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 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
195 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
196 an easily guessable temporary filename in a way which might open
197 applications using LOAD-FOREIGN to hijacking by malicious users
198 on the same machine. Incantations for doing this safely are
199 floating around the net in various "how to write secure programs
200 despite Unix" documents, and it would be good to (1) fix this in
201 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
202 files and make it share the same new safe logic.
204 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
205 make the temporary filename less easily guessable)
208 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
209 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
210 the mean of the distribution can be systematically O(0.1%) wrong.
211 Just increasing R-I-E-B is probably not a good solution, since
212 it would decrease efficiency more than is probably necessary. Perhaps
213 using some sort of accept/reject method would be better.
216 Internally the compiler sometimes evaluates
217 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
218 (I stumbled across this when I added an
219 (assert (not (eq type1 *wild-type*)))
220 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
221 in a type context should probably be translated to T, and so it's
222 probably wrong to ask whether it's equal to the T type and then (using
223 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
224 (I haven't tried to investigate this bug enough to guess whether
225 there might be any user-level symptoms.)
227 In fact, the type system is likely to depend on this inequality not
228 holding... * is not equivalent to T in many cases, such as
229 (VECTOR *) /= (VECTOR T).
232 The facility for dumping a running Lisp image to disk gets confused
233 when run without the PURIFY option, and creates an unnecessarily large
234 core file (apparently representing memory usage up to the previous
235 high-water mark). Moreover, when the file is loaded, it confuses the
236 GC, so that thereafter memory usage can never be reduced below that
239 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
240 The interaction between gencgc and the variables used by
241 save-lisp-and-die is still nonoptimal, though, so no respite from
245 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
246 CL), out-of-line structure slot setters are horribly inefficient
247 whenever the type of the slot is declared, because out-of-line
248 structure slot setters are implemented as closures to save space,
249 so the compiler doesn't compile the type test into code, but
250 instead just saves the type in a lexical closure and interprets it
252 To exercise the problem, compile and load
253 (cl:in-package :cl-user)
255 (bar (error "missing") :type bar))
258 (loop (setf (foo-bar *foo*) x)))
260 (defvar *bar* (make-bar))
261 (defvar *foo* (make-foo :bar *bar*))
262 (defvar *setf-foo-bar* #'(setf foo-bar))
264 (loop (funcall *setf-foo-bar* x *foo*)))
265 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
266 use BACKTRACE, to see it's spending all essentially all its time
267 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
268 One possible solution would be simply to give up on
269 representing structure slot accessors as functions, and represent
270 them as macroexpansions instead. This can be inconvenient for users,
271 but it's not clear that it's worse than trying to help by expanding
272 into a horribly inefficient implementation.
273 As a workaround for the problem, #'(SETF FOO) expressions
274 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
275 (defmacro efficient-setf-function (place-function-name)
276 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
277 ;; a workaround for the problem, encouraging the
278 ;; inline expansion of the structure accessor, so
279 ;; that the compiler can optimize its type test
280 (let ((new-value (gensym "NEW-VALUE-"))
281 (structure-value (gensym "STRUCTURE-VALUE-")))
282 `(lambda (,new-value ,structure-value)
283 (setf (,place-function-name ,structure-value)
285 ;; no problem, can just use the ordinary expansion
286 `(function (setf ,place-function-name))))
289 There's apparently a bug in CEILING optimization which caused
290 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
291 applied the patches to SBCL and they didn't seem to cause problems
292 (as reported sbcl-devel 2001-05-04). However, since the patches
293 modify nontrivial code which was apparently written incorrectly
294 the first time around, until regression tests are written I'm not
295 comfortable merging the patches in the CVS version of SBCL.
298 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
299 a clean, just-started SBCL system. And it seems to be right:
300 (ROOM T) can bring a small computer to its knees for a *long*
301 time trying to GC afterwards. Surely there's some more economical
302 way to implement (ROOM T).
304 Daniel Barlow doesn't know what fixed this, but observes that it
305 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
306 in a fresh SBCL causes
308 debugger invoked on a SB-INT:BUG in thread 5911:
309 failed AVER: "(SAP= CURRENT END)"
311 unless a GC has happened beforehand.
314 When the compiler inline expands functions, it may be that different
315 kinds of return values are generated from different code branches.
316 E.g. an inline expansion of POSITION generates integer results
317 from one branch, and NIL results from another. When that inline
318 expansion is used in a context where only one of those results
321 (aref *a1* (position x *a2*)))
322 and the compiler can't prove that the unacceptable branch is
323 never taken, then bogus type mismatch warnings can be generated.
324 If you need to suppress the type mismatch warnings, you can
325 suppress the inline expansion,
327 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
328 (aref *a1* (position x *a2*)))
329 or, sometimes, suppress them by declaring the result to be of an
332 (aref *a1* (the integer (position x *a2*))))
334 This is not a new compiler problem in 0.7.0, but the new compiler
335 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
336 more conspicuous. If you don't need performance from these functions,
337 and the bogus warnings are a nuisance for you, you can return to
338 your pre-0.7.0 state of grace with
339 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
344 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
345 the entire lexical environment at the point of MACROLET available
346 in the bodies of the macroexpander functions. In particular, it
347 allows the function bodies (which run at compile time) to try to
348 access lexical variables (which are only defined at runtime).
