3 Bugs can be reported on the help mailing list
4 sbcl-help@lists.sourceforge.net
5 or on the development mailing list
6 sbcl-devel@lists.sourceforge.net
8 Please include enough information in a bug report that someone reading
9 it can reproduce the problem, i.e. don't write
10 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
11 PRINT-OBJECT doesn't seem to work with *PRINT-LENGTH*. Is this a bug?
13 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
14 In sbcl-1.2.3 running under OpenBSD 4.5 on my Alpha box, when
15 I compile and load the file
16 (DEFSTRUCT (FOO (:PRINT-OBJECT (LAMBDA (X Y)
17 (LET ((*PRINT-LENGTH* 4))
20 then at the command line type
22 the program loops endlessly instead of printing the object.
27 There is also some information on bugs in the manual page and
28 in the TODO file. Eventually more such information may move here.
30 The gaps in the number sequence belong to old bug descriptions which
31 have gone away (typically because they were fixed, but sometimes for
32 other reasons, e.g. because they were moved elsewhere).
36 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
37 instead of just punting when a contradictory structure definition
38 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
39 changes their layout, you probably have to rebuild your entire
40 program, even if you know or guess enough about the internals of
41 SBCL to wager that this (undefined in ANSI) operation would be safe.
43 3: "type checking of structure slots"
45 ANSI specifies that a type mismatch in a structure slot
46 initialization value should not cause a warning.
48 This one might not be fixed for a while because while we're big
49 believers in ANSI compatibility and all, (1) there's no obvious
50 simple way to do it (short of disabling all warnings for type
51 mismatches everywhere), and (2) there's a good portable
52 workaround, and (3) by their own reasoning, it looks as though
53 ANSI may have gotten it wrong. ANSI justifies this specification
55 The restriction against issuing a warning for type mismatches
56 between a slot-initform and the corresponding slot's :TYPE
57 option is necessary because a slot-initform must be specified
58 in order to specify slot options; in some cases, no suitable
60 However, in SBCL (as in CMU CL or, for that matter, any compiler
61 which really understands Common Lisp types) a suitable default
62 does exist, in all cases, because the compiler understands the
63 concept of functions which never return (i.e. has return type NIL).
64 Thus, as a portable workaround, you can use a call to some
65 known-never-to-return function as the default. E.g.
67 (BAR (ERROR "missing :BAR argument")
68 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
70 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
71 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
72 (ERROR "missing required argument"))
74 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
75 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
76 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
77 Such code should compile without complaint and work correctly either
78 on SBCL or on any other completely compliant Common Lisp system.
80 b: &AUX argument in a boa-constructor without a default value means
81 "do not initilize this slot" and does not cause type error. But
82 an error may be signalled at read time and it would be good if
88 The "compiling top-level form:" output ought to be condensed.
89 Perhaps any number of such consecutive lines ought to turn into a
90 single "compiling top-level forms:" line.
93 (I *think* this is a bug. It certainly seems like strange behavior. But
94 the ANSI spec is scary, dark, and deep.. -- WHN)
95 (FORMAT NIL "~,1G" 1.4) => "1. "
96 (FORMAT NIL "~3,1G" 1.4) => "1. "
99 Sometimes (SB-EXT:QUIT) fails with
100 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
101 Process inferior-lisp exited abnormally with code 1
102 I haven't noticed a repeatable case of this yet.
105 The printer doesn't report closures very well. This is true in
109 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
110 It would be nice to make closures have a settable name slot,
111 and make things like DEFSTRUCT and FLET, which create closures,
112 set helpful values into this slot.
115 And as long as we're wishing, it would be awfully nice if INSPECT could
116 also report on closures, telling about the values of the bound variables.
119 The compiler assumes that any time a function of declared FTYPE
120 doesn't signal an error, its arguments were of the declared type.
121 E.g. compiling and loading
122 (DECLAIM (OPTIMIZE (SAFETY 3)))
123 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
125 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
127 (COND ((> (FACTORIAL X) 1.0E6)
128 (FORMAT T "too big~%"))
130 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
132 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
135 will cause the INTEGERP case to be selected, giving bogus output a la
137 This violates the "declarations are assertions" principle.
138 According to the ANSI spec, in the section "System Class FUNCTION",
139 this is a case of "lying to the compiler", but the lying is done
140 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
141 not by the unexpectedly general definition of FACTORIAL. In any case,
142 "declarations are assertions" means that lying to the compiler should
143 cause an error to be signalled, and should not cause a bogus
144 result to be returned. Thus, the compiler should not assume
145 that arbitrary functions check their argument types. (It might
146 make sense to add another flag (CHECKED?) to DEFKNOWN to
147 identify functions which *do* check their argument types.)
148 (Also, verify that the compiler handles declared function
149 return types as assertions.)
152 The definitions of SIGCONTEXT-FLOAT-REGISTER and
153 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
154 supported on FreeBSD because the floating point state is not saved,
155 but at least as of FreeBSD 4.0, the floating point state *is* saved,
156 so they could be supported after all. Very likely
157 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
160 The debugger LIST-LOCATIONS command doesn't work properly.
161 (How should it work properly?)
164 Compiling and loading
165 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
167 then requesting a BACKTRACE at the debugger prompt gives no information
168 about where in the user program the problem occurred.
171 Using the pretty-printer from the command prompt gives funny
172 results, apparently because the pretty-printer doesn't know
173 about user's command input, including the user's carriage return
174 that the user, and therefore the pretty-printer thinks that
175 the new output block should start indented 2 or more characters
176 rightward of the correct location.
179 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
180 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
181 crashes SBCL. In general tracing anything which is used in the
182 implementation of TRACE is likely to have the same problem.
185 ANSI says in one place that type declarations can be abbreviated even
186 when the type name is not a symbol, e.g.
187 (DECLAIM ((VECTOR T) *FOOVECTOR*))
188 SBCL doesn't support this. But ANSI says in another place that this
189 isn't allowed. So it's not clear this is a bug after all. (See the
190 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
194 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
195 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
196 the mean of the distribution can be systematically O(0.1%) wrong.
197 Just increasing R-I-E-B is probably not a good solution, since
198 it would decrease efficiency more than is probably necessary. Perhaps
199 using some sort of accept/reject method would be better.
