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 a slew of floating-point-related errors reported by Peter Van Eynde
162 c: Many expressions generate floating infinity on x86/Linux:
167 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
168 on x86/Linux generates the infinities instead. That might or
169 might not be conforming behavior, but it's also inconsistent,
170 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
171 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
172 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
174 d: (in section12.erg) various forms a la
175 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
176 don't give the right behavior.
179 The debugger LIST-LOCATIONS command doesn't work properly.
180 (How should it work properly?)
183 Compiling and loading
184 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
186 then requesting a BACKTRACE at the debugger prompt gives no information
187 about where in the user program the problem occurred.
190 Using the pretty-printer from the command prompt gives funny
191 results, apparently because the pretty-printer doesn't know
192 about user's command input, including the user's carriage return
193 that the user, and therefore the pretty-printer thinks that
194 the new output block should start indented 2 or more characters
195 rightward of the correct location.
198 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
199 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
200 crashes SBCL. In general tracing anything which is used in the
201 implementation of TRACE is likely to have the same problem.
204 ANSI says in one place that type declarations can be abbreviated even
205 when the type name is not a symbol, e.g.
206 (DECLAIM ((VECTOR T) *FOOVECTOR*))
207 SBCL doesn't support this. But ANSI says in another place that this
208 isn't allowed. So it's not clear this is a bug after all. (See the
209 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
213 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
214 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
215 an easily guessable temporary filename in a way which might open
216 applications using LOAD-FOREIGN to hijacking by malicious users
217 on the same machine. Incantations for doing this safely are
218 floating around the net in various "how to write secure programs
219 despite Unix" documents, and it would be good to (1) fix this in
220 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
221 files and make it share the same new safe logic.
223 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
224 make the temporary filename less easily guessable)
227 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
228 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
229 the mean of the distribution can be systematically O(0.1%) wrong.
230 Just increasing R-I-E-B is probably not a good solution, since
231 it would decrease efficiency more than is probably necessary. Perhaps
232 using some sort of accept/reject method would be better.
235 Internally the compiler sometimes evaluates
236 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
237 (I stumbled across this when I added an
238 (assert (not (eq type1 *wild-type*)))
239 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
240 in a type context should probably be translated to T, and so it's
241 probably wrong to ask whether it's equal to the T type and then (using
242 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
243 (I haven't tried to investigate this bug enough to guess whether
244 there might be any user-level symptoms.)
246 In fact, the type system is likely to depend on this inequality not
247 holding... * is not equivalent to T in many cases, such as
248 (VECTOR *) /= (VECTOR T).
251 The facility for dumping a running Lisp image to disk gets confused
252 when run without the PURIFY option, and creates an unnecessarily large
253 core file (apparently representing memory usage up to the previous
254 high-water mark). Moreover, when the file is loaded, it confuses the
255 GC, so that thereafter memory usage can never be reduced below that
258 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
259 The interaction between gencgc and the variables used by
260 save-lisp-and-die is still nonoptimal, though, so no respite from
264 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
265 CL), out-of-line structure slot setters are horribly inefficient
266 whenever the type of the slot is declared, because out-of-line
267 structure slot setters are implemented as closures to save space,
268 so the compiler doesn't compile the type test into code, but
269 instead just saves the type in a lexical closure and interprets it
271 To exercise the problem, compile and load
272 (cl:in-package :cl-user)
274 (bar (error "missing") :type bar))
277 (loop (setf (foo-bar *foo*) x)))
279 (defvar *bar* (make-bar))
280 (defvar *foo* (make-foo :bar *bar*))
281 (defvar *setf-foo-bar* #'(setf foo-bar))
283 (loop (funcall *setf-foo-bar* x *foo*)))
284 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
285 use BACKTRACE, to see it's spending all essentially all its time
286 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
287 One possible solution would be simply to give up on
288 representing structure slot accessors as functions, and represent
289 them as macroexpansions instead. This can be inconvenient for users,
290 but it's not clear that it's worse than trying to help by expanding
291 into a horribly inefficient implementation.
292 As a workaround for the problem, #'(SETF FOO) expressions
293 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
294 (defmacro efficient-setf-function (place-function-name)
295 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
296 ;; a workaround for the problem, encouraging the
297 ;; inline expansion of the structure accessor, so
298 ;; that the compiler can optimize its type test
299 (let ((new-value (gensym "NEW-VALUE-"))
300 (structure-value (gensym "STRUCTURE-VALUE-")))
301 `(lambda (,new-value ,structure-value)
302 (setf (,place-function-name ,structure-value)
304 ;; no problem, can just use the ordinary expansion
305 `(function (setf ,place-function-name))))
308 There's apparently a bug in CEILING optimization which caused
309 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
310 applied the patches to SBCL and they didn't seem to cause problems
311 (as reported sbcl-devel 2001-05-04). However, since the patches
312 modify nontrivial code which was apparently written incorrectly
313 the first time around, until regression tests are written I'm not
314 comfortable merging the patches in the CVS version of SBCL.
317 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
318 a clean, just-started SBCL system. And it seems to be right:
319 (ROOM T) can bring a small computer to its knees for a *long*
320 time trying to GC afterwards. Surely there's some more economical
321 way to implement (ROOM T).
323 Daniel Barlow doesn't know what fixed this, but observes that it
324 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
325 in a fresh SBCL causes
327 debugger invoked on a SB-INT:BUG in thread 5911:
328 failed AVER: "(SAP= CURRENT END)"
330 unless a GC has happened beforehand.
