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).
35 KNOWN BUGS OF NO SPECIAL CLASS:
38 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
39 instead of just punting when a contradictory structure definition
40 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
41 changes their layout, you probably have to rebuild your entire
42 program, even if you know or guess enough about the internals of
43 SBCL to wager that this (undefined in ANSI) operation would be safe.
45 3: "type checking of structure slots"
47 ANSI specifies that a type mismatch in a structure slot
48 initialization value should not cause a warning.
50 This one might not be fixed for a while because while we're big
51 believers in ANSI compatibility and all, (1) there's no obvious
52 simple way to do it (short of disabling all warnings for type
53 mismatches everywhere), and (2) there's a good portable
54 workaround, and (3) by their own reasoning, it looks as though
55 ANSI may have gotten it wrong. ANSI justifies this specification
57 The restriction against issuing a warning for type mismatches
58 between a slot-initform and the corresponding slot's :TYPE
59 option is necessary because a slot-initform must be specified
60 in order to specify slot options; in some cases, no suitable
62 However, in SBCL (as in CMU CL or, for that matter, any compiler
63 which really understands Common Lisp types) a suitable default
64 does exist, in all cases, because the compiler understands the
65 concept of functions which never return (i.e. has return type NIL).
66 Thus, as a portable workaround, you can use a call to some
67 known-never-to-return function as the default. E.g.
69 (BAR (ERROR "missing :BAR argument")
70 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
72 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
73 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
74 (ERROR "missing required argument"))
76 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
77 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
78 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
79 Such code should compile without complaint and work correctly either
80 on SBCL or on any other completely compliant Common Lisp system.
82 b: &AUX argument in a boa-constructor without a default value means
83 "do not initilize this slot" and does not cause type error. But
84 an error may be signalled at read time and it would be good if
87 c: Reading of not initialized slot sometimes causes SEGV (for inline
88 accessors it is fixed, but out-of-line still do not perform type
92 (declaim (optimize (safety 3) (speed 1) (space 1)))
95 (defstruct (stringwise-foo (:include foo
96 (x "x" :type simple-string)
97 (y "y" :type simple-string))))
98 (defparameter *stringwise-foo*
99 (make-stringwise-foo))
100 (setf (foo-x *stringwise-foo*) 0)
101 (defun frob-stringwise-foo (sf)
102 (aref (stringwise-foo-x sf) 0))
103 (frob-stringwise-foo *stringwise-foo*)
107 The "compiling top-level form:" output ought to be condensed.
108 Perhaps any number of such consecutive lines ought to turn into a
109 single "compiling top-level forms:" line.
112 The way that the compiler munges types with arguments together
113 with types with no arguments (in e.g. TYPE-EXPAND) leads to
114 weirdness visible to the user:
115 (DEFTYPE FOO () 'FIXNUM)
117 (TYPEP 11 '(FOO)) => T, which seems weird
118 (TYPEP 11 'FIXNUM) => T
119 (TYPEP 11 '(FIXNUM)) signals an error, as it should
120 The situation is complicated by the presence of Common Lisp types
121 like UNSIGNED-BYTE (which can either be used in list form or alone)
122 so I'm not 100% sure that the behavior above is actually illegal.
123 But I'm 90+% sure, and the following related behavior,
125 treating the bare symbol AND as equivalent to '(AND), is specifically
126 forbidden (by the ANSI specification of the AND type).
129 It would be nice if the
131 (during macroexpansion)
132 said what macroexpansion was at fault, e.g.
134 (during macroexpansion of IN-PACKAGE,
135 during macroexpansion of DEFFOO)
138 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
139 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
140 (Also, when this is fixed, we can enable the code in PROCLAIM which
141 checks for incompatible FTYPE redeclarations.)
144 (I *think* this is a bug. It certainly seems like strange behavior. But
145 the ANSI spec is scary, dark, and deep.. -- WHN)
146 (FORMAT NIL "~,1G" 1.4) => "1. "
147 (FORMAT NIL "~3,1G" 1.4) => "1. "
150 Sometimes (SB-EXT:QUIT) fails with
151 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
152 Process inferior-lisp exited abnormally with code 1
153 I haven't noticed a repeatable case of this yet.
156 The printer doesn't report closures very well. This is true in
160 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
161 It would be nice to make closures have a settable name slot,
162 and make things like DEFSTRUCT and FLET, which create closures,
163 set helpful values into this slot.
166 And as long as we're wishing, it would be awfully nice if INSPECT could
167 also report on closures, telling about the values of the bound variables.
170 The compiler assumes that any time a function of declared FTYPE
171 doesn't signal an error, its arguments were of the declared type.