202 Internally the compiler sometimes evaluates
203 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
204 (I stumbled across this when I added an
205 (assert (not (eq type1 *wild-type*)))
206 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
207 in a type context should probably be translated to T, and so it's
208 probably wrong to ask whether it's equal to the T type and then (using
209 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
210 (I haven't tried to investigate this bug enough to guess whether
211 there might be any user-level symptoms.)
213 In fact, the type system is likely to depend on this inequality not
214 holding... * is not equivalent to T in many cases, such as
215 (VECTOR *) /= (VECTOR T).
218 The facility for dumping a running Lisp image to disk gets confused
219 when run without the PURIFY option, and creates an unnecessarily large
220 core file (apparently representing memory usage up to the previous
221 high-water mark). Moreover, when the file is loaded, it confuses the
222 GC, so that thereafter memory usage can never be reduced below that
225 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
226 The interaction between gencgc and the variables used by
227 save-lisp-and-die is still nonoptimal, though, so no respite from
231 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
232 CL), out-of-line structure slot setters are horribly inefficient
233 whenever the type of the slot is declared, because out-of-line
234 structure slot setters are implemented as closures to save space,
235 so the compiler doesn't compile the type test into code, but
236 instead just saves the type in a lexical closure and interprets it
238 To exercise the problem, compile and load
239 (cl:in-package :cl-user)
241 (bar (error "missing") :type bar))
244 (loop (setf (foo-bar *foo*) x)))
246 (defvar *bar* (make-bar))
247 (defvar *foo* (make-foo :bar *bar*))
248 (defvar *setf-foo-bar* #'(setf foo-bar))
250 (loop (funcall *setf-foo-bar* x *foo*)))
251 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
252 use BACKTRACE, to see it's spending all essentially all its time
253 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
254 One possible solution would be simply to give up on
255 representing structure slot accessors as functions, and represent
256 them as macroexpansions instead. This can be inconvenient for users,
257 but it's not clear that it's worse than trying to help by expanding
258 into a horribly inefficient implementation.
259 As a workaround for the problem, #'(SETF FOO) expressions
260 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
261 (defmacro efficient-setf-function (place-function-name)
262 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
263 ;; a workaround for the problem, encouraging the
264 ;; inline expansion of the structure accessor, so
265 ;; that the compiler can optimize its type test
266 (let ((new-value (gensym "NEW-VALUE-"))
267 (structure-value (gensym "STRUCTURE-VALUE-")))
268 `(lambda (,new-value ,structure-value)
269 (setf (,place-function-name ,structure-value)
271 ;; no problem, can just use the ordinary expansion
272 `(function (setf ,place-function-name))))
275 There's apparently a bug in CEILING optimization which caused
276 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
277 applied the patches to SBCL and they didn't seem to cause problems
278 (as reported sbcl-devel 2001-05-04). However, since the patches
279 modify nontrivial code which was apparently written incorrectly
280 the first time around, until regression tests are written I'm not
281 comfortable merging the patches in the CVS version of SBCL.
284 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
285 a clean, just-started SBCL system. And it seems to be right:
286 (ROOM T) can bring a small computer to its knees for a *long*
287 time trying to GC afterwards. Surely there's some more economical
288 way to implement (ROOM T).
290 Daniel Barlow doesn't know what fixed this, but observes that it
291 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
292 in a fresh SBCL causes
294 debugger invoked on a SB-INT:BUG in thread 5911:
295 failed AVER: "(SAP= CURRENT END)"
297 unless a GC has happened beforehand.
300 When the compiler inline expands functions, it may be that different
301 kinds of return values are generated from different code branches.
302 E.g. an inline expansion of POSITION generates integer results
303 from one branch, and NIL results from another. When that inline
304 expansion is used in a context where only one of those results
307 (aref *a1* (position x *a2*)))
308 and the compiler can't prove that the unacceptable branch is
309 never taken, then bogus type mismatch warnings can be generated.
310 If you need to suppress the type mismatch warnings, you can
311 suppress the inline expansion,
313 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
314 (aref *a1* (position x *a2*)))
315 or, sometimes, suppress them by declaring the result to be of an
318 (aref *a1* (the integer (position x *a2*))))
320 This is not a new compiler problem in 0.7.0, but the new compiler
321 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
322 more conspicuous. If you don't need performance from these functions,
323 and the bogus warnings are a nuisance for you, you can return to
324 your pre-0.7.0 state of grace with
325 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
330 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
331 the entire lexical environment at the point of MACROLET available
332 in the bodies of the macroexpander functions. In particular, it
333 allows the function bodies (which run at compile time) to try to
334 access lexical variables (which are only defined at runtime).
335 It doesn't even issue a warning, which is bad.
337 The SBCL behavior arguably conforms to the ANSI spec (since the
338 spec says that the behavior is undefined, ergo anything conforms).
339 However, it would be better to issue a compile-time error.
340 Unfortunately I (WHN) don't see any simple way to detect this
341 condition in order to issue such an error, so for the meantime
342 SBCL just does this weird broken "conforming" thing.
344 The ANSI standard says, in the definition of the special operator
346 The macro-expansion functions defined by MACROLET are defined
347 in the lexical environment in which the MACROLET form appears.
348 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
349 the local macro definitions in a MACROLET, but the consequences
350 are undefined if the local macro definitions reference any
351 local variable or function bindings that are visible in that
353 Then it seems to contradict itself by giving the example
355 (macrolet ((fudge (z)
356 ;The parameters x and flag are not accessible
357 ; at this point; a reference to flag would be to
358 ; the global variable of that name.
359 ` (if flag (* ,z ,z) ,z)))
360 ;The parameters x and flag are accessible here.
364 The comment "a reference to flag would be to the global variable
365 of the same name" sounds like good behavior for the system to have.
366 but actual specification quoted above says that the actual behavior
369 (Since 0.7.8.23 macroexpanders are defined in a restricted version
370 of the lexical environment, containing no lexical variables and
371 functions, which seems to conform to ANSI and CLtL2, but signalling
372 a STYLE-WARNING for references to variables similar to locals might
376 (as reported by Gabe Garza on cmucl-help 2001-09-21)
378 (defun test-pred (x y)
382 (func (lambda () x)))
383 (print (eq func func))
384 (print (test-pred func func))
385 (delete func (list func))))
386 Now calling (TEST-CASE) gives output
389 (#<FUNCTION {500A9EF9}>)
390 Evidently Python thinks of the lambda as a code transformation so
391 much that it forgets that it's also an object.