333 When the compiler inline expands functions, it may be that different
334 kinds of return values are generated from different code branches.
335 E.g. an inline expansion of POSITION generates integer results
336 from one branch, and NIL results from another. When that inline
337 expansion is used in a context where only one of those results
340 (aref *a1* (position x *a2*)))
341 and the compiler can't prove that the unacceptable branch is
342 never taken, then bogus type mismatch warnings can be generated.
343 If you need to suppress the type mismatch warnings, you can
344 suppress the inline expansion,
346 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
347 (aref *a1* (position x *a2*)))
348 or, sometimes, suppress them by declaring the result to be of an
351 (aref *a1* (the integer (position x *a2*))))
353 This is not a new compiler problem in 0.7.0, but the new compiler
354 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
355 more conspicuous. If you don't need performance from these functions,
356 and the bogus warnings are a nuisance for you, you can return to
357 your pre-0.7.0 state of grace with
358 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
363 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
364 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
365 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
366 when of course it should be NIL. (He says it only fails for X86,
367 not SPARC; dunno about Alpha.)
369 Also, "the same problem exists for LONG-FLOAT-EPSILON,
370 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
371 for the -negative- the + is replaced by a - in the test)."
373 Raymond Toy comments that this is tricky on the X86 since its FPU
374 uses 80-bit precision internally.
376 Bruno Haible comments:
377 The values are those that are expected for an IEEE double-float
378 arithmetic. The problem appears to be that the rounding is not
379 IEEE on x86 compliant: namely, values are first rounded to 64
380 bits mantissa precision, then only to 53 bits mantissa
381 precision. This gives different results than rounding to 53 bits
382 mantissa precision in a single step.
384 The quick "fix", to permanently change the FPU control word from
385 0x037f to 0x027f, will give problems with the fdlibm code that is
386 used for computing transcendental functions like sinh() etc.
387 so maybe we need to change the FPU control word to that for Lisp
388 code, and adjust it to the safe 0x037f for calls to C?
391 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
392 the entire lexical environment at the point of MACROLET available
393 in the bodies of the macroexpander functions. In particular, it
394 allows the function bodies (which run at compile time) to try to
395 access lexical variables (which are only defined at runtime).
396 It doesn't even issue a warning, which is bad.
398 The SBCL behavior arguably conforms to the ANSI spec (since the
399 spec says that the behavior is undefined, ergo anything conforms).
400 However, it would be better to issue a compile-time error.
401 Unfortunately I (WHN) don't see any simple way to detect this
402 condition in order to issue such an error, so for the meantime
403 SBCL just does this weird broken "conforming" thing.
405 The ANSI standard says, in the definition of the special operator
407 The macro-expansion functions defined by MACROLET are defined
408 in the lexical environment in which the MACROLET form appears.
409 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
410 the local macro definitions in a MACROLET, but the consequences
411 are undefined if the local macro definitions reference any
412 local variable or function bindings that are visible in that
414 Then it seems to contradict itself by giving the example
416 (macrolet ((fudge (z)
417 ;The parameters x and flag are not accessible
418 ; at this point; a reference to flag would be to
419 ; the global variable of that name.
420 ` (if flag (* ,z ,z) ,z)))
421 ;The parameters x and flag are accessible here.
425 The comment "a reference to flag would be to the global variable
426 of the same name" sounds like good behavior for the system to have.
427 but actual specification quoted above says that the actual behavior
430 (Since 0.7.8.23 macroexpanders are defined in a restricted version
431 of the lexical environment, containing no lexical variables and
432 functions, which seems to conform to ANSI and CLtL2, but signalling
433 a STYLE-WARNING for references to variables similar to locals might
437 (as reported by Gabe Garza on cmucl-help 2001-09-21)
439 (defun test-pred (x y)
443 (func (lambda () x)))
444 (print (eq func func))
445 (print (test-pred func func))
446 (delete func (list func))))
447 Now calling (TEST-CASE) gives output
450 (#<FUNCTION {500A9EF9}>)
451 Evidently Python thinks of the lambda as a code transformation so
452 much that it forgets that it's also an object.
455 Ideally, uninterning a symbol would allow it, and its associated
456 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
457 at least as of sbcl-0.7.0, this isn't the case. Information about
458 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
459 essentially in ordinary (non-weak) hash tables keyed by symbols.
460 Thus, once a system has an entry in this system, it tends to live
461 forever, even when it is uninterned and all other references to it
464 141: "pretty printing and backquote"
467 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
469 c. (reported by Paul F. Dietz)
471 `(LAMBDA (SB-IMPL::BACKQ-COMMA X))
474 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
475 prominent SourceForge web/db bug tracking system, which is
476 unfortunately not a reliable way to get a timely response from
477 the SBCL maintainers)
478 In the course of trying to build a test case for an
479 application error, I encountered this behavior:
480 If you start up sbcl, and then lay on CTRL-C for a
481 minute or two, the lisp process will eventually say:
482 %PRIMITIVE HALT called; the party is over.
483 and throw you into the monitor. If I start up lisp,
484 attach to the process with strace, and then do the same
485 (abusive) thing, I get instead:
486 access failure in heap page not marked as write-protected
487 and the monitor again. I don't know enough to have the
488 faintest idea of what is going on here.
489 This is with sbcl 6.12, uname -a reports:
490 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
491 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
492 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
493 it took more than two minutes (but less than five) for me.