349 It doesn't even issue a warning, which is bad.
351 The SBCL behavior arguably conforms to the ANSI spec (since the
352 spec says that the behavior is undefined, ergo anything conforms).
353 However, it would be better to issue a compile-time error.
354 Unfortunately I (WHN) don't see any simple way to detect this
355 condition in order to issue such an error, so for the meantime
356 SBCL just does this weird broken "conforming" thing.
358 The ANSI standard says, in the definition of the special operator
360 The macro-expansion functions defined by MACROLET are defined
361 in the lexical environment in which the MACROLET form appears.
362 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
363 the local macro definitions in a MACROLET, but the consequences
364 are undefined if the local macro definitions reference any
365 local variable or function bindings that are visible in that
367 Then it seems to contradict itself by giving the example
369 (macrolet ((fudge (z)
370 ;The parameters x and flag are not accessible
371 ; at this point; a reference to flag would be to
372 ; the global variable of that name.
373 ` (if flag (* ,z ,z) ,z)))
374 ;The parameters x and flag are accessible here.
378 The comment "a reference to flag would be to the global variable
379 of the same name" sounds like good behavior for the system to have.
380 but actual specification quoted above says that the actual behavior
383 (Since 0.7.8.23 macroexpanders are defined in a restricted version
384 of the lexical environment, containing no lexical variables and
385 functions, which seems to conform to ANSI and CLtL2, but signalling
386 a STYLE-WARNING for references to variables similar to locals might
390 (as reported by Gabe Garza on cmucl-help 2001-09-21)
392 (defun test-pred (x y)
396 (func (lambda () x)))
397 (print (eq func func))
398 (print (test-pred func func))
399 (delete func (list func))))
400 Now calling (TEST-CASE) gives output
403 (#<FUNCTION {500A9EF9}>)
404 Evidently Python thinks of the lambda as a code transformation so
405 much that it forgets that it's also an object.
408 Ideally, uninterning a symbol would allow it, and its associated
409 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
410 at least as of sbcl-0.7.0, this isn't the case. Information about
411 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
412 essentially in ordinary (non-weak) hash tables keyed by symbols.
413 Thus, once a system has an entry in this system, it tends to live
414 forever, even when it is uninterned and all other references to it
417 141: "pretty printing and backquote"
420 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
422 c. (reported by Paul F. Dietz)
424 `(LAMBDA (SB-IMPL::BACKQ-COMMA X))
427 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
428 prominent SourceForge web/db bug tracking system, which is
429 unfortunately not a reliable way to get a timely response from
430 the SBCL maintainers)
431 In the course of trying to build a test case for an
432 application error, I encountered this behavior:
433 If you start up sbcl, and then lay on CTRL-C for a
434 minute or two, the lisp process will eventually say:
435 %PRIMITIVE HALT called; the party is over.
436 and throw you into the monitor. If I start up lisp,
437 attach to the process with strace, and then do the same
438 (abusive) thing, I get instead:
439 access failure in heap page not marked as write-protected
440 and the monitor again. I don't know enough to have the
441 faintest idea of what is going on here.
442 This is with sbcl 6.12, uname -a reports:
443 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
444 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
445 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
446 it took more than two minutes (but less than five) for me.
450 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
451 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
452 COMPLEX implementation didn't deal with this, and hasn't been
453 upgraded to do so. (This doesn't seem to be a high priority
454 conformance problem, since seems hard to construct useful code
457 b. (fixed in 0.8.3.43)
460 Floating point errors are reported poorly. E.g. on x86 OpenBSD
463 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
464 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
465 No traps are enabled? How can this be?
466 It should be possible to be much more specific (overflow, division
467 by zero, etc.) and of course the "How can this be?" should be fixable.
469 See also bugs #45.c and #183
472 (reported by Robert E. Brown 2002-04-16)
473 When a function is called with too few arguments, causing the
474 debugger to be entered, the uninitialized slots in the bad call frame
475 seem to cause GCish problems, being interpreted as tagged data even
476 though they're not. In particular, executing ROOM in the
477 debugger at that point causes AVER failures:
480 * (lisp-implementation-version)
486 failed AVER: "(SAP= CURRENT END)"
487 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
488 isn't too surprising since there are many differences in stack
489 implementation and GC conservatism between the X86 and other ports.)
491 This is probably the same bug as 216
494 In sbcl-0.7.3.11, compiling the (illegal) code
495 (in-package :cl-user)
496 (defmethod prove ((uustk uustk))
499 gives the (not terribly clear) error message
501 ; (during macroexpansion of (DEFMETHOD PROVE ...))