394 Ideally, uninterning a symbol would allow it, and its associated
395 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
396 at least as of sbcl-0.7.0, this isn't the case. Information about
397 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
398 essentially in ordinary (non-weak) hash tables keyed by symbols.
399 Thus, once a system has an entry in this system, it tends to live
400 forever, even when it is uninterned and all other references to it
404 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
405 prominent SourceForge web/db bug tracking system, which is
406 unfortunately not a reliable way to get a timely response from
407 the SBCL maintainers)
408 In the course of trying to build a test case for an
409 application error, I encountered this behavior:
410 If you start up sbcl, and then lay on CTRL-C for a
411 minute or two, the lisp process will eventually say:
412 %PRIMITIVE HALT called; the party is over.
413 and throw you into the monitor. If I start up lisp,
414 attach to the process with strace, and then do the same
415 (abusive) thing, I get instead:
416 access failure in heap page not marked as write-protected
417 and the monitor again. I don't know enough to have the
418 faintest idea of what is going on here.
419 This is with sbcl 6.12, uname -a reports:
420 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
421 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
422 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
423 it took more than two minutes (but less than five) for me.
427 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
428 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
429 COMPLEX implementation didn't deal with this, and hasn't been
430 upgraded to do so. (This doesn't seem to be a high priority
431 conformance problem, since seems hard to construct useful code
434 b. (fixed in 0.8.3.43)
437 Floating point errors are reported poorly. E.g. on x86 OpenBSD
440 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
441 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
442 No traps are enabled? How can this be?
443 It should be possible to be much more specific (overflow, division
444 by zero, etc.) and of course the "How can this be?" should be fixable.
446 See also bugs #45.c and #183
449 (reported by Robert E. Brown 2002-04-16)
450 When a function is called with too few arguments, causing the
451 debugger to be entered, the uninitialized slots in the bad call frame
452 seem to cause GCish problems, being interpreted as tagged data even
453 though they're not. In particular, executing ROOM in the
454 debugger at that point causes AVER failures:
457 * (lisp-implementation-version)
463 failed AVER: "(SAP= CURRENT END)"
464 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
465 isn't too surprising since there are many differences in stack
466 implementation and GC conservatism between the X86 and other ports.)
468 This is probably the same bug as 216
471 The compiler sometimes tries to constant-fold expressions before
472 it checks to see whether they can be reached. This can lead to
473 bogus warnings about errors in the constant folding, e.g. in code
476 (WRITE-STRING (> X 0) "+" "0"))
477 compiled in a context where the compiler can prove that X is NIL,
478 and the compiler complains that (> X 0) causes a type error because
479 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
480 caused a full WARNING, which made the bug really annoying because then
481 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
482 code. Since then the warning has been downgraded to STYLE-WARNING,
483 so it's still a bug but at least it's a little less annoying.
485 183: "IEEE floating point issues"
486 Even where floating point handling is being dealt with relatively
487 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
488 accrued-exceptions and current-exceptions part of the fp control
489 word don't seem to bear much relation to reality. E.g. on
493 debugger invoked on condition of type DIVISION-BY-ZERO:
494 arithmetic error DIVISION-BY-ZERO signalled
495 0] (sb-vm::get-floating-point-modes)
497 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
498 :ROUNDING-MODE :NEAREST
499 :CURRENT-EXCEPTIONS NIL
500 :ACCRUED-EXCEPTIONS (:INEXACT)
503 * (sb-vm::get-floating-point-modes)
504 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
505 :ROUNDING-MODE :NEAREST
506 :CURRENT-EXCEPTIONS (:INEXACT)
507 :ACCRUED-EXCEPTIONS (:INEXACT)
510 188: "compiler performance fiasco involving type inference and UNION-TYPE"
514 (declare (optimize (safety 3)))
515 (declare (optimize (compilation-speed 2)))
516 (declare (optimize (speed 1) (debug 1) (space 1)))
518 (declare (type (integer 0) start))
519 (print (incf start 22))
520 (print (incf start 26))
521 (print (incf start 28)))
523 (declare (type (integer 0) start))
524 (print (incf start 22))
525 (print (incf start 26)))
527 (declare (type (integer 0) start))
528 (print (incf start 22))
529 (print (incf start 26))))))
531 This example could be solved with clever enough constraint
532 propagation or with SSA, but consider
537 The careful type of X is {2k} :-(. Is it really important to be
538 able to work with unions of many intervals?
540 191: "Miscellaneous PCL deficiencies"
541 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
542 a. DEFCLASS does not inform the compiler about generated
543 functions. Compiling a file with
547 (WITH-SLOTS (A-CLASS-X) A
549 results in a STYLE-WARNING:
551 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
553 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
554 Mai points out that the declamation of functions is in fact
555 incorrect in some cases (most notably for structure
556 classes). This means that at present erroneous attempts to use
557 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
558 won't get the corresponding STYLE-WARNING.
559 c. (fixed in 0.8.4.23)
561 201: "Incautious type inference from compound types"
562 a. (reported by APD sbcl-devel 2002-09-17)
564 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
566 (FORMAT NIL "~S IS ~S, Y = ~S"
573 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
577 (declare (type (array * (4 4)) x))
579 (setq x (make-array '(4 4)))
580 (adjust-array y '(3 5))
581 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
583 * (foo (make-array '(4 4) :adjustable t))
586 205: "environment issues in cross compiler"
587 (These bugs have no impact on user code, but should be fixed or
589 a. Macroexpanders introduced with MACROLET are defined in the null
591 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
592 the null lexical environment.
593 c. The cross-compiler cannot inline functions defined in a non-null
596 206: ":SB-FLUID feature broken"
597 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
598 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
601 207: "poorly distributed SXHASH results for compound data"
602 SBCL's SXHASH could probably try a little harder. ANSI: "the
603 intent is that an implementation should make a good-faith
604 effort to produce hash-codes that are well distributed
605 within the range of non-negative fixnums". But
606 (let ((hits (make-hash-table)))
609 (let* ((ij (cons i j))
610 (newlist (push ij (gethash (sxhash ij) hits))))
612 (format t "~&collision: ~S~%" newlist))))))
613 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
614 would be an obvious way of fix. Maybe it would be acceptably efficient
615 to redo MIX using a lookup into a 256-entry s-box containing
616 29-bit pseudorandom numbers?
618 211: "keywords processing"
619 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
620 number of keyword arguments.