497 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
498 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
499 COMPLEX implementation didn't deal with this, and hasn't been
500 upgraded to do so. (This doesn't seem to be a high priority
501 conformance problem, since seems hard to construct useful code
504 b. (fixed in 0.8.3.43)
507 Floating point errors are reported poorly. E.g. on x86 OpenBSD
510 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
511 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
512 No traps are enabled? How can this be?
513 It should be possible to be much more specific (overflow, division
514 by zero, etc.) and of course the "How can this be?" should be fixable.
516 See also bugs #45.c and #183
519 (reported by Robert E. Brown 2002-04-16)
520 When a function is called with too few arguments, causing the
521 debugger to be entered, the uninitialized slots in the bad call frame
522 seem to cause GCish problems, being interpreted as tagged data even
523 though they're not. In particular, executing ROOM in the
524 debugger at that point causes AVER failures:
527 * (lisp-implementation-version)
533 failed AVER: "(SAP= CURRENT END)"
534 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
535 isn't too surprising since there are many differences in stack
536 implementation and GC conservatism between the X86 and other ports.)
538 This is probably the same bug as 216
541 In sbcl-0.7.3.11, compiling the (illegal) code
542 (in-package :cl-user)
543 (defmethod prove ((uustk uustk))
546 gives the (not terribly clear) error message
548 ; (during macroexpansion of (DEFMETHOD PROVE ...))
549 ; can't get template for (FROB NIL NIL)
550 The problem seems to be that the code walker used by the DEFMETHOD
551 macro is unhappy with the illegal syntax in the method body, and
552 is giving an unclear error message.
555 The compiler sometimes tries to constant-fold expressions before
556 it checks to see whether they can be reached. This can lead to
557 bogus warnings about errors in the constant folding, e.g. in code
560 (WRITE-STRING (> X 0) "+" "0"))
561 compiled in a context where the compiler can prove that X is NIL,
562 and the compiler complains that (> X 0) causes a type error because
563 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
564 caused a full WARNING, which made the bug really annoying because then
565 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
566 code. Since then the warning has been downgraded to STYLE-WARNING,
567 so it's still a bug but at least it's a little less annoying.
569 183: "IEEE floating point issues"
570 Even where floating point handling is being dealt with relatively
571 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
572 accrued-exceptions and current-exceptions part of the fp control
573 word don't seem to bear much relation to reality. E.g. on
577 debugger invoked on condition of type DIVISION-BY-ZERO:
578 arithmetic error DIVISION-BY-ZERO signalled
579 0] (sb-vm::get-floating-point-modes)
581 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
582 :ROUNDING-MODE :NEAREST
583 :CURRENT-EXCEPTIONS NIL
584 :ACCRUED-EXCEPTIONS (:INEXACT)
587 * (sb-vm::get-floating-point-modes)
588 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
589 :ROUNDING-MODE :NEAREST
590 :CURRENT-EXCEPTIONS (:INEXACT)
591 :ACCRUED-EXCEPTIONS (:INEXACT)
594 188: "compiler performance fiasco involving type inference and UNION-TYPE"
598 (declare (optimize (safety 3)))
599 (declare (optimize (compilation-speed 2)))
600 (declare (optimize (speed 1) (debug 1) (space 1)))
602 (declare (type (integer 0) start))
603 (print (incf start 22))
604 (print (incf start 26))
605 (print (incf start 28)))
607 (declare (type (integer 0) start))
608 (print (incf start 22))
609 (print (incf start 26)))
611 (declare (type (integer 0) start))
612 (print (incf start 22))
613 (print (incf start 26))))))
615 This example could be solved with clever enough constraint
616 propagation or with SSA, but consider
621 The careful type of X is {2k} :-(. Is it really important to be
622 able to work with unions of many intervals?
624 191: "Miscellaneous PCL deficiencies"
625 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
626 a. DEFCLASS does not inform the compiler about generated
627 functions. Compiling a file with
631 (WITH-SLOTS (A-CLASS-X) A
633 results in a STYLE-WARNING:
635 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
637 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
638 Mai points out that the declamation of functions is in fact
639 incorrect in some cases (most notably for structure
640 classes). This means that at present erroneous attempts to use
641 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
642 won't get the corresponding STYLE-WARNING.
643 c. (fixed in 0.8.4.23)
645 201: "Incautious type inference from compound types"
646 a. (reported by APD sbcl-devel 2002-09-17)
648 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
650 (FORMAT NIL "~S IS ~S, Y = ~S"
657 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
661 (declare (type (array * (4 4)) x))
663 (setq x (make-array '(4 4)))
664 (adjust-array y '(3 5))
665 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
667 * (foo (make-array '(4 4) :adjustable t))
670 205: "environment issues in cross compiler"
671 (These bugs have no impact on user code, but should be fixed or
673 a. Macroexpanders introduced with MACROLET are defined in the null
675 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
676 the null lexical environment.
677 c. The cross-compiler cannot inline functions defined in a non-null
680 206: ":SB-FLUID feature broken"
681 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
682 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
685 207: "poorly distributed SXHASH results for compound data"
686 SBCL's SXHASH could probably try a little harder. ANSI: "the
687 intent is that an implementation should make a good-faith
688 effort to produce hash-codes that are well distributed
689 within the range of non-negative fixnums". But
690 (let ((hits (make-hash-table)))
693 (let* ((ij (cons i j))
694 (newlist (push ij (gethash (sxhash ij) hits))))
696 (format t "~&collision: ~S~%" newlist))))))
697 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
698 would be an obvious way of fix. Maybe it would be acceptably efficient
699 to redo MIX using a lookup into a 256-entry s-box containing
700 29-bit pseudorandom numbers?
