the (S367-G367 Y) form of FROB, when Y is NIL
1899 instead of an instance of S367). Instead (on x86/Linux at least)
1900 we end up with a segfault.
1901 bug 367b: Comment out the (DECLAIM (FTYPE ... R367)), and compile
1902 the file. The compiler fails with TYPE-ERROR at compile time.
1904 368: miscompiled OR (perhaps related to bug 367)
1905 Trying to relax type declarations to find a workaround for bug 367,
1906 it turns out that even when the return type isn't declared (or
1907 declared to be T, anyway) the system remains confused about type
1908 inference in code similar to that for bug 367:
1909 (in-package :cl-user)
1910 (declaim (optimize (safety 3) (debug 2) (speed 2) (space 1)))
1914 (i368s (make-array 0 :fill-pointer t) :type (or (vector i368) null)))
1916 (g368 (error "missing :G368") :type g368 :read-only t))
1917 (declaim (ftype (function (fixnum (vector i368) e368) t) r368))
1918 (declaim (ftype (function (fixnum (vector e368)) t) h368))
1919 (defparameter *h368-was-called-p* nil)
1920 (defun nsu (vertices e368)
1921 (let ((i368s (g368-i368s (make-g368))))
1922 (let ((fuis (r368 0 i368s e368)))
1923 (format t "~&FUIS=~S~%" fuis)
1924 (or fuis (h368 0 i368s)))))
1926 (declare (ignore w x y))
1929 (declare (ignore w x))
1930 (setf *h368-was-called-p* t)
1931 (make-s368 :g368 (make-g368)))
1933 (format t "~&calling NSU~%")
1934 (let ((nsu (nsu #() (make-e368))))
1935 (format t "~&NSU returned ~S~%" nsu)
1936 (format t "~&*H368-WAS-CALLED-P*=~S~%" *h368-was-called-p*)
1937 (assert (s368-p nsu))
1938 (assert *h368-was-called-p*))
1939 In sbcl-0.8.18, both ASSERTs fail, and (DISASSEMBLE 'NSU) shows
1940 that no call to H368 is compiled.
1942 369: unlike-an-intersection behavior of VALUES-TYPE-INTERSECTION
1943 In sbcl-0.8.18.2, the identity $(x \cap y \cap y)=(x \cap y)$
1944 does not hold for VALUES-TYPE-INTERSECTION, even for types which
1945 can be intersected exactly, so that ASSERTs fail in this test case:
1946 (in-package :cl-user)
1947 (let ((types (mapcar #'sb-c::values-specifier-type
1948 '((values (vector package) &optional)
1949 (values (vector package) &rest t)
1950 (values (vector hash-table) &rest t)
1951 (values (vector hash-table) &optional)
1952 (values t &optional)
1954 (values nil &optional)
1955 (values nil &rest t)
1956 (values sequence &optional)
1957 (values sequence &rest t)
1958 (values list &optional)
1959 (values list &rest t)))))
1962 (let ((i (sb-c::values-type-intersection x y)))
1963 (assert (sb-c::type= i (sb-c::values-type-intersection i x)))
1964 (assert (sb-c::type= i (sb-c::values-type-intersection i y)))))))
1966 370: reader misbehaviour on large-exponent floats
1967 (read-from-string "1.0s1000000000000000000000000000000000000000")
1968 causes the reader to attempt to create a very large bignum (which it
1969 will then attempt to coerce to a rational). While this isn't
1970 completely wrong, it is probably not ideal -- checking the floating
1971 point control word state and then returning the relevant float
1972 (most-positive-short-float or short-float-infinity) or signalling an
1973 error immediately would seem to make more sense.
1975 372: floating-point overflow not signalled on ppc/darwin
1976 The following assertions in float.pure.lisp fail on ppc/darwin
1977 (Mac OS X version 10.3.7):
1978 (assert (raises-error? (scale-float 1.0 most-positive-fixnum)
1979 floating-point-overflow))
1980 (assert (raises-error? (scale-float 1.0d0 (1+ most-positive-fixnum))
1981 floating-point-overflow)))
1982 as the SCALE-FLOAT just returns
1983 #.SB-EXT:SINGLE/DOUBLE-FLOAT-POSITIVE-INFINITY. These tests have been
1984 disabled on Darwin for now.
1986 374: BIT-AND problem on ppc/darwin:
1987 The BIT-AND test in bit-vector.impure-cload.lisp results in
1988 fatal error encountered in SBCL pid 8356:
1989 GC invariant lost, file "gc-common.c", line 605
1990 on ppc/darwin. Test disabled for the duration.
1993 (compile nil '(lambda (p1)
1994 (declare (optimize (speed 1) (safety 2) (debug 2) (space 0))
1998 fails on hairy type check in IR2.
2000 1. KEYWORDP is MAYBE-INLINE expanded (before TYPEP-like
2001 transformation could eliminate it).
2003 2. From the only call of KEYWORDP the type of its argument is
2004 derived to be KEYWORD.