623 (flet ((foo (&key y) (list y)))
624 (list (foo :y 1 :y 2)))
626 issues confusing message
631 ; caught STYLE-WARNING:
632 ; The variable #:G15 is defined but never used.
634 212: "Sequence functions and circular arguments"
635 COERCE, MERGE and CONCATENATE go into an infinite loop when given
636 circular arguments; it would be good for the user if they could be
637 given an error instead (ANSI 17.1.1 allows this behaviour on the part
638 of the implementation, as conforming code cannot give non-proper
639 sequences to these functions. MAP also has this problem (and
640 solution), though arguably the convenience of being able to do
641 (MAP 'LIST '+ FOO '#1=(1 . #1#))
642 might be classed as more important (though signalling an error when
643 all of the arguments are circular is probably desireable).
645 213: "Sequence functions and type checking"
646 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
647 check that it will return a sequence of the given type. Fixing
648 it along the same lines as the others (cf. work done around
649 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
650 entirely straightforward.
651 c. All of these functions will silently accept a type of the form
653 whether or not the return value is of this type. This is
654 probably permitted by ANSI (see "Exceptional Situations" under
655 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
656 know about this escape clause, so code of the form
657 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
658 can erroneously return T.
660 215: ":TEST-NOT handling by functions"
661 a. FIND and POSITION currently signal errors when given non-NIL for
662 both their :TEST and (deprecated) :TEST-NOT arguments, but by
663 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
664 means that the effect is "unpredictable but harmless". It's not
665 clear what that actually means; it may preclude conforming
666 implementations from signalling errors.
667 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
668 when conflict occurs. As a quality of implementation issue, it
669 might be preferable to treat :TEST and :TEST-NOT as being in some
670 sense the same &KEY, and effectively take the first test function in
672 c. Again, a quality of implementation issue: it would be good to issue a
673 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
674 WARNING for calls with both :TEST and :TEST-NOT; possibly this
675 latter should be WARNed about at execute-time too.
677 216: "debugger confused by frames with invalid number of arguments"
678 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
679 leaves the system confused, enough so that (QUIT) no longer works.
680 It's as though the process of working with the uninitialized slot in
681 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
682 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
684 This is probably the same bug as 162
686 217: "Bad type operations with FUNCTION types"
689 * (values-type-union (specifier-type '(function (base-char)))
690 (specifier-type '(function (integer))))
692 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
694 It causes insertion of wrong type assertions into generated
698 (let ((f (etypecase x
699 (character #'write-char)
700 (integer #'write-byte))))
703 (character (write-char x s))
704 (integer (write-byte x s)))))
706 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
708 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
709 produce invalid code, but type checking is not accurate.)
711 233: bugs in constraint propagation
713 (declaim (optimize (speed 2) (safety 3)))
715 (if (typep (prog1 x (setq x y)) 'double-float)
718 (foo 1d0 5) => segmentation violation
720 235: "type system and inline expansion"
722 (declaim (ftype (function (cons) number) acc))
723 (declaim (inline acc))
725 (the number (car c)))
728 (values (locally (declare (optimize (safety 0)))
730 (locally (declare (optimize (safety 3)))
733 (foo '(nil) '(t)) => NIL, T.
735 237: "Environment arguments to type functions"
736 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
737 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
738 argument, but they ignore it completely. This is almost
739 certainly not correct.
740 b. Also, the compiler's optimizers for TYPEP have not been informed
741 about the new argument; consequently, they will not transform
742 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
743 just as optimizeable as (TYPEP 1 'INTEGER).
745 238: "REPL compiler overenthusiasm for CLOS code"
747 * (defclass foo () ())
748 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
749 causes approximately 100 lines of code deletion notes. Some
750 discussion on this issue happened under the title 'Three "interesting"
751 bugs in PCL', resulting in a fix for this oververbosity from the
752 compiler proper; however, the problem persists in the interactor
753 because the notion of original source is not preserved: for the
754 compiler, the original source of the above expression is (DEFMETHOD
755 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
756 compiler gets its hands on the code needing compilation from the REPL,
757 it has been macroexpanded several times.
759 A symptom of the same underlying problem, reported by Tony Martinez:
761 (with-input-from-string (*query-io* " no")
763 (simple-type-error () 'error))
765 ; (SB-KERNEL:FLOAT-WAIT)
767 ; note: deleting unreachable code
768 ; compilation unit finished
771 242: "WRITE-SEQUENCE suboptimality"
772 (observed from clx performance)
773 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
774 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
775 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
776 rather than writing the sequence in one go, leading to severe
777 performance degradation.
779 243: "STYLE-WARNING overenthusiasm for unused variables"
780 (observed from clx compilation)
781 In sbcl-0.7.14, in the presence of the macros
782 (DEFMACRO FOO (X) `(BAR ,X))
783 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
784 somewhat surprising style warnings are emitted for
785 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
787 ; (LAMBDA (Y) (FOO Y))
789 ; caught STYLE-WARNING:
790 ; The variable Y is defined but never used.
792 245: bugs in disassembler
793 b. On X86 operand size prefix is not recognized.
796 (defun foo (&key (a :x))
800 does not cause a warning. (BTW: old SBCL issued a warning, but for a
801 function, which was never called!)
804 Compiler does not emit warnings for
806 a. (lambda () (svref (make-array 8 :adjustable t) 1))
809 (list (let ((y (the real x)))
810 (unless (floatp y) (error ""))
815 (declare (optimize (debug 0)))
816 (declare (type vector x))
817 (list (fill-pointer x)
821 Complex array type does not have corresponding type specifier.
823 This is a problem because the compiler emits optimization notes when
824 you use a non-simple array, and without a type specifier for hairy
825 array types, there's no good way to tell it you're doing it
826 intentionally so that it should shut up and just compile the code.
828 Another problem is confusing error message "asserted type ARRAY
829 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
830 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
832 The last problem is that when type assertions are converted to type
833 checks, types are represented with type specifiers, so we could lose
834 complex attribute. (Now this is probably not important, because
835 currently checks for complex arrays seem to be performed by
839 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
840 warning. Analogous cases with the index and length being equal and
841 greater than 0 are warned for; the problem here seems to be that the
842 type required for an array reference of this type is (INTEGER 0 (0))
843 which is canonicalized to NIL.
848 (t1 (specifier-type s)))
849 (eval `(defstruct ,s))
850 (type= t1 (specifier-type s)))
855 b. The same for CSUBTYPEP.