502 ; can't get template for (FROB NIL NIL)
503 The problem seems to be that the code walker used by the DEFMETHOD
504 macro is unhappy with the illegal syntax in the method body, and
505 is giving an unclear error message.
508 The compiler sometimes tries to constant-fold expressions before
509 it checks to see whether they can be reached. This can lead to
510 bogus warnings about errors in the constant folding, e.g. in code
513 (WRITE-STRING (> X 0) "+" "0"))
514 compiled in a context where the compiler can prove that X is NIL,
515 and the compiler complains that (> X 0) causes a type error because
516 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
517 caused a full WARNING, which made the bug really annoying because then
518 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
519 code. Since then the warning has been downgraded to STYLE-WARNING,
520 so it's still a bug but at least it's a little less annoying.
522 183: "IEEE floating point issues"
523 Even where floating point handling is being dealt with relatively
524 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
525 accrued-exceptions and current-exceptions part of the fp control
526 word don't seem to bear much relation to reality. E.g. on
530 debugger invoked on condition of type DIVISION-BY-ZERO:
531 arithmetic error DIVISION-BY-ZERO signalled
532 0] (sb-vm::get-floating-point-modes)
534 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
535 :ROUNDING-MODE :NEAREST
536 :CURRENT-EXCEPTIONS NIL
537 :ACCRUED-EXCEPTIONS (:INEXACT)
540 * (sb-vm::get-floating-point-modes)
541 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
542 :ROUNDING-MODE :NEAREST
543 :CURRENT-EXCEPTIONS (:INEXACT)
544 :ACCRUED-EXCEPTIONS (:INEXACT)
547 188: "compiler performance fiasco involving type inference and UNION-TYPE"
551 (declare (optimize (safety 3)))
552 (declare (optimize (compilation-speed 2)))
553 (declare (optimize (speed 1) (debug 1) (space 1)))
555 (declare (type (integer 0) start))
556 (print (incf start 22))
557 (print (incf start 26))
558 (print (incf start 28)))
560 (declare (type (integer 0) start))
561 (print (incf start 22))
562 (print (incf start 26)))
564 (declare (type (integer 0) start))
565 (print (incf start 22))
566 (print (incf start 26))))))
568 This example could be solved with clever enough constraint
569 propagation or with SSA, but consider
574 The careful type of X is {2k} :-(. Is it really important to be
575 able to work with unions of many intervals?
577 191: "Miscellaneous PCL deficiencies"
578 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
579 a. DEFCLASS does not inform the compiler about generated
580 functions. Compiling a file with
584 (WITH-SLOTS (A-CLASS-X) A
586 results in a STYLE-WARNING:
588 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
590 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
591 Mai points out that the declamation of functions is in fact
592 incorrect in some cases (most notably for structure
593 classes). This means that at present erroneous attempts to use
594 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
595 won't get the corresponding STYLE-WARNING.
596 c. (fixed in 0.8.4.23)
598 201: "Incautious type inference from compound types"
599 a. (reported by APD sbcl-devel 2002-09-17)
601 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
603 (FORMAT NIL "~S IS ~S, Y = ~S"
610 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
614 (declare (type (array * (4 4)) x))
616 (setq x (make-array '(4 4)))
617 (adjust-array y '(3 5))
618 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
620 * (foo (make-array '(4 4) :adjustable t))
623 205: "environment issues in cross compiler"
624 (These bugs have no impact on user code, but should be fixed or
626 a. Macroexpanders introduced with MACROLET are defined in the null
628 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
629 the null lexical environment.
630 c. The cross-compiler cannot inline functions defined in a non-null
633 206: ":SB-FLUID feature broken"
634 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
635 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
638 207: "poorly distributed SXHASH results for compound data"
639 SBCL's SXHASH could probably try a little harder. ANSI: "the
640 intent is that an implementation should make a good-faith
641 effort to produce hash-codes that are well distributed
642 within the range of non-negative fixnums". But
643 (let ((hits (make-hash-table)))
646 (let* ((ij (cons i j))
647 (newlist (push ij (gethash (sxhash ij) hits))))
649 (format t "~&collision: ~S~%" newlist))))))
650 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
651 would be an obvious way of fix. Maybe it would be acceptably efficient
652 to redo MIX using a lookup into a 256-entry s-box containing
653 29-bit pseudorandom numbers?
655 211: "keywords processing"
656 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
657 number of keyword arguments.
660 (flet ((foo (&key y) (list y)))
661 (list (foo :y 1 :y 2)))
663 issues confusing message
668 ; caught STYLE-WARNING:
669 ; The variable #:G15 is defined but never used.
671 212: "Sequence functions and circular arguments"
672 COERCE, MERGE and CONCATENATE go into an infinite loop when given
673 circular arguments; it would be good for the user if they could be
674 given an error instead (ANSI 17.1.1 allows this behaviour on the part
675 of the implementation, as conforming code cannot give non-proper
676 sequences to these functions. MAP also has this problem (and
677 solution), though arguably the convenience of being able to do
678 (MAP 'LIST '+ FOO '#1=(1 . #1#))
679 might be classed as more important (though signalling an error when
680 all of the arguments are circular is probably desireable).