702 211: "keywords processing"
703 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
704 number of keyword arguments.
707 (flet ((foo (&key y) (list y)))
708 (list (foo :y 1 :y 2)))
710 issues confusing message
715 ; caught STYLE-WARNING:
716 ; The variable #:G15 is defined but never used.
718 212: "Sequence functions and circular arguments"
719 COERCE, MERGE and CONCATENATE go into an infinite loop when given
720 circular arguments; it would be good for the user if they could be
721 given an error instead (ANSI 17.1.1 allows this behaviour on the part
722 of the implementation, as conforming code cannot give non-proper
723 sequences to these functions. MAP also has this problem (and
724 solution), though arguably the convenience of being able to do
725 (MAP 'LIST '+ FOO '#1=(1 . #1#))
726 might be classed as more important (though signalling an error when
727 all of the arguments are circular is probably desireable).
729 213: "Sequence functions and type checking"
730 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
731 check that it will return a sequence of the given type. Fixing
732 it along the same lines as the others (cf. work done around
733 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
734 entirely straightforward.
735 c. All of these functions will silently accept a type of the form
737 whether or not the return value is of this type. This is
738 probably permitted by ANSI (see "Exceptional Situations" under
739 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
740 know about this escape clause, so code of the form
741 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
742 can erroneously return T.
744 215: ":TEST-NOT handling by functions"
745 a. FIND and POSITION currently signal errors when given non-NIL for
746 both their :TEST and (deprecated) :TEST-NOT arguments, but by
747 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
748 means that the effect is "unpredictable but harmless". It's not
749 clear what that actually means; it may preclude conforming
750 implementations from signalling errors.
751 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
752 when conflict occurs. As a quality of implementation issue, it
753 might be preferable to treat :TEST and :TEST-NOT as being in some
754 sense the same &KEY, and effectively take the first test function in
756 c. Again, a quality of implementation issue: it would be good to issue a
757 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
758 WARNING for calls with both :TEST and :TEST-NOT; possibly this
759 latter should be WARNed about at execute-time too.
761 216: "debugger confused by frames with invalid number of arguments"
762 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
763 leaves the system confused, enough so that (QUIT) no longer works.
764 It's as though the process of working with the uninitialized slot in
765 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
766 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
768 This is probably the same bug as 162
770 217: "Bad type operations with FUNCTION types"
773 * (values-type-union (specifier-type '(function (base-char)))
774 (specifier-type '(function (integer))))
776 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
778 It causes insertion of wrong type assertions into generated
782 (let ((f (etypecase x
783 (character #'write-char)
784 (integer #'write-byte))))
787 (character (write-char x s))
788 (integer (write-byte x s)))))
790 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
792 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
793 produce invalid code, but type checking is not accurate.)
795 233: bugs in constraint propagation
797 (declaim (optimize (speed 2) (safety 3)))
799 (if (typep (prog1 x (setq x y)) 'double-float)
802 (foo 1d0 5) => segmentation violation
804 235: "type system and inline expansion"
806 (declaim (ftype (function (cons) number) acc))
807 (declaim (inline acc))
809 (the number (car c)))
812 (values (locally (declare (optimize (safety 0)))
814 (locally (declare (optimize (safety 3)))
817 (foo '(nil) '(t)) => NIL, T.
819 237: "Environment arguments to type functions"
820 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
821 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
822 argument, but they ignore it completely. This is almost
823 certainly not correct.
824 b. Also, the compiler's optimizers for TYPEP have not been informed
825 about the new argument; consequently, they will not transform
826 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
827 just as optimizeable as (TYPEP 1 'INTEGER).
829 238: "REPL compiler overenthusiasm for CLOS code"
831 * (defclass foo () ())
832 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
833 causes approximately 100 lines of code deletion notes. Some
834 discussion on this issue happened under the title 'Three "interesting"
835 bugs in PCL', resulting in a fix for this oververbosity from the
836 compiler proper; however, the problem persists in the interactor
837 because the notion of original source is not preserved: for the
838 compiler, the original source of the above expression is (DEFMETHOD
839 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
840 compiler gets its hands on the code needing compilation from the REPL,
841 it has been macroexpanded several times.
843 A symptom of the same underlying problem, reported by Tony Martinez:
845 (with-input-from-string (*query-io* " no")
847 (simple-type-error () 'error))
849 ; (SB-KERNEL:FLOAT-WAIT)
851 ; note: deleting unreachable code
852 ; compilation unit finished
855 242: "WRITE-SEQUENCE suboptimality"
856 (observed from clx performance)
857 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
858 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
859 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
860 rather than writing the sequence in one go, leading to severe
861 performance degradation.
863 243: "STYLE-WARNING overenthusiasm for unused variables"
864 (observed from clx compilation)
865 In sbcl-0.7.14, in the presence of the macros
866 (DEFMACRO FOO (X) `(BAR ,X))
867 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
868 somewhat surprising style warnings are emitted for
869 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
871 ; (LAMBDA (Y) (FOO Y))
873 ; caught STYLE-WARNING:
874 ; The variable Y is defined but never used.
876 245: bugs in disassembler
877 b. On X86 operand size prefix is not recognized.
880 (defun foo (&key (a :x))
884 does not cause a warning. (BTW: old SBCL issued a warning, but for a
885 function, which was never called!)