172 E.g. compiling and loading
173 (DECLAIM (OPTIMIZE (SAFETY 3)))
174 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
176 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
178 (COND ((> (FACTORIAL X) 1.0E6)
179 (FORMAT T "too big~%"))
181 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
183 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
186 will cause the INTEGERP case to be selected, giving bogus output a la
188 This violates the "declarations are assertions" principle.
189 According to the ANSI spec, in the section "System Class FUNCTION",
190 this is a case of "lying to the compiler", but the lying is done
191 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
192 not by the unexpectedly general definition of FACTORIAL. In any case,
193 "declarations are assertions" means that lying to the compiler should
194 cause an error to be signalled, and should not cause a bogus
195 result to be returned. Thus, the compiler should not assume
196 that arbitrary functions check their argument types. (It might
197 make sense to add another flag (CHECKED?) to DEFKNOWN to
198 identify functions which *do* check their argument types.)
199 (Also, verify that the compiler handles declared function
200 return types as assertions.)
203 The definitions of SIGCONTEXT-FLOAT-REGISTER and
204 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
205 supported on FreeBSD because the floating point state is not saved,
206 but at least as of FreeBSD 4.0, the floating point state *is* saved,
207 so they could be supported after all. Very likely
208 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
211 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
212 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
213 issue with 'union, 'and, 'or etc. So even though according to the
214 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
215 (and now SBCL) interpret them as legal types.
218 a slew of floating-point-related errors reported by Peter Van Eynde
220 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT on the x86 is
221 bogus, and should probably be 1.4012985e-45. In SBCL,
222 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
223 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
224 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
225 and LEAST-NEGATIVE-LONG-FLOAT.
226 c: Many expressions generate floating infinity on x86/Linux:
231 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
232 on x86/Linux generates the infinities instead. That might or
233 might not be conforming behavior, but it's also inconsistent,
234 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
235 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
236 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
238 d: (in section12.erg) various forms a la
239 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
240 don't give the right behavior.
243 type safety errors reported by Peter Van Eynde July 25, 2000:
244 c: (COERCE 'AND 'FUNCTION) returns something related to
245 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
246 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
247 not a binary input stream, but instead cheerfully reads from
248 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
251 The debugger LIST-LOCATIONS command doesn't work properly.
254 Compiling and loading
255 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
257 then requesting a BACKTRACE at the debugger prompt gives no information
258 about where in the user program the problem occurred.
261 Using the pretty-printer from the command prompt gives funny
262 results, apparently because the pretty-printer doesn't know
263 about user's command input, including the user's carriage return
264 that the user, and therefore the pretty-printer thinks that
265 the new output block should start indented 2 or more characters
266 rightward of the correct location.
269 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
270 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
271 crashes SBCL. In general tracing anything which is used in the
272 implementation of TRACE is likely to have the same problem.
275 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
276 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
277 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
278 WITH-OUTPUT-TO-STRING.
281 ANSI says in one place that type declarations can be abbreviated even
282 when the type name is not a symbol, e.g.
283 (DECLAIM ((VECTOR T) *FOOVECTOR*))
284 SBCL doesn't support this. But ANSI says in another place that this
285 isn't allowed. So it's not clear this is a bug after all. (See the
286 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
290 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
291 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
292 an easily guessable temporary filename in a way which might open
293 applications using LOAD-FOREIGN to hijacking by malicious users
294 on the same machine. Incantations for doing this safely are
295 floating around the net in various "how to write secure programs
296 despite Unix" documents, and it would be good to (1) fix this in
297 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
298 files and make it share the same new safe logic.
300 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
301 make the temporary filename less easily guessable)
304 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
305 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
306 the mean of the distribution can be systematically O(0.1%) wrong.
307 Just increasing R-I-E-B is probably not a good solution, since
308 it would decrease efficiency more than is probably necessary. Perhaps
309 using some sort of accept/reject method would be better.
312 Internally the compiler sometimes evaluates
313 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
314 (I stumbled across this when I added an
315 (assert (not (eq type1 *wild-type*)))
316 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
317 in a type context should probably be translated to T, and so it's
318 probably wrong to ask whether it's equal to the T type and then (using
319 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
320 (I haven't tried to investigate this bug enough to guess whether
321 there might be any user-level symptoms.)
323 In fact, the type system is likely to depend on this inequality not
324 holding... * is not equivalent to T in many cases, such as
325 (VECTOR *) /= (VECTOR T).
328 Inconsistencies between derived and declared VALUES return types for
329 DEFUN aren't checked very well. E.g. the logic which successfully
330 catches problems like
331 (declaim (ftype (function (fixnum) float) foo))
333 (declare (type integer x))
334 (values x)) ; wrong return type, detected, gives warning, good!
336 (declaim (ftype (function (t) (values t t)) bar))
338 (values x)) ; wrong number of return values, no warning, bad!
339 The cause of this is seems to be that (1) the internal function
340 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
341 arguments symmetrically, and (2) when the type checking code was
342 written back when when SBCL's code was still CMU CL, the intent
344 (declaim (ftype (function (t) t) bar))
346 (values x x)) ; wrong number of return values; should give warning?