857 262: "yet another bug in inline expansion of local functions"
861 (declare (integer x y))
864 (declare (integer u))
865 (if (> (1+ (the unsigned-byte u)) 0)
867 (return (+ 38 (cos (/ u 78)))))))
868 (declare (inline xyz))
870 (* (funcall (eval #'xyz) x)
872 (funcall (if (> x 5) #'xyz #'identity)
877 Urgh... It's time to write IR1-copier.
880 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
881 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
882 looks reasonable, if not obviously correct; however, it caused the
883 PPC/Linux port to segfault during warm-init while loading
884 src/pcl/std-class.fasl. A workaround patch was made, but it would
885 be nice to understand why the first patch caused problems, and to
886 fix the cause if possible.
888 268: "wrong free declaration scope"
889 The following code must signal type error:
891 (locally (declare (optimize (safety 3)))
892 (flet ((foo (x &optional (y (car x)))
893 (declare (optimize (safety 0)))
895 (funcall (eval #'foo) 1)))
898 In the following function constraint propagator optimizes nothing:
901 (declare (integer x))
902 (declare (optimize speed))
910 Compilation of the following two forms causes "X is unbound" error:
912 (symbol-macrolet ((x pi))
913 (macrolet ((foo (y) (+ x y)))
914 (declaim (inline bar))
920 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
923 CLHS says that type declaration of a symbol macro should not affect
924 its expansion, but in SBCL it does. (If you like magic and want to
925 fix it, don't forget to change all uses of MACROEXPAND to
929 The following code (taken from CLOCC) takes a lot of time to compile:
932 (declare (type (integer 0 #.large-constant) n))
935 (fixed in 0.8.2.51, but a test case would be good)
940 (declare (optimize speed))
941 (loop for i of-type (integer 0) from 0 by 2 below 10
944 uses generic arithmetic.
946 b. (fixed in 0.8.3.6)
948 279: type propagation error -- correctly inferred type goes astray?
949 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
950 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
951 &OPTIONAL), not a (INTEGER 1 536870911)
952 is emitted when compiling this file:
953 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
954 (integer #.most-negative-fixnum 0))
959 (let* (;; Uncomment this for a type mismatch warning indicating
960 ;; that the type of (FOO X) is correctly understood.
961 #+nil (fs-foo (float-sign (foo x)))
962 ;; Uncomment this for a type mismatch warning
963 ;; indicating that the type of (ABS (FOO X)) is
964 ;; correctly understood.
965 #+nil (fs-abs-foo (float-sign (abs (foo x))))
966 ;; something wrong with this one though
967 (abs-foo (abs (foo x))))
968 (declare (type (integer 1 100) abs-foo))
973 281: COMPUTE-EFFECTIVE-METHOD error signalling.
974 (slightly obscured by a non-0 default value for
975 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
976 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
977 when it finds a method with invalid qualifiers. However, it
978 shouldn't signal errors when any such methods are not applicable to
979 the particular call being evaluated, and certainly it shouldn't when
980 simply precomputing effective methods that may never be called.
981 (setf sb-pcl::*max-emf-precompute-methods* 0)
983 (:method-combination +)
984 (:method ((x symbol)) 1)
985 (:method + ((x number)) x))
986 (foo 1) -> ERROR, but should simply return 1
988 The issue seems to be that construction of a discriminating function
989 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
991 283: Thread safety: libc functions
992 There are places that we call unsafe-for-threading libc functions
993 that we should find alternatives for, or put locks around. Known or
994 strongly suspected problems, as of 0.8.3.10: please update this
995 bug instead of creating new ones
997 localtime() - called for timezone calculations in code/time.lisp
999 284: Thread safety: special variables
1000 There are lots of special variables in SBCL, and I feel sure that at
1001 least some of them are indicative of potentially thread-unsafe
1002 parts of the system. See doc/internals/notes/threading-specials
1004 286: "recursive known functions"
1005 Self-call recognition conflicts with known function
1006 recognition. Currently cross compiler and target COMPILE do not
1007 recognize recursion, and in target compiler it can be disabled. We
1008 can always disable it for known functions with RECURSIVE attribute,
1009 but there remains a possibility of a function with a
1010 (tail)-recursive simplification pass and transforms/VOPs for base
1013 287: PPC/Linux miscompilation or corruption in first GC
1014 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
1015 segmentation fault is reported at the point of first triggered GC,
1016 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
1017 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
1018 that this merely obscures, not solves, the underlying problem; as and when
1019 underlying problems are fixed, it would be worth trying again to provoke
1022 288: fundamental cross-compilation issues (from old UGLINESS file)
1023 Using host floating point numbers to represent target floating point
1024 numbers, or host characters to represent target characters, is
1025 theoretically shaky. (The characters are OK as long as the characters
1026 are in the ANSI-guaranteed character set, though, so they aren't a
1027 real problem as long as the sources don't need anything but that;
1028 the floats are a real problem.)
1030 289: "type checking and source-transforms"
1032 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1035 Our policy is to check argument types at the moment of a call. It
1036 disagrees with ANSI, which says that type assertions are put
1037 immediately onto argument expressions, but is easier to implement in
1038 IR1 and is more compatible to type inference, inline expansion,
1039 etc. IR1-transforms automatically keep this policy, but source
1040 transforms for associative functions (such as +), being applied
1041 during IR1-convertion, do not. It may be tolerable for direct calls
1042 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1044 b. Another aspect of this problem is efficiency. [x y + z +]
1045 requires less registers than [x y z + +]. This transformation is
1046 currently performed with source transforms, but it would be good to
1047 also perform it in IR1 optimization phase.
1049 290: Alpha floating point and denormalized traps
1050 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1051 hardware or kernel deficiency: the status of the enable/disable
1052 denormalized-float traps bit seems to be ambiguous; by the time we
1053 get to os_restore_fp_control after a trap, denormalized traps seem
1054 to be enabled. Since we don't want a trap every time someone uses a
1055 denormalized float, in general, we mask out that bit when we restore
1056 the control word; however, this clobbers any change the user might
1060 (reported by Adam Warner, sbcl-devel 2003-09-23)
1062 The --load toplevel argument does not perform any sanitization of its
1063 argument. As a result, files with Lisp pathname pattern characters
1064 (#\* or #\?, for instance) or quotation marks can cause the system
1065 to perform arbitrary behaviour.