888 Compiler does not emit warnings for
890 a. (lambda () (svref (make-array 8 :adjustable t) 1))
893 (list (let ((y (the real x)))
894 (unless (floatp y) (error ""))
899 (declare (optimize (debug 0)))
900 (declare (type vector x))
901 (list (fill-pointer x)
905 Complex array type does not have corresponding type specifier.
907 This is a problem because the compiler emits optimization notes when
908 you use a non-simple array, and without a type specifier for hairy
909 array types, there's no good way to tell it you're doing it
910 intentionally so that it should shut up and just compile the code.
912 Another problem is confusing error message "asserted type ARRAY
913 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
914 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
916 The last problem is that when type assertions are converted to type
917 checks, types are represented with type specifiers, so we could lose
918 complex attribute. (Now this is probably not important, because
919 currently checks for complex arrays seem to be performed by
923 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
924 warning. Analogous cases with the index and length being equal and
925 greater than 0 are warned for; the problem here seems to be that the
926 type required for an array reference of this type is (INTEGER 0 (0))
927 which is canonicalized to NIL.
932 (t1 (specifier-type s)))
933 (eval `(defstruct ,s))
934 (type= t1 (specifier-type s)))
939 b. The same for CSUBTYPEP.
941 262: "yet another bug in inline expansion of local functions"
945 (declare (integer x y))
948 (declare (integer u))
949 (if (> (1+ (the unsigned-byte u)) 0)
951 (return (+ 38 (cos (/ u 78)))))))
952 (declare (inline xyz))
954 (* (funcall (eval #'xyz) x)
956 (funcall (if (> x 5) #'xyz #'identity)
961 Urgh... It's time to write IR1-copier.
964 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
965 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
966 looks reasonable, if not obviously correct; however, it caused the
967 PPC/Linux port to segfault during warm-init while loading
968 src/pcl/std-class.fasl. A workaround patch was made, but it would
969 be nice to understand why the first patch caused problems, and to
970 fix the cause if possible.
972 268: "wrong free declaration scope"
973 The following code must signal type error:
975 (locally (declare (optimize (safety 3)))
976 (flet ((foo (x &optional (y (car x)))
977 (declare (optimize (safety 0)))
979 (funcall (eval #'foo) 1)))
982 SCALE-FLOAT should accept any integer for its second argument.
985 In the following function constraint propagator optimizes nothing:
988 (declare (integer x))
989 (declare (optimize speed))
997 Compilation of the following two forms causes "X is unbound" error:
999 (symbol-macrolet ((x pi))
1000 (macrolet ((foo (y) (+ x y)))
1001 (declaim (inline bar))
1007 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
1010 CLHS says that type declaration of a symbol macro should not affect
1011 its expansion, but in SBCL it does. (If you like magic and want to
1012 fix it, don't forget to change all uses of MACROEXPAND to
1016 The following code (taken from CLOCC) takes a lot of time to compile:
1019 (declare (type (integer 0 #.large-constant) n))
1022 (fixed in 0.8.2.51, but a test case would be good)
1025 (defmethod fee ((x fixnum))
1028 (fee 1) => type error
1035 (declare (optimize speed))
1036 (loop for i of-type (integer 0) from 0 by 2 below 10
1039 uses generic arithmetic.
1041 b. (fixed in 0.8.3.6)
1043 279: type propagation error -- correctly inferred type goes astray?
1044 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
1045 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
1046 &OPTIONAL), not a (INTEGER 1 536870911)
1047 is emitted when compiling this file:
1048 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
1049 (integer #.most-negative-fixnum 0))
1054 (let* (;; Uncomment this for a type mismatch warning indicating
1055 ;; that the type of (FOO X) is correctly understood.
1056 #+nil (fs-foo (float-sign (foo x)))
1057 ;; Uncomment this for a type mismatch warning
1058 ;; indicating that the type of (ABS (FOO X)) is
1059 ;; correctly understood.
1060 #+nil (fs-abs-foo (float-sign (abs (foo x))))
1061 ;; something wrong with this one though
1062 (abs-foo (abs (foo x))))
1063 (declare (type (integer 1 100) abs-foo))
1068 281: COMPUTE-EFFECTIVE-METHOD error signalling.
1069 (slightly obscured by a non-0 default value for
1070 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
1071 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
1072 when it finds a method with invalid qualifiers. However, it
1073 shouldn't signal errors when any such methods are not applicable to
1074 the particular call being evaluated, and certainly it shouldn't when
1075 simply precomputing effective methods that may never be called.
1076 (setf sb-pcl::*max-emf-precompute-methods* 0)
1078 (:method-combination +)
1079 (:method ((x symbol)) 1)
1080 (:method + ((x number)) x))
1081 (foo 1) -> ERROR, but should simply return 1
1083 The issue seems to be that construction of a discriminating function
1084 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
1086 283: Thread safety: libc functions
1087 There are places that we call unsafe-for-threading libc functions
1088 that we should find alternatives for, or put locks around. Known or
1089 strongly suspected problems, as of 0.8.3.10: please update this
1090 bug instead of creating new ones
1092 localtime() - called for timezone calculations in code/time.lisp
1094 284: Thread safety: special variables
1095 There are lots of special variables in SBCL, and I feel sure that at
1096 least some of them are indicative of potentially thread-unsafe
1097 parts of the system. See doc/internals/notes/threading-specials
1099 286: "recursive known functions"
1100 Self-call recognition conflicts with known function
1101 recognition. Currently cross compiler and target COMPILE do not
1102 recognize recursion, and in target compiler it can be disabled. We
1103 can always disable it for known functions with RECURSIVE attribute,
1104 but there remains a possibility of a function with a
1105 (tail)-recursive simplification pass and transforms/VOPs for base
1108 287: PPC/Linux miscompilation or corruption in first GC
1109 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
1110 segmentation fault is reported at the point of first triggered GC,
1111 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
1112 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
1113 that this merely obscures, not solves, the underlying problem; as and when
1114 underlying problems are fixed, it would be worth trying again to provoke
1117 288: fundamental cross-compilation issues (from old UGLINESS file)
1118 Using host floating point numbers to represent target floating point
1119 numbers, or host characters to represent target characters, is
1120 theoretically shaky. (The characters are OK as long as the characters
1121 are in the ANSI-guaranteed character set, though, so they aren't a
1122 real problem as long as the sources don't need anything but that;
1123 the floats are a real problem.)