347 not be warned for, because a two-valued return value is considered
348 to be compatible with callers who expects a single value to be
349 returned. That intent is probably not appropriate for modern ANSI
350 Common Lisp, but fixing this might be complicated because of other
351 divergences between auld-style and new-style handling of
352 multiple-VALUES types. (Some issues related to this were discussed
353 on cmucl-imp at some length sometime in 2000.)
356 The facility for dumping a running Lisp image to disk gets confused
357 when run without the PURIFY option, and creates an unnecessarily large
358 core file (apparently representing memory usage up to the previous
359 high-water mark). Moreover, when the file is loaded, it confuses the
360 GC, so that thereafter memory usage can never be reduced below that
364 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
365 CL), out-of-line structure slot setters are horribly inefficient
366 whenever the type of the slot is declared, because out-of-line
367 structure slot setters are implemented as closures to save space,
368 so the compiler doesn't compile the type test into code, but
369 instead just saves the type in a lexical closure and interprets it
371 A proper solution involves deciding whether it's really worth
372 saving space by implementing structure slot accessors as closures.
373 (If it's not worth it, the problem vanishes automatically. If it
374 is worth it, there are hacks we could use to force type tests to
375 be compiled anyway, and even shared. E.g. we could implement
376 an EQUAL hash table mapping from types to compiled type tests,
377 and save the appropriate compiled type test as part of each lexical
378 closure; or we could make the lexical closures be placeholders
379 which overwrite their old definition as a lexical closure with
380 a new compiled definition the first time that they're called.)
381 As a workaround for the problem, #'(SETF FOO) expressions can
382 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
383 (defmacro efficient-setf-function (place-function-name)
384 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
385 ;; a workaround for the problem, encouraging the
386 ;; inline expansion of the structure accessor, so
387 ;; that the compiler can optimize its type test
388 (let ((new-value (gensym "NEW-VALUE-"))
389 (structure-value (gensym "STRUCTURE-VALUE-")))
390 `(lambda (,new-value ,structure-value)
391 (setf (,place-function-name ,structure-value)
393 ;; no problem, can just use the ordinary expansion
394 `(function (setf ,place-function-name))))
397 There's apparently a bug in CEILING optimization which caused
398 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
399 applied the patches to SBCL and they didn't seem to cause problems
400 (as reported sbcl-devel 2001-05-04). However, since the patches
401 modify nontrivial code which was apparently written incorrectly
402 the first time around, until regression tests are written I'm not
403 comfortable merging the patches in the CVS version of SBCL.
406 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
407 a clean, just-started SBCL system. And it seems to be right:
408 (ROOM T) can bring a small computer to its knees for a *long*
409 time trying to GC afterwards. Surely there's some more economical
410 way to implement (ROOM T).
413 When the compiler inline expands functions, it may be that different
414 kinds of return values are generated from different code branches.
415 E.g. an inline expansion of POSITION generates integer results
416 from one branch, and NIL results from another. When that inline
417 expansion is used in a context where only one of those results
420 (aref *a1* (position x *a2*)))
421 and the compiler can't prove that the unacceptable branch is
422 never taken, then bogus type mismatch warnings can be generated.
423 If you need to suppress the type mismatch warnings, you can
424 suppress the inline expansion,
426 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
427 (aref *a1* (position x *a2*)))
428 or, sometimes, suppress them by declaring the result to be of an
431 (aref *a1* (the integer (position x *a2*))))
433 This is not a new compiler problem in 0.7.0, but the new compiler
434 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
435 more conspicuous. If you don't need performance from these functions,
436 and the bogus warnings are a nuisance for you, you can return to
437 your pre-0.7.0 state of grace with
438 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
441 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
442 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
443 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
444 when of course it should be NIL. (He says it only fails for X86,
445 not SPARC; dunno about Alpha.)
447 Also, "the same problem exists for LONG-FLOAT-EPSILON,
448 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
449 for the -negative- the + is replaced by a - in the test)."
451 Raymond Toy comments that this is tricky on the X86 since its FPU
452 uses 80-bit precision internally.
455 Even in sbcl-0.pre7.x, which is supposed to be free of the old
456 non-ANSI behavior of treating the function return type inferred
457 from the current function definition as a declaration of the
458 return type from any function of that name, the return type of NIL
459 is attached to FOO in 120a above, and used to optimize code which
463 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
464 the entire lexical environment at the point of MACROLET available
465 in the bodies of the macroexpander functions. In particular, it
466 allows the function bodies (which run at compile time) to try to
467 access lexical variables (which are only defined at runtime).
468 It doesn't even issue a warning, which is bad.
470 The SBCL behavior arguably conforms to the ANSI spec (since the
471 spec says that the behavior is undefined, ergo anything conforms).