1068 LOOP with non-constant arithmetic step clauses suffers from overzealous
1069 type constraint: code of the form
1070 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1071 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1072 (0))). However, an integral value of X should be legal, because
1073 successive adds of integers to double-floats produces double-floats,
1074 so none of the type restrictions in the code is violated.
1076 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1077 argument type only after having read a character. This is caused
1078 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1079 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1080 less error prone to have EXPLICIT-CHECK be a local declaration,
1081 being put into the definition, instead of an attribute being kept in
1082 a separate file; maybe also put it into SB-EXT?
1084 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1085 in NOTE-ASSUMED-TYPES
1086 In sbcl-0.8.7.32, compiling the file
1088 (declare (type integer x))
1089 (declare (type (vector (or hash-table bit)) y))
1092 (declare (type integer x))
1093 (declare (type (simple-array base (2)) y))
1096 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1098 302: Undefined type messes up DATA-VECTOR-REF expansion.
1100 (defun dis (s ei x y)
1101 (declare (type (simple-array function (2)) s) (type ei ei))
1102 (funcall (aref s ei) x y))
1103 on sbcl-0.8.7.36/X86/Linux causes a BUG to be signalled:
1104 full call to SB-KERNEL:DATA-VECTOR-REF
1106 303: "nonlinear LVARs" (aka MISC.293)
1108 (multiple-value-call #'list
1110 (multiple-value-prog1
1111 (eval '(values :a :b :c))
1117 (throw 'bar (values 3 4)))))))))))
1119 (BUU 1) returns garbage.
1121 The problem is that both EVALs sequentially write to the same LVAR.
1124 (Reported by Dave Roberts.)
1125 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1128 (declare (ftype (function () (integer 0 10)) fee)
1132 uses generic arithmetic with INLINE and fixnum without.
1134 306: "Imprecise unions of array types"
1136 (declare (optimize speed)
1137 (type (or (array cons) (array vector)) x))
1139 (foo #((0))) => TYPE-ERROR
1146 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1147 collect `(array ,(sb-vm:saetp-specifier x)))))
1148 => NIL, T (when it should be T, T)
1150 308: "Characters without names"
1151 (reported by Bruno Haible sbcl-devel "character names are missing"
1153 (graphic-char-p (code-char 255))
1155 (char-name (code-char 255))
1158 SBCL is unsure of what to do about characters with codes in the
1159 range 128-255. Currently they are treated as non-graphic, but don't
1160 have names, which is not compliant with the standard. Various fixes
1161 are possible, such as
1162 * giving them names such as NON-ASCII-128;
1163 * reducing CHAR-CODE-LIMIT to 127 (almost certainly unpopular);
1164 * making the characters graphic (makes a certain amount of sense);
1165 * biting the bullet and implementing Unicode (probably quite hard).
1167 309: "Dubious values for implementation limits"
1168 (reported by Bruno Haible sbcl-devel "Incorrect value of
1169 multiple-values-limit" 2004-04-19)
1170 (values-list (make-list 1000000)), on x86/linux, signals a stack
1171 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1172 significantly larger than 1000000. There are probably similar
1173 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1174 around the same time regarding a call to LIST on sparc with 1000
1175 arguments) and other implementation limit constants.
1177 311: "Tokeniser not thread-safe"
1178 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1180 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1181 spurious errors should two threads attempt to tokenise at the same
1184 314: "LOOP :INITIALLY clauses and scope of initializers"
1185 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1186 test suite, originally by Thomas F. Burdick.
1187 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1188 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1189 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1190 ;; initially clauses are evaluated in the loop prologue, which precedes all
1191 ;; loop code except for the initial settings provided by with, for, or as.
1192 (loop :for x = 0 :then (1+ x)
1193 :for y = (1+ x) :then (ash y 1)
1194 :for z :across #(1 3 9 27 81 243)
1196 :initially (assert (zerop x)) :initially (assert (= 2 w))
1197 :until (>= w 100) :collect w)
1198 Expected: (2 6 15 38)
1201 317: "FORMAT of floating point numbers"
1202 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1204 (format nil "~1F" 10) => "0." ; "10." expected
1205 (format nil "~0F" 10) => "0." ; "10." expected
1206 (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected
1207 it would be nice if whatever fixed this also untangled the two
1208 competing implementations of floating point printing (Steele and
1209 White, and Burger and Dybvig) present in src/code/print.lisp
1211 318: "stack overflow in compiler warning with redefined class"
1212 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1214 (setq *print-pretty* nil)
1216 (setf (find-class 'foo) nil)
1217 (defstruct foo slot-1)
1219 ...#<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTUREControl stack guard page temporarily disabled: proceed with caution
1220 (it's not really clear what it should give: is (SETF FIND-CLASS)
1221 meant to be enough to delete structure classes from the system?
1222 Giving a stack overflow is definitely suboptimal, though.)
1224 319: "backquote with comma inside array"
1225 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1227 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1229 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1230 which probably isn't intentional.
1232 323: "REPLACE, BIT-BASH and large strings"
1233 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1234 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1236 (declare (optimize speed (safety 1)))
1237 (let ((x (make-string 140000000))
1238 (y (make-string 140000000)))
1239 (length (replace x y))))
1242 debugger invoked on a TYPE-ERROR in thread 2412:
1243 The value 1120000000 is not of type (MOD 536870911).
1244 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1246 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1247 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1248 for all positive integral <x>. At present, it only works for <x> up
1249 to about 1024 (and similarly for signed-byte), so
1250 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1251 gives an error in sbcl-0.8.10.
1253 325: "CLOSE :ABORT T on supeseding streams"
1254 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1255 file on disk, even if one existed before opening.
1257 The illegality of this is not crystal clear, as the ANSI dictionary
1258 entry for CLOSE says that when :ABORT is T superseded files are not
1259 superseded (ie. the original should be restored), whereas the OPEN
1260 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1261 implementation should not destroy the old file until the new stream
1262 is closed." -- implying that even though undesirable, early deletion
1263 is legal. Restoring the original would none the less be the polite
1266 326: "*PRINT-CIRCLE* crosstalk between streams"
1267 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1268 mixed between streams when output operations are intermingled closely
1269 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1270 test case below), so that e.g. the references #2# appears on a stream
1271 with no preceding #2= on that stream to define it (because the #2= was
1272 sent to another stream).