682 213: "Sequence functions and type checking"
683 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
684 check that it will return a sequence of the given type. Fixing
685 it along the same lines as the others (cf. work done around
686 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
687 entirely straightforward.
688 c. All of these functions will silently accept a type of the form
690 whether or not the return value is of this type. This is
691 probably permitted by ANSI (see "Exceptional Situations" under
692 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
693 know about this escape clause, so code of the form
694 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
695 can erroneously return T.
697 215: ":TEST-NOT handling by functions"
698 a. FIND and POSITION currently signal errors when given non-NIL for
699 both their :TEST and (deprecated) :TEST-NOT arguments, but by
700 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
701 means that the effect is "unpredictable but harmless". It's not
702 clear what that actually means; it may preclude conforming
703 implementations from signalling errors.
704 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
705 when conflict occurs. As a quality of implementation issue, it
706 might be preferable to treat :TEST and :TEST-NOT as being in some
707 sense the same &KEY, and effectively take the first test function in
709 c. Again, a quality of implementation issue: it would be good to issue a
710 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
711 WARNING for calls with both :TEST and :TEST-NOT; possibly this
712 latter should be WARNed about at execute-time too.
714 216: "debugger confused by frames with invalid number of arguments"
715 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
716 leaves the system confused, enough so that (QUIT) no longer works.
717 It's as though the process of working with the uninitialized slot in
718 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
719 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
721 This is probably the same bug as 162
723 217: "Bad type operations with FUNCTION types"
726 * (values-type-union (specifier-type '(function (base-char)))
727 (specifier-type '(function (integer))))
729 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
731 It causes insertion of wrong type assertions into generated
735 (let ((f (etypecase x
736 (character #'write-char)
737 (integer #'write-byte))))
740 (character (write-char x s))
741 (integer (write-byte x s)))))
743 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
745 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
746 produce invalid code, but type checking is not accurate.)
748 233: bugs in constraint propagation
750 (declaim (optimize (speed 2) (safety 3)))
752 (if (typep (prog1 x (setq x y)) 'double-float)
755 (foo 1d0 5) => segmentation violation
757 235: "type system and inline expansion"
759 (declaim (ftype (function (cons) number) acc))
760 (declaim (inline acc))
762 (the number (car c)))
765 (values (locally (declare (optimize (safety 0)))
767 (locally (declare (optimize (safety 3)))
770 (foo '(nil) '(t)) => NIL, T.
772 237: "Environment arguments to type functions"
773 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
774 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
775 argument, but they ignore it completely. This is almost
776 certainly not correct.
777 b. Also, the compiler's optimizers for TYPEP have not been informed
778 about the new argument; consequently, they will not transform
779 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
780 just as optimizeable as (TYPEP 1 'INTEGER).
782 238: "REPL compiler overenthusiasm for CLOS code"
784 * (defclass foo () ())
785 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
786 causes approximately 100 lines of code deletion notes. Some
787 discussion on this issue happened under the title 'Three "interesting"
788 bugs in PCL', resulting in a fix for this oververbosity from the
789 compiler proper; however, the problem persists in the interactor
790 because the notion of original source is not preserved: for the
791 compiler, the original source of the above expression is (DEFMETHOD
792 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
793 compiler gets its hands on the code needing compilation from the REPL,
794 it has been macroexpanded several times.
796 A symptom of the same underlying problem, reported by Tony Martinez:
798 (with-input-from-string (*query-io* " no")
800 (simple-type-error () 'error))
802 ; (SB-KERNEL:FLOAT-WAIT)
804 ; note: deleting unreachable code
805 ; compilation unit finished
808 242: "WRITE-SEQUENCE suboptimality"
809 (observed from clx performance)
810 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
811 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
812 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
813 rather than writing the sequence in one go, leading to severe
814 performance degradation.
816 243: "STYLE-WARNING overenthusiasm for unused variables"
817 (observed from clx compilation)
818 In sbcl-0.7.14, in the presence of the macros
819 (DEFMACRO FOO (X) `(BAR ,X))
820 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
821 somewhat surprising style warnings are emitted for
822 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
824 ; (LAMBDA (Y) (FOO Y))
826 ; caught STYLE-WARNING:
827 ; The variable Y is defined but never used.
829 245: bugs in disassembler
830 b. On X86 operand size prefix is not recognized.
833 (defun foo (&key (a :x))
837 does not cause a warning. (BTW: old SBCL issued a warning, but for a
838 function, which was never called!)
841 Compiler does not emit warnings for
843 a. (lambda () (svref (make-array 8 :adjustable t) 1))
846 (list (let ((y (the real x)))
847 (unless (floatp y) (error ""))
852 (declare (optimize (debug 0)))
853 (declare (type vector x))
854 (list (fill-pointer x)
858 Complex array type does not have corresponding type specifier.
860 This is a problem because the compiler emits optimization notes when
861 you use a non-simple array, and without a type specifier for hairy
862 array types, there's no good way to tell it you're doing it
863 intentionally so that it should shut up and just compile the code.