1125 289: "type checking and source-transforms"
1127 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1130 Our policy is to check argument types at the moment of a call. It
1131 disagrees with ANSI, which says that type assertions are put
1132 immediately onto argument expressions, but is easier to implement in
1133 IR1 and is more compatible to type inference, inline expansion,
1134 etc. IR1-transforms automatically keep this policy, but source
1135 transforms for associative functions (such as +), being applied
1136 during IR1-convertion, do not. It may be tolerable for direct calls
1137 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1139 b. Another aspect of this problem is efficiency. [x y + z +]
1140 requires less registers than [x y z + +]. This transformation is
1141 currently performed with source transforms, but it would be good to
1142 also perform it in IR1 optimization phase.
1144 290: Alpha floating point and denormalized traps
1145 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1146 hardware or kernel deficiency: the status of the enable/disable
1147 denormalized-float traps bit seems to be ambiguous; by the time we
1148 get to os_restore_fp_control after a trap, denormalized traps seem
1149 to be enabled. Since we don't want a trap every time someone uses a
1150 denormalized float, in general, we mask out that bit when we restore
1151 the control word; however, this clobbers any change the user might
1155 (reported by Adam Warner, sbcl-devel 2003-09-23)
1157 The --load toplevel argument does not perform any sanitization of its
1158 argument. As a result, files with Lisp pathname pattern characters
1159 (#\* or #\?, for instance) or quotation marks can cause the system
1160 to perform arbitrary behaviour.
1163 LOOP with non-constant arithmetic step clauses suffers from overzealous
1164 type constraint: code of the form
1165 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1166 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1167 (0))). However, an integral value of X should be legal, because
1168 successive adds of integers to double-floats produces double-floats,
1169 so none of the type restrictions in the code is violated.
1171 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1172 argument type only after having read a character. This is caused
1173 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1174 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1175 less error prone to have EXPLICIT-CHECK be a local declaration,
1176 being put into the definition, instead of an attribute being kept in
1177 a separate file; maybe also put it into SB-EXT?
1179 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1180 in NOTE-ASSUMED-TYPES
1181 In sbcl-0.8.7.32, compiling the file
1183 (declare (type integer x))
1184 (declare (type (vector (or hash-table bit)) y))
1187 (declare (type integer x))
1188 (declare (type (simple-array base (2)) y))
1191 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1193 302: Undefined type messes up DATA-VECTOR-REF expansion.
1195 (defun dis (s ei x y)
1196 (declare (type (simple-array function (2)) s) (type ei ei))
1197 (funcall (aref s ei) x y))
1198 on sbcl-0.8.7.36/X86/Linux causes a BUG to be signalled:
1199 full call to SB-KERNEL:DATA-VECTOR-REF
1201 303: "nonlinear LVARs" (aka MISC.293)
1203 (multiple-value-call #'list
1205 (multiple-value-prog1
1206 (eval '(values :a :b :c))
1212 (throw 'bar (values 3 4)))))))))))
1214 (BUU 1) returns garbage.
1216 The problem is that both EVALs sequentially write to the same LVAR.
1219 (Reported by Dave Roberts.)
1220 Local INLINE/NOTINLINE declaration removes local FTYPE declaration:
1223 (declare (ftype (function () (integer 0 10)) fee)
1227 uses generic arithmetic with INLINE and fixnum without.
1229 306: "Imprecise unions of array types"
1231 (declare (optimize speed)
1232 (type (or (array cons) (array vector)) x))
1234 (foo #((0))) => TYPE-ERROR
1241 ,@(loop for x across sb-vm:*specialized-array-element-type-properties*
1242 collect `(array ,(sb-vm:saetp-specifier x)))))
1243 => NIL, T (when it should be T, T)
1245 308: "Characters without names"
1246 (reported by Bruno Haible sbcl-devel "character names are missing"
1248 (graphic-char-p (code-char 255))
1250 (char-name (code-char 255))
1253 SBCL is unsure of what to do about characters with codes in the
1254 range 128-255. Currently they are treated as non-graphic, but don't
1255 have names, which is not compliant with the standard. Various fixes
1256 are possible, such as
1257 * giving them names such as NON-ASCII-128;
1258 * reducing CHAR-CODE-LIMIT to 127 (almost certainly unpopular);
1259 * making the characters graphic (makes a certain amount of sense);
1260 * biting the bullet and implementing Unicode (probably quite hard).
1262 309: "Dubious values for implementation limits"
1263 (reported by Bruno Haible sbcl-devel "Incorrect value of
1264 multiple-values-limit" 2004-04-19)
1265 (values-list (make-list 1000000)), on x86/linux, signals a stack
1266 exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
1267 significantly larger than 1000000. There are probably similar
1268 dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
1269 around the same time regarding a call to LIST on sparc with 1000
1270 arguments) and other implementation limit constants.