1273 (cl:in-package :cl-user)
1274 (defstruct foo index)
1275 (defparameter *foo* (make-foo :index 4))
1277 (defparameter *bar* (make-bar))
1278 (defparameter *tangle* (list *foo* *bar* *foo*))
1279 (defmethod print-object ((foo foo) stream)
1280 (let ((index (foo-index foo)))
1281 (format *trace-output*
1282 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1284 (format stream "[FOO ~D]" index))
1286 (let ((tsos (make-string-output-stream))
1287 (ssos (make-string-output-stream)))
1288 (let ((*print-circle* t)
1289 (*trace-output* tsos)
1290 (*standard-output* ssos))
1291 (prin1 *tangle* *standard-output*))
1292 (let ((string (get-output-stream-string ssos)))
1293 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1294 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1295 (error "oops: ~S" string)))))
1296 It might be straightforward to fix this by turning the
1297 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1298 per-stream slots, but (1) it would probably be sort of messy faking
1299 up the special variable binding semantics using UNWIND-PROTECT and
1300 (2) it might be sort of a pain to test that no other bugs had been
1303 328: "Profiling generic functions", transplanted from #241
1304 (from tonyms on #lisp IRC 2003-02-25)
1305 In sbcl-0.7.12.55, typing
1306 (defclass foo () ((bar :accessor foo-bar)))
1309 (defclass foo () ((bar :accessor foo-bar)))
1310 gives the error message
1311 "#:FOO-BAR already names an ordinary function or a macro."
1313 Problem: when a generic function is profiled, it appears as an ordinary
1314 function to PCL. (Remembering the uninterned accessor is OK, as the
1315 redefinition must be able to remove old accessors from their generic
1318 329: "Sequential class redefinition"
1319 reported by Bruno Haible:
1320 (defclass reactor () ((max-temp :initform 10000000)))
1321 (defvar *r1* (make-instance 'reactor))
1322 (defvar *r2* (make-instance 'reactor))
1323 (slot-value *r1* 'max-temp)
1324 (slot-value *r2* 'max-temp)
1325 (defclass reactor () ((uptime :initform 0)))
1326 (slot-value *r1* 'uptime)
1327 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1328 (slot-value *r1* 'max-temp) ; => 10000
1329 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1332 The method effective when the wrapper is obsoleted can be saved
1333 in the wrapper, and then to update the instance just run through
1334 all the old wrappers in order from oldest to newest.
1336 331: "lazy creation of CLOS classes for user-defined conditions"
1338 (defstruct (bar (:include foo)))
1339 (sb-mop:class-direct-subclasses (find-class 'foo))
1340 returns NIL, rather than a singleton list containing the BAR class.
1342 332: "fasl stack inconsistency in structure redefinition"
1343 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1344 Even though structure redefinition is undefined by the standard, the
1345 following behaviour is suboptimal: running
1346 (defun stimulate-sbcl ()
1347 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1348 ;;create a file which redefines a structure incompatibly
1349 (with-open-file (f filename :direction :output :if-exists :supersede)
1350 (print '(defstruct astruct foo) f)
1351 (print '(defstruct astruct foo bar) f))
1352 ;;compile and load the file, then invoke the continue restart on
1353 ;;the structure redefinition error
1354 (handler-bind ((error (lambda (c) (continue c))))
1355 (load (compile-file filename)))))
1357 and choosing the CONTINUE restart yields the message
1358 debugger invoked on a SB-INT:BUG in thread 27726:
1359 fasl stack not empty when it should be
1361 336: "slot-definitions must retain the generic functions of accessors"
1362 reported by Tony Martinez:
1363 (defclass foo () ((bar :reader foo-bar)))
1364 (defun foo-bar (x) x)
1365 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1367 Note: just punting the accessor removal if the fdefinition
1368 is not a generic function is not enough:
1370 (defclass foo () ((bar :reader foo-bar)))
1371 (defvar *reader* #'foo-bar)
1372 (defun foo-bar (x) x)
1373 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1374 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1375 ; the method must be removed
1376 ; by the class redefinition
1378 Fixing this should also fix a subset of #328 -- update the
1379 description with a new test-case then.
1381 337: MAKE-METHOD and user-defined method classes
1382 (reported by Bruno Haible sbcl-devel 2004-06-11)
1386 (defclass user-method (standard-method) (myslot))
1387 (defmacro def-user-method (name &rest rest)
1388 (let* ((lambdalist-position (position-if #'listp rest))
1389 (qualifiers (subseq rest 0 lambdalist-position))
1390 (lambdalist (elt rest lambdalist-position))
1391 (body (subseq rest (+ lambdalist-position 1)))
1393 (subseq lambdalist 0 (or
1395 (lambda (x) (member x lambda-list-keywords))
1397 (length lambdalist))))
1398 (specializers (mapcar #'find-class
1399 (mapcar (lambda (x) (if (consp x) (second x) t))
1401 (unspecialized-required-part
1402 (mapcar (lambda (x) (if (consp x) (first x) x)) required-part))
1403 (unspecialized-lambdalist
1404 (append unspecialized-required-part
1405 (subseq lambdalist (length required-part)))))
1408 (MAKE-INSTANCE 'USER-METHOD
1409 :QUALIFIERS ',qualifiers
1410 :LAMBDA-LIST ',unspecialized-lambdalist
1411 :SPECIALIZERS ',specializers
1413 (LAMBDA (ARGUMENTS NEXT-METHODS-LIST)
1414 (FLET ((NEXT-METHOD-P () NEXT-METHODS-LIST)
1415 (CALL-NEXT-METHOD (&REST NEW-ARGUMENTS)
1416 (UNLESS NEW-ARGUMENTS (SETQ NEW-ARGUMENTS ARGUMENTS))
1417 (IF (NULL NEXT-METHODS-LIST)
1418 (ERROR "no next method for arguments ~:S" ARGUMENTS)
1419 (FUNCALL (SB-PCL:METHOD-FUNCTION
1420 (FIRST NEXT-METHODS-LIST))
1421 NEW-ARGUMENTS (REST NEXT-METHODS-LIST)))))
1422 (APPLY #'(LAMBDA ,unspecialized-lambdalist ,@body) ARGUMENTS)))))
1426 (defgeneric test-um03 (x))
1427 (defmethod test-um03 ((x integer))
1428 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1429 (def-user-method test-um03 ((x rational))
1430 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1431 (defmethod test-um03 ((x real))
1432 (list 'real x (not (null (next-method-p)))))
1437 (defgeneric test-um10 (x))
1438 (defmethod test-um10 ((x integer))
1439 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1440 (defmethod test-um10 ((x rational))
1441 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1442 (defmethod test-um10 ((x real))
1443 (list 'real x (not (null (next-method-p)))))
1444 (defmethod test-um10 :after ((x real)))
1445 (def-user-method test-um10 :around ((x integer))
1446 (list* 'around-integer x
1447 (not (null (next-method-p))) (call-next-method)))
1448 (defmethod test-um10 :around ((x rational))
1449 (list* 'around-rational x
1450 (not (null (next-method-p))) (call-next-method)))
1451 (defmethod test-um10 :around ((x real))
1452 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1454 fails with a type error, and
1457 (defgeneric test-um12 (x))
1458 (defmethod test-um12 ((x integer))
1459 (list* 'integer x (not (null (next-method-p))) (call-next-method)))
1460 (defmethod test-um12 ((x rational))
1461 (list* 'rational x (not (null (next-method-p))) (call-next-method)))
1462 (defmethod test-um12 ((x real))
1463 (list 'real x (not (null (next-method-p)))))
1464 (defmethod test-um12 :after ((x real)))
1465 (defmethod test-um12 :around ((x integer))
1466 (list* 'around-integer x
1467 (not (null (next-method-p))) (call-next-method)))
1468 (defmethod test-um12 :around ((x rational))
1469 (list* 'around-rational x
1470 (not (null (next-method-p))) (call-next-method)))
1471 (def-user-method test-um12 :around ((x real))
1472 (list* 'around-real x (not (null (next-method-p))) (call-next-method)))
1474 fails with NO-APPLICABLE-METHOD.