865 Another problem is confusing error message "asserted type ARRAY
866 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
867 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
869 The last problem is that when type assertions are converted to type
870 checks, types are represented with type specifiers, so we could lose
871 complex attribute. (Now this is probably not important, because
872 currently checks for complex arrays seem to be performed by
876 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
877 warning. Analogous cases with the index and length being equal and
878 greater than 0 are warned for; the problem here seems to be that the
879 type required for an array reference of this type is (INTEGER 0 (0))
880 which is canonicalized to NIL.
885 (t1 (specifier-type s)))
886 (eval `(defstruct ,s))
887 (type= t1 (specifier-type s)))
892 b. The same for CSUBTYPEP.
894 262: "yet another bug in inline expansion of local functions"
898 (declare (integer x y))
901 (declare (integer u))
902 (if (> (1+ (the unsigned-byte u)) 0)
904 (return (+ 38 (cos (/ u 78)))))))
905 (declare (inline xyz))
907 (* (funcall (eval #'xyz) x)
909 (funcall (if (> x 5) #'xyz #'identity)
914 Urgh... It's time to write IR1-copier.
917 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
918 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
919 looks reasonable, if not obviously correct; however, it caused the
920 PPC/Linux port to segfault during warm-init while loading
921 src/pcl/std-class.fasl. A workaround patch was made, but it would
922 be nice to understand why the first patch caused problems, and to
923 fix the cause if possible.
925 268: "wrong free declaration scope"
926 The following code must signal type error:
928 (locally (declare (optimize (safety 3)))
929 (flet ((foo (x &optional (y (car x)))
930 (declare (optimize (safety 0)))
932 (funcall (eval #'foo) 1)))
935 SCALE-FLOAT should accept any integer for its second argument.
938 In the following function constraint propagator optimizes nothing:
941 (declare (integer x))
942 (declare (optimize speed))
950 Compilation of the following two forms causes "X is unbound" error:
952 (symbol-macrolet ((x pi))
953 (macrolet ((foo (y) (+ x y)))
954 (declaim (inline bar))
960 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
963 CLHS says that type declaration of a symbol macro should not affect
964 its expansion, but in SBCL it does. (If you like magic and want to
965 fix it, don't forget to change all uses of MACROEXPAND to
969 The following code (taken from CLOCC) takes a lot of time to compile:
972 (declare (type (integer 0 #.large-constant) n))
975 (fixed in 0.8.2.51, but a test case would be good)
978 (defmethod fee ((x fixnum))
981 (fee 1) => type error
988 (declare (optimize speed))
989 (loop for i of-type (integer 0) from 0 by 2 below 10
992 uses generic arithmetic.
994 b. (fixed in 0.8.3.6)
996 279: type propagation error -- correctly inferred type goes astray?
997 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
998 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
999 &OPTIONAL), not a (INTEGER 1 536870911)
1000 is emitted when compiling this file:
1001 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
1002 (integer #.most-negative-fixnum 0))
1007 (let* (;; Uncomment this for a type mismatch warning indicating
1008 ;; that the type of (FOO X) is correctly understood.
1009 #+nil (fs-foo (float-sign (foo x)))
1010 ;; Uncomment this for a type mismatch warning
1011 ;; indicating that the type of (ABS (FOO X)) is
1012 ;; correctly understood.
1013 #+nil (fs-abs-foo (float-sign (abs (foo x))))
1014 ;; something wrong with this one though
1015 (abs-foo (abs (foo x))))
1016 (declare (type (integer 1 100) abs-foo))
1021 281: COMPUTE-EFFECTIVE-METHOD error signalling.
1022 (slightly obscured by a non-0 default value for
1023 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
1024 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
1025 when it finds a method with invalid qualifiers. However, it
1026 shouldn't signal errors when any such methods are not applicable to
1027 the particular call being evaluated, and certainly it shouldn't when
1028 simply precomputing effective methods that may never be called.