1272 311: "Tokeniser not thread-safe"
1273 (see also Robert Marlow sbcl-help "Multi threaded read chucking a
1275 The tokenizer's use of *read-buffer* and *read-buffer-length* causes
1276 spurious errors should two threads attempt to tokenise at the same
1279 314: "LOOP :INITIALLY clauses and scope of initializers"
1280 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1281 test suite, originally by Thomas F. Burdick.
1282 ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
1283 ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
1284 ;; initialized to the result of evaluating form1. 6.1.7.2 says that
1285 ;; initially clauses are evaluated in the loop prologue, which precedes all
1286 ;; loop code except for the initial settings provided by with, for, or as.
1287 (loop :for x = 0 :then (1+ x)
1288 :for y = (1+ x) :then (ash y 1)
1289 :for z :across #(1 3 9 27 81 243)
1291 :initially (assert (zerop x)) :initially (assert (= 2 w))
1292 :until (>= w 100) :collect w)
1293 Expected: (2 6 15 38)
1296 317: "FORMAT of floating point numbers"
1297 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1299 (format nil "~1F" 10) => "0." ; "10." expected
1300 (format nil "~0F" 10) => "0." ; "10." expected
1301 (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected
1302 it would be nice if whatever fixed this also untangled the two
1303 competing implementations of floating point printing (Steele and
1304 White, and Burger and Dybvig) present in src/code/print.lisp
1306 318: "stack overflow in compiler warning with redefined class"
1307 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1309 (setq *print-pretty* nil)
1311 (setf (find-class 'foo) nil)
1312 (defstruct foo slot-1)
1314 ...#<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTUREControl stack guard page temporarily disabled: proceed with caution
1315 (it's not really clear what it should give: is (SETF FIND-CLASS)
1316 meant to be enough to delete structure classes from the system?
1317 Giving a stack overflow is definitely suboptimal, though.)
1319 319: "backquote with comma inside array"
1320 reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
1322 (read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
1324 #(1 2 ((SB-IMPL::|,|) + 2 2) 4)
1325 which probably isn't intentional.
1327 323: "REPLACE, BIT-BASH and large strings"
1328 The transform for REPLACE on simple-base-strings uses BIT-BASH, which
1329 at present has an upper limit in size. Consequently, in sbcl-0.8.10
1331 (declare (optimize speed (safety 1)))
1332 (let ((x (make-string 140000000))
1333 (y (make-string 140000000)))
1334 (length (replace x y))))
1337 debugger invoked on a TYPE-ERROR in thread 2412:
1338 The value 1120000000 is not of type (MOD 536870911).
1339 (see also "more and better sequence transforms" sbcl-devel 2004-05-10)
1341 324: "STREAMs and :ELEMENT-TYPE with large bytesize"
1342 In theory, (open foo :element-type '(unsigned-byte <x>)) should work
1343 for all positive integral <x>. At present, it only works for <x> up
1344 to about 1024 (and similarly for signed-byte), so
1345 (open "/dev/zero" :element-type '(unsigned-byte 1025))
1346 gives an error in sbcl-0.8.10.
1348 325: "CLOSE :ABORT T on supeseding streams"
1349 Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
1350 file on disk, even if one existed before opening.
1352 The illegality of this is not crystal clear, as the ANSI dictionary
1353 entry for CLOSE says that when :ABORT is T superseded files are not
1354 superseded (ie. the original should be restored), whereas the OPEN
1355 entry says about :IF-EXISTS :SUPERSEDE "If possible, the
1356 implementation should not destroy the old file until the new stream
1357 is closed." -- implying that even though undesirable, early deletion
1358 is legal. Restoring the original would none the less be the polite
1361 326: "*PRINT-CIRCLE* crosstalk between streams"
1362 In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
1363 mixed between streams when output operations are intermingled closely
1364 enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
1365 test case below), so that e.g. the references #2# appears on a stream
1366 with no preceding #2= on that stream to define it (because the #2= was
1367 sent to another stream).
1368 (cl:in-package :cl-user)
1369 (defstruct foo index)
1370 (defparameter *foo* (make-foo :index 4))
1372 (defparameter *bar* (make-bar))
1373 (defparameter *tangle* (list *foo* *bar* *foo*))
1374 (defmethod print-object ((foo foo) stream)
1375 (let ((index (foo-index foo)))
1376 (format *trace-output*
1377 "~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
1379 (format stream "[FOO ~D]" index))
1381 (let ((tsos (make-string-output-stream))
1382 (ssos (make-string-output-stream)))
1383 (let ((*print-circle* t)
1384 (*trace-output* tsos)
1385 (*standard-output* ssos))
1386 (prin1 *tangle* *standard-output*))
1387 (let ((string (get-output-stream-string ssos)))
1388 (unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
1389 ;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
1390 (error "oops: ~S" string))))
1391 It might be straightforward to fix this by turning the
1392 *CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
1393 per-stream slots, but (1) it would probably be sort of messy faking
1394 up the special variable binding semantics using UNWIND-PROTECT and
1395 (2) it might be sort of a pain to test that no other bugs had been
1398 328: "Profiling generic functions", transplanted from #241
1399 (from tonyms on #lisp IRC 2003-02-25)
1400 In sbcl-0.7.12.55, typing
1401 (defclass foo () ((bar :accessor foo-bar)))
1404 (defclass foo () ((bar :accessor foo-bar)))
1405 gives the error message
1406 "#:FOO-BAR already names an ordinary function or a macro."