1476 339: "DEFINE-METHOD-COMBINATION bugs"
1477 (reported by Bruno Haible via the clisp test suite)
1479 a. Syntax checking laxity (should produce errors):
1480 i. (define-method-combination foo :documentation :operator)
1481 ii. (define-method-combination foo :documentation nil)
1482 iii. (define-method-combination foo nil)
1483 iv. (define-method-combination foo nil nil
1484 (:arguments order &aux &key))
1485 v. (define-method-combination foo nil nil (:arguments &whole))
1486 vi. (define-method-combination foo nil nil (:generic-function))
1487 vii. (define-method-combination foo nil nil (:generic-function bar baz))
1488 viii. (define-method-combination foo nil nil (:generic-function (bar)))
1489 ix. (define-method-combination foo nil ((3)))
1490 x. (define-method-combination foo nil ((a)))
1492 b. define-method-combination arguments lambda list badness
1493 i. &aux args are currently unsupported;
1494 ii. default values of &optional and &key arguments are ignored;
1495 iii. supplied-p variables for &optional and &key arguments are not
1498 c. qualifier matching incorrect
1500 (define-method-combination mc27 ()
1502 (ignored (:ignore :unused)))
1504 ,@(mapcar #'(lambda (method) `(call-method ,method)) normal)))
1505 (defgeneric test-mc27 (x)
1506 (:method-combination mc27)
1507 (:method :ignore ((x number)) (/ 0)))
1510 should signal an invalid-method-error, as the :IGNORE (NUMBER)
1511 method is applicable, and yet matches neither of the method group
1514 340: SETF of VALUES using too many values
1515 (fixed in sbcl-0.8.12.10)
1517 341: PPRINT-LOGICAL-BLOCK / PPRINT-FILL / PPRINT-LINEAR sharing detection.
1518 (from Paul Dietz' test suite)
1520 CLHS on PPRINT-LINEAR and PPRINT-FILL (and PPRINT-TABULAR, though
1521 that's slightly different) states that these functions perform
1522 circular and shared structure detection on their object. Therefore,
1524 a.(let ((*print-circle* t))
1525 (pprint-linear *standard-output* (let ((x '(a))) (list x x))))
1526 should print "(#1=(A) #1#)"
1528 b.(let ((*print-circle* t))
1529 (pprint-linear *standard-output*
1530 (let ((x (cons nil nil))) (setf (cdr x) x) x)))
1531 should print "#1=(NIL . #1#)"
1533 (it is likely that the fault lies in PPRINT-LOGICAL-BLOCK, as
1534 suggested by the suggested implementation of PPRINT-TABULAR)
1536 342: PPRINT-TABULAR / PPRINT-LOGICAL-BLOCK logical block start position
1537 The logical block introduced by PPRINT-LOGICAL-BLOCK should not
1538 include the prefix, so that
1539 (pprint-tabular *standard-output* '(1 2 3) t nil 2)
1541 "(1 2 3)" rather than "(1 2 3)".
1543 343: MOP:COMPUTE-DISCRIMINATING-FUNCTION overriding causes error
1544 Even the simplest possible overriding of
1545 COMPUTE-DISCRIMINATING-FUNCTION, suggested in the PCL implementation
1546 as "canonical", does not work:
1547 (defclass my-generic-function (standard-generic-function) ()
1548 (:metaclass funcallable-standard-class))
1549 (defmethod compute-discriminating-function ((gf my-generic-function))
1550 (let ((dfun (call-next-method)))
1551 (lambda (&rest args)
1552 (apply dfun args))))
1554 (:generic-function-class my-generic-function))
1555 (defmethod foo (x) (+ x x))
1557 signals an error. This error is the same even if the LAMBDA is
1558 replaced by (FUNCTION (SB-KERNEL:INSTANCE-LAMBDA ...)). Maybe the
1559 SET-FUNCALLABLE-INSTANCE-FUN scary stuff in
1560 src/code/target-defstruct.lisp is broken? This seems to be broken
1561 in CMUCL 18e, so it's not caused by a recent change.
1563 344: more (?) ROOM T problems (possibly part of bug 108)
1564 In sbcl-0.8.12.51, and off and on leading up to it, the
1565 SB!VM:MEMORY-USAGE operations in ROOM T caused
1566 unhandled condition (of type SB-INT:BUG):
1567 failed AVER: "(SAP= CURRENT END)"
1568 Several clever people have taken a shot at this without fixing
1569 it; this time around (before sbcl-0.8.13 release) I (WHN) just
1570 commented out the SB!VM:MEMORY-USAGE calls until someone figures
1571 out how to make them work reliably with the rest of the GC.