1029 (setf sb-pcl::*max-emf-precompute-methods* 0)
1031 (:method-combination +)
1032 (:method ((x symbol)) 1)
1033 (:method + ((x number)) x))
1034 (foo 1) -> ERROR, but should simply return 1
1036 The issue seems to be that construction of a discriminating function
1037 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
1039 283: Thread safety: libc functions
1040 There are places that we call unsafe-for-threading libc functions
1041 that we should find alternatives for, or put locks around. Known or
1042 strongly suspected problems, as of 0.8.3.10: please update this
1043 bug instead of creating new ones
1045 localtime() - called for timezone calculations in code/time.lisp
1047 284: Thread safety: special variables
1048 There are lots of special variables in SBCL, and I feel sure that at
1049 least some of them are indicative of potentially thread-unsafe
1050 parts of the system. See doc/internals/notes/threading-specials
1052 286: "recursive known functions"
1053 Self-call recognition conflicts with known function
1054 recognition. Currently cross compiler and target COMPILE do not
1055 recognize recursion, and in target compiler it can be disabled. We
1056 can always disable it for known functions with RECURSIVE attribute,
1057 but there remains a possibility of a function with a
1058 (tail)-recursive simplification pass and transforms/VOPs for base
1061 287: PPC/Linux miscompilation or corruption in first GC
1062 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
1063 segmentation fault is reported at the point of first triggered GC,
1064 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
1065 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
1066 that this merely obscures, not solves, the underlying problem; as and when
1067 underlying problems are fixed, it would be worth trying again to provoke
1070 288: fundamental cross-compilation issues (from old UGLINESS file)
1071 Using host floating point numbers to represent target floating point
1072 numbers, or host characters to represent target characters, is
1073 theoretically shaky. (The characters are OK as long as the characters
1074 are in the ANSI-guaranteed character set, though, so they aren't a
1075 real problem as long as the sources don't need anything but that;
1076 the floats are a real problem.)
1078 289: "type checking and source-transforms"
1080 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1083 Our policy is to check argument types at the moment of a call. It
1084 disagrees with ANSI, which says that type assertions are put
1085 immediately onto argument expressions, but is easier to implement in
1086 IR1 and is more compatible to type inference, inline expansion,
1087 etc. IR1-transforms automatically keep this policy, but source
1088 transforms for associative functions (such as +), being applied
1089 during IR1-convertion, do not. It may be tolerable for direct calls
1090 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1092 b. Another aspect of this problem is efficiency. [x y + z +]
1093 requires less registers than [x y z + +]. This transformation is
1094 currently performed with source transforms, but it would be good to
1095 also perform it in IR1 optimization phase.
1097 290: Alpha floating point and denormalized traps
1098 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1099 hardware or kernel deficiency: the status of the enable/disable
1100 denormalized-float traps bit seems to be ambiguous; by the time we
1101 get to os_restore_fp_control after a trap, denormalized traps seem
1102 to be enabled. Since we don't want a trap every time someone uses a
1103 denormalized float, in general, we mask out that bit when we restore
1104 the control word; however, this clobbers any change the user might
1108 (reported by Adam Warner, sbcl-devel 2003-09-23)
1110 The --load toplevel argument does not perform any sanitization of its
1111 argument. As a result, files with Lisp pathname pattern characters
1112 (#\* or #\?, for instance) or quotation marks can cause the system
1113 to perform arbitrary behaviour.
1116 LOOP with non-constant arithmetic step clauses suffers from overzealous
1117 type constraint: code of the form
1118 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1119 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1120 (0))). However, an integral value of X should be legal, because
1121 successive adds of integers to double-floats produces double-floats,
1122 so none of the type restrictions in the code is violated.
1124 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1125 argument type only after having read a character. This is caused
1126 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1127 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1128 less error prone to have EXPLICIT-CHECK be a local declaration,
1129 being put into the definition, instead of an attribute being kept in
1130 a separate file; maybe also put it into SB-EXT?
1132 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1133 in NOTE-ASSUMED-TYPES
1134 In sbcl-0.8.7.32, compiling the file
1136 (declare (type integer x))
1137 (declare (type (vector (or hash-table bit)) y))
1140 (declare (type integer x))
1141 (declare (type (simple-array base (2)) y))
1144 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1146 302: Undefined type messes up DATA-VECTOR-REF expansion.
1148 (defun dis (s ei x y)
1149 (declare (type (simple-array function (2)) s) (type ei ei))
1150 (funcall (aref s ei) x y))
1151 on sbcl-0.8.7.36/X86/Linux causes a BUG to be signalled:
1152 full call to SB-KERNEL:DATA-VECTOR-REF
1154 303: "nonlinear LVARs" (aka MISC.293)
1156 (multiple-value-call #'list
1158 (multiple-value-prog1
1159 (eval '(values :a :b :c))
1165 (throw 'bar (values 3 4)))))))))))
1167 (BUU 1) returns garbage.
1169 The problem is that both EVALs sequentially write to the same LVAR.
1172 (Reported by Dave Roberts.)
1173 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1176 (declare (ftype (function () (integer 0 10)) fee)
1180 uses generic arithmetic with INLINE and fixnum without.
1182 306: "Imprecise unions of array types"
1184 (declare (optimize speed)
1185 (type (or (array cons) (array vector)) x))
1187 (foo #((0))) => TYPE-ERROR
1194 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1195 collect `(array ,(sb-vm:saetp-specifier x)))))
1196 => NIL, T (when it should be T, T)
1198 308: "Characters without names"
1199 (reported by Bruno Haible sbcl-devel "character names are missing"
1201 (graphic-char-p (code-char 255))
1203 (char-name (code-char 255))
1206 SBCL is unsure of what to do about characters with codes in the
1207 range 128-255. Currently they are treated as non-graphic, but don't
1208 have names, which is not compliant with the standard. Various fixes
1209 are possible, such as
1210 * giving them names such as NON-ASCII-128;
1211 * reducing CHAR-CODE-LIMIT to 127 (almost certainly unpopular);
1212 * making the characters graphic (makes a certain amount of sense);
1213 * biting the bullet and implementing Unicode (probably quite hard).