1408 Problem: when a generic function is profiled, it appears as an ordinary
1409 function to PCL. (Remembering the uninterned accessor is OK, as the
1410 redefinition must be able to remove old accessors from their generic
1413 329: "Sequential class redefinition"
1414 reported by Bruno Haible:
1415 (defclass reactor () ((max-temp :initform 10000000)))
1416 (defvar *r1* (make-instance 'reactor))
1417 (defvar *r2* (make-instance 'reactor))
1418 (slot-value *r1* 'max-temp)
1419 (slot-value *r2* 'max-temp)
1420 (defclass reactor () ((uptime :initform 0)))
1421 (slot-value *r1* 'uptime)
1422 (defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
1423 (slot-value *r1* 'max-temp) ; => 10000
1424 (slot-value *r2* 'max-temp) ; => 10000000 oops...
1427 The method effective when the wrapper is obsoleted can be saved
1428 in the wrapper, and then to update the instance just run through
1429 all the old wrappers in order from oldest to newest.
1431 331: "lazy creation of CLOS classes for user-defined conditions"
1433 (defstruct (bar (:include foo)))
1434 (sb-mop:class-direct-subclasses (find-class 'foo))
1435 returns NIL, rather than a singleton list containing the BAR class.
1437 332: "fasl stack inconsistency in structure redefinition"
1438 (reported by Tim Daly Jr sbcl-devel 2004-05-06)
1439 Even though structure redefinition is undefined by the standard, the
1440 following behaviour is suboptimal: running
1441 (defun stimulate-sbcl ()
1442 (let ((filename (format nil "/tmp/~A.lisp" (gensym))))
1443 ;;create a file which redefines a structure incompatibly
1444 (with-open-file (f filename :direction :output :if-exists :supersede)
1445 (print '(defstruct astruct foo) f)
1446 (print '(defstruct astruct foo bar) f))
1447 ;;compile and load the file, then invoke the continue restart on
1448 ;;the structure redefinition error
1449 (handler-bind ((error (lambda (c) (continue c))))
1450 (load (compile-file filename)))))
1452 and choosing the CONTINUE restart yields the message
1453 debugger invoked on a SB-INT:BUG in thread 27726:
1454 fasl stack not empty when it should be
1456 333: "CHECK-TYPE TYPE-ERROR-DATUM place"
1457 (reported by Tony Martinez sbcl-devel 2004-05-23)
1458 When CHECK-TYPE signals a TYPE-ERROR, the TYPE-ERROR-DATUM holds the
1459 lisp symbolic place in question rather than the place's value. This
1462 334: "COMPUTE-SLOTS used to add slots to classes"
1463 (reported by Bruno Haible sbcl-devel 2004-06-01)
1464 a. Adding a local slot does not work:
1465 (use-package "SB-PCL")
1467 (defmethod compute-slots ((class (eql (find-class 'b))))
1468 (append (call-next-method)
1469 (list (make-instance 'standard-effective-slot-definition
1471 :allocation :instance))))
1472 (defclass a () ((x :allocation :class)))
1473 ;; A should now have a shared slot, X, and a local slot, Y.
1474 (mapcar #'slot-definition-location (class-slots (find-class 'b)))
1476 There is no applicable method for the generic function
1477 #<STANDARD-GENERIC-FUNCTION CLASS-SLOTS (3)>
1478 when called with arguments
1481 b. Adding a class slot does not work:
1482 (use-package "SB-PCL")
1484 (defmethod compute-slots ((class (eql (find-class 'b))))
1485 (append (call-next-method)
1486 (list (make-instance 'standard-effective-slot-definition
1488 :allocation :class))))
1489 (defclass a () ((x :allocation :class)))
1490 ;; A should now have two shared slots, X and Y.
1491 (mapcar #'slot-definition-location (class-slots (find-class 'b)))
1493 There is no applicable method for the generic function
1494 #<STANDARD-GENERIC-FUNCTION SB-PCL::CLASS-SLOT-CELLS (1)>
1495 when called with arguments
1498 335: "ATANH completely broken"
1499 a. (reported by Peter Graves sbcl-devel 2004-06-01)
1500 (atanh #c(1 2)), and more generally atanh of any complex with real
1501 part 1, computes entirely the wrong answer.
1502 b. (discovered by CSR when investigating a.)
1503 (atanh most-positive-double-float), and more generally atanh of any
1504 number with magnitude larger than
1505 sqrt(most-positive-double-float), computes a number whose real
1506 part is the imaginary part of the correct answer, and whose
1507 imaginary part is the real part of the correct answer.
1508 (fixes for both of these were sent CSR sbcl-devel 2004-06-02, to be merged
1511 336: "slot-definitions must retain the generic functions of accessors"
1512 reported by Tony Martinez:
1513 (defclass foo () ((bar :reader foo-bar)))
1514 (defun foo-bar (x) x)
1515 (defclass foo () ((bar :reader get-bar))) ; => error, should work
1517 Note: just punting the accessor removal if the fdefinition
1518 is not a generic function is not enough:
1520 (defclass foo () ((bar :reader foo-bar)))
1521 (defvar *reader* #'foo-bar)
1522 (defun foo-bar (x) x)
1523 (defclass foo () ((bar :initform 'ok :reader get-bar)))
1524 (funcall *reader* (make-instance 'foo)) ; should be an error, since
1525 ; the method must be removed
1526 ; by the class redefinition
1528 Fixing this should also fix a subset of #328 -- update the
1529 description with a new test-case then.