2006 2. Type check for P1 is generated; it uses KEYWORDP to perform the
2007 check, and so references the local function; from the KEYWORDP
2008 argument type new CAST to KEYWORD is generated. The compiler
2011 377: Memory fault error reporting
2012 On those architectures where :C-STACK-IS-CONTROL-STACK is in
2013 *FEATURES*, we handle SIG_MEMORY_FAULT (SEGV or BUS) on an altstack,
2014 so we cannot handle the signal directly (as in interrupt_handle_now())
2015 in the case when the signal comes from some external agent (the user
2016 using kill(1), or a fault in some foreign code, for instance). As
2017 of sbcl-0.8.20.20, this is fixed by calling
2018 arrange_return_to_lisp_function() to a new error-signalling
2019 function, but as a result the error reporting is poor: we cannot
2020 even tell the user at which address the fault occurred. We should
2021 arrange such that arguments can be passed to the function called from
2022 arrange_return_to_lisp_function(), but this looked hard to do in
2023 general without suffering from memory leaks.
2025 379: TRACE :ENCAPSULATE NIL broken on ppc/darwin
2026 See commented-out test-case in debug.impure.lisp.
2028 380: Accessor redefinition fails because of old accessor name
2029 When redefining an accessor, SB-PCL::FIX-SLOT-ACCESSORS may try to
2030 find the generic function named by the old accessor name using
2031 ENSURE-GENERIC-FUNCTION and then remove the old accessor's method in
2032 the GF. If the old name does not name a function, or if the old name
2033 does not name a generic function, no attempt to find the GF or remove
2034 any methods is made.
2036 However, if an unrelated GF with an incompatible lambda list exists,
2037 the class redefinition will fail when SB-PCL::REMOVE-READER-METHOD
2038 tries to find and remove a method with an incompatible lambda list
2039 from the unrelated generic function.
2041 381: incautious calls to EQUAL in fasl dumping
2043 (frob #(#1=(a #1#)))
2044 (frob #(#1=(b #1#)))
2045 (frob #(#1=(a #1#)))
2046 in sbcl-0.9.0 causes CONTROL-STACK-EXHAUSTED. My (WHN) impression
2047 is that this follows from the use of (MAKE-HASH-TABLE :TEST 'EQUAL)
2048 to detect sharing, in which case fixing it might require either
2049 getting less ambitious about detecting shared list structure, or
2050 implementing the moral equivalent of EQUAL hash tables in a
2053 382: externalization unexpectedly changes array simplicity
2054 COMPILE-FILE and LOAD
2056 (let ((x #.(make-array 4 :fill-pointer 0)))
2057 (values (eval `(typep ',x 'simple-array))
2058 (typep x 'simple-array))))
2059 then (FOO) => T, NIL.
2061 Similar problems exist with SIMPLE-ARRAY-P, ARRAY-HEADER accessors
2062 and all array dimension functions.
2064 383: ASH'ing non-constant zeros
2067 (declare (type (integer -2 14) b))
2068 (declare (ignorable b))
2069 (ash (imagpart b) 57))
2070 on PPC (and other platforms, presumably) gives an error during the
2071 emission of FASH-ASH-LEFT/FIXNUM=>FIXNUM as the assembler attempts to
2072 stuff a too-large constant into the immediate field of a PPC
2073 instruction. Either the VOP should be fixed or the compiler should be
2074 taught how to transform this case away, paying particular attention
2075 to side-effects that might occur in the arguments to ASH.
2077 384: Compiler runaway on very large character types
2079 (compile nil '(lambda (x)
2080 (declare (type (member #\a 1) x))
2081 (the (member 1 nil) x)))
2083 The types apparently normalize into a very large type, and the compiler
2084 gets lost in REMOVE-DUPLICATES. Perhaps the latter should use
2085 a better algorithm (one based on hash tables, say) on very long lists
2086 when :TEST has its default value?
2090 (compile nil '(lambda (x) (the (not (eql #\a)) x)))
2092 (partially fixed in 0.9.3.1, but a better representation for these
2096 (format nil "~4,1F" 0.001) => "0.00" (should be " 0.0");
2097 (format nil "~4,1@F" 0.001) => "+.00" (should be "+0.0").
2099 386: SunOS/x86 stack exhaustion handling broken
2100 According to <http://alfa.s145.xrea.com/sbcl/solaris-x86.html>, the
2101 stack exhaustion checking (implemented with a write-protected guard
2102 page) does not work on SunOS/x86.
2105 12:10 < jsnell> the package-lock test is basically due to a change in the test
2106 behaviour when you install a handler for error around it. I
2107 thought I'd disabled the test for now, but apparently that was
2109 12:19 < Xophe> jsnell: ah, I see the problem in the package-locks stuff
2110 12:19 < Xophe> it's the same problem as we had with compiler-error conditions
2111 12:19 < Xophe> the thing that's signalled up and down the stack is a subtype of
2112 ERROR, where it probably shouldn't be
2115 (found by Dmitry Bogomolov)
2117 (defclass foo () ((x :type (unsigned-byte 8))))
2118 (defclass bar () ((x :type symbol)))
2119 (defclass baz (foo bar) ())
2123 SB-PCL::SPECIALIZER-APPLICABLE-USING-TYPE-P cannot handle the second argument