1215 309: "Dubious values for implementation limits"
1216 (reported by Bruno Haible sbcl-devel "Incorrect value of
1217 multiple-values-limit" 2004-04-19)
1218 (values-list (make-list 1000000)), on x86/linux, signals a stack
1219 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1220 significantly larger than 1000000. There are probably similar
1221 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1222 around the same time regarding a call to LIST on sparc with 1000
1223 arguments) and other implementation limit constants.
1225 311: "Tokeniser not thread-safe"
1226 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1228 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1229 spurious errors should two threads attempt to tokenise at the same
1232 314: "LOOP :INITIALLY clauses and scope of initializers"
1233 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1234 test suite, originally by Thomas F. Burdick.
1235 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1236 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1237 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1238 ;; initially clauses are evaluated in the loop prologue, which precedes all
1239 ;; loop code except for the initial settings provided by with, for, or as.
1240 (loop :for x = 0 :then (1+ x)
1241 :for y = (1+ x) :then (ash y 1)
1242 :for z :across #(1 3 9 27 81 243)
1244 :initially (assert (zerop x)) :initially (assert (= 2 w))
1245 :until (>= w 100) :collect w)
1246 Expected: (2 6 15 38)
1249 317: "FORMAT of floating point numbers"
1250 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1252 (format nil "~1F" 10) => "0." ; "10." expected
1253 (format nil "~0F" 10) => "0." ; "10." expected
1254 (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected
1255 it would be nice if whatever fixed this also untangled the two
1256 competing implementations of floating point printing (Steele and
1257 White, and Burger and Dybvig) present in src/code/print.lisp
1259 318: "stack overflow in compiler warning with redefined class"
1260 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1262 (setq *print-pretty* nil)
1264 (setf (find-class 'foo) nil)
1265 (defstruct foo slot-1)
1267 ...#<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTUREControl stack guard page temporarily disabled: proceed with caution
1268 (it's not really clear what it should give: is (SETF FIND-CLASS)
1269 meant to be enough to delete structure classes from the system?
1270 Giving a stack overflow is definitely suboptimal, though.)
1272 319: "backquote with comma inside array"
1273 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1275 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1277 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1278 which probably isn't intentional.
1280 323: "REPLACE, BIT-BASH and large strings"
1281 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1282 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1284 (declare (optimize speed (safety 1)))
1285 (let ((x (make-string 140000000))
1286 (y (make-string 140000000)))
1287 (length (replace x y))))
1290 debugger invoked on a TYPE-ERROR in thread 2412:
1291 The value 1120000000 is not of type (MOD 536870911).
1292 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1294 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1295 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1296 for all positive integral <x>. At present, it only works for <x> up
1297 to about 1024 (and similarly for signed-byte), so
1298 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1299 gives an error in sbcl-0.8.10.
1301 325: "CLOSE :ABORT T on supeseding streams"
1302 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1303 file on disk, even if one existed before opening.
1305 The illegality of this is not crystal clear, as the ANSI dictionary
1306 entry for CLOSE says that when :ABORT is T superseded files are not
1307 superseded (ie. the original should be restored), whereas the OPEN
1308 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1309 implementation should not destroy the old file until the new stream
1310 is closed." -- implying that even though undesirable, early deletion
1311 is legal. Restoring the original would none the less be the polite
1314 326: "*PRINT-CIRCLE* crosstalk between streams"
1315 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1316 mixed between streams when output operations are intermingled closely
1317 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1318 test case below), so that e.g. the references #2# appears on a stream
1319 with no preceding #2= on that stream to define it (because the #2= was
1320 sent to another stream).
1321 (cl:in-package :cl-user)
1322 (defstruct foo index)
1323 (defparameter *foo* (make-foo :index 4))
1325 (defparameter *bar* (make-bar))
1326 (defparameter *tangle* (list *foo* *bar* *foo*))
1327 (defmethod print-object ((foo foo) stream)
1328 (let ((index (foo-index foo)))
1329 (format *trace-output*
1330 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1332 (format stream "[FOO ~D]" index))
1334 (let ((tsos (make-string-output-stream))
1335 (ssos (make-string-output-stream)))
1336 (let ((*print-circle* t)
1337 (*trace-output* tsos)
1338 (*standard-output* ssos))
1339 (prin1 *tangle* *standard-output*))
1340 (let ((string (get-output-stream-string ssos)))
1341 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1342 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1343 (error "oops: ~S" string))))
1344 It might be straightforward to fix this by turning the
1345 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1346 per-stream slots, but (1) it would probably be sort of messy faking
1347 up the special variable binding semantics using UNWIND-PROTECT and
1348 (2) it might be sort of a pain to test that no other bugs had been
1351 328: "Profiling generic functions", transplanted from #241
1352 (from tonyms on #lisp IRC 2003-02-25)
1353 In sbcl-0.7.12.55, typing
1354 (defclass foo () ((bar :accessor foo-bar)))
1357 (defclass foo () ((bar :accessor foo-bar)))
1358 gives the error message
1359 "#:FOO-BAR already names an ordinary function or a macro."
1361 Problem: when a generic function is profiled, it appears as an ordinary
1362 function to PCL. (Remembering the uninterned accessor is OK, as the
1363 redefinition must be able to remove old accessors from their generic
1366 329: "Sequential class redefinition"
1367 reported by Bruno Haible:
1368 (defclass reactor () ((max-temp :initform 10000000)))
1369 (defvar *r1* (make-instance 'reactor))
1370 (defvar *r2* (make-instance 'reactor))
1371 (slot-value *r1* 'max-temp)
1372 (slot-value *r2* 'max-temp)
1373 (defclass reactor () ((uptime :initform 0)))
1374 (slot-value *r1* 'uptime)
1375 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1376 (slot-value *r1* 'max-temp) ; => 10000
1377 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1380 The method effective when the wrapper is obsoleted can be saved
1381 in the wrapper, and then to update the instance just run through
1382 all the old wrappers in order from oldest to newest.
1384 331: "lazy creation of CLOS classes for user-defined conditions"
1386 (defstruct (bar (:include foo)))
1387 (sb-mop:class-direct-subclasses (find-class 'foo))
1388 returns NIL, rather than a singleton list containing the BAR class.
1390 332: "fasl stack inconsistency in structure redefinition"
1391 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1392 Even though structure redefinition is undefined by the standard, the
1393 following behaviour is suboptimal: running
1394 (defun stimulate-sbcl ()
1395 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1396 ;;create a file which redefines a structure incompatibly
1397 (with-open-file (f filename :direction :output :if-exists :supersede)
1398 (print '(defstruct astruct foo) f)
1399 (print '(defstruct astruct foo bar) f))
1400 ;;compile and load the file, then invoke the continue restart on
1401 ;;the structure redefinition error
1402 (handler-bind ((error (lambda (c) (continue c))))
1403 (load (compile-file filename)))))
1405 and choosing the CONTINUE restart yields the message
1406 debugger invoked on a SB-INT:BUG in thread 27726:
1407 fasl stack not empty when it should be
1409 333: "CHECK-TYPE TYPE-ERROR-DATUM place"
1410 (reported by Tony Martinez sbcl-devel 2004-05-23)
1411 When CHECK-TYPE signals a TYPE-ERROR, the TYPE-ERROR-DATUM holds the
1412 lisp symbolic place in question rather than the place's value. This
1415 334: "COMPUTE-SLOTS used to add slots to classes"
1416 (reported by Bruno Haible sbcl-devel 2004-06-01)
1417 a. Adding a local slot does not work:
1418 (use-package "SB-PCL")
1420 (defmethod compute-slots ((class (eql (find-class 'b))))
1421 (append (call-next-method)
1422 (list (make-instance 'standard-effective-slot-definition
1424 :allocation :instance))))
1425 (defclass a () ((x :allocation :class)))
1426 ;; A should now have a shared slot, X, and a local slot, Y.
1427 (mapcar #'slot-definition-location (class-slots (find-class 'b)))
1429 There is no applicable method for the generic function
1430 #<STANDARD-GENERIC-FUNCTION CLASS-SLOTS (3)>
1431 when called with arguments
1434 b. Adding a class slot does not work:
1435 (use-package "SB-PCL")
1437 (defmethod compute-slots ((class (eql (find-class 'b))))
1438 (append (call-next-method)
1439 (list (make-instance 'standard-effective-slot-definition
1441 :allocation :class))))
1442 (defclass a () ((x :allocation :class)))
1443 ;; A should now have two shared slots, X and Y.
1444 (mapcar #'slot-definition-location (class-slots (find-class 'b)))
1446 There is no applicable method for the generic function
1447 #<STANDARD-GENERIC-FUNCTION SB-PCL::CLASS-SLOT-CELLS (1)>
1448 when called with arguments
1451 336: "slot-definitions must retain the generic functions of accessors"
1452 reported by Tony Martinez:
1453 (defclass foo () ((bar :reader foo-bar)))
1454 (defun foo-bar (x) x)
1455 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1457 Note: just punting the accessor removal if the fdefinition
1458 is not a generic function is not enough:
1460 (defclass foo () ((bar :reader foo-bar)))
1461 (defvar *reader* #'foo-bar)
1462 (defun foo-bar (x) x)
1463 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1464 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1465 ; the method must be removed
1466 ; by the class redefinition
1468 Fixing this should also fix a subset of #328 -- update the
1469 description with a new test-case then.