472 However, it would be better to issue a compile-time error.
473 Unfortunately I (WHN) don't see any simple way to detect this
474 condition in order to issue such an error, so for the meantime
475 SBCL just does this weird broken "conforming" thing.
477 The ANSI standard says, in the definition of the special operator
479 The macro-expansion functions defined by MACROLET are defined
480 in the lexical environment in which the MACROLET form appears.
481 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
482 the local macro definitions in a MACROLET, but the consequences
483 are undefined if the local macro definitions reference any
484 local variable or function bindings that are visible in that
486 Then it seems to contradict itself by giving the example
488 (macrolet ((fudge (z)
489 ;The parameters x and flag are not accessible
490 ; at this point; a reference to flag would be to
491 ; the global variable of that name.
492 ` (if flag (* ,z ,z) ,z)))
493 ;The parameters x and flag are accessible here.
497 The comment "a reference to flag would be to the global variable
498 of the same name" sounds like good behavior for the system to have.
499 but actual specification quoted above says that the actual behavior
502 (Since 0.7.8.23 macroexpanders are defined in a restricted version
503 of the lexical environment, containing no lexical variables and
504 functions, which seems to conform to ANSI and CLtL2, but signalling
505 a STYLE-WARNING for references to variables similar to locals might
509 (as reported by Gabe Garza on cmucl-help 2001-09-21)
511 (defun test-pred (x y)
515 (func (lambda () x)))
516 (print (eq func func))
517 (print (test-pred func func))
518 (delete func (list func))))
519 Now calling (TEST-CASE) gives output
522 (#<FUNCTION {500A9EF9}>)
523 Evidently Python thinks of the lambda as a code transformation so
524 much that it forgets that it's also an object.
527 Ideally, uninterning a symbol would allow it, and its associated
528 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
529 at least as of sbcl-0.7.0, this isn't the case. Information about
530 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
531 essentially in ordinary (non-weak) hash tables keyed by symbols.
532 Thus, once a system has an entry in this system, it tends to live
533 forever, even when it is uninterned and all other references to it
536 141: "pretty printing and backquote"
539 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
542 * (write '`(, .ala.) :readably t :pretty t)
545 (note the space between the comma and the point)
548 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
549 prominent SourceForge web/db bug tracking system, which is
550 unfortunately not a reliable way to get a timely response from
551 the SBCL maintainers)
552 In the course of trying to build a test case for an
553 application error, I encountered this behavior:
554 If you start up sbcl, and then lay on CTRL-C for a
555 minute or two, the lisp process will eventually say:
556 %PRIMITIVE HALT called; the party is over.
557 and throw you into the monitor. If I start up lisp,
558 attach to the process with strace, and then do the same
559 (abusive) thing, I get instead:
560 access failure in heap page not marked as write-protected
561 and the monitor again. I don't know enough to have the
562 faintest idea of what is going on here.
563 This is with sbcl 6.12, uname -a reports:
564 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
565 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
566 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
567 it took more than two minutes (but less than five) for me.
570 (This was once known as IR1-4, but it lived on even after the
571 IR1 interpreter went to the big bit bucket in the sky.)
572 The system accepts DECLAIM in most places where DECLARE would be
573 accepted, without even issuing a warning. ANSI allows this, but since
574 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
575 meaning is rather different, and it's unlikely that the user
576 has a good reason for doing DECLAIM not at top level, it would be
577 good to issue a STYLE-WARNING when this happens. A possible
578 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
579 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
580 [This is considered an IR1-interpreter-related bug because until
581 EVAL-WHEN is rewritten, which won't happen until after the IR1
582 interpreter is gone, the system's notion of what's a top-level form
583 and what's not will remain too confused to fix this problem.]
586 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
587 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
588 COMPLEX implementation didn't deal with this, and hasn't been
589 upgraded to do so. (This doesn't seem to be a high priority
590 conformance problem, since seems hard to construct useful code
594 Floating point errors are reported poorly. E.g. on x86 OpenBSD
597 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
598 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
599 No traps are enabled? How can this be?
600 It should be possible to be much more specific (overflow, division
601 by zero, etc.) and of course the "How can this be?" should be fixable.
603 See also bugs #45.c and #183
606 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
607 (in-package :cl-user)
610 (defstruct foo bar bletch)
612 (labels ((kidify1 (kid)
620 (declare (inline kid-frob))
623 (the simple-vector (foo-bar perd)))))
625 debugger invoked on condition of type TYPE-ERROR:
626 The value NIL is not of type SB-C::NODE.
627 The location of this failure has moved around as various related
628 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
629 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
631 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
632 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
633 of KIDIFY1, which already does not exist. So it gives up on
634 expansion, leaving garbage consisting of infinished blocks of the
635 partially converted function.)
637 (due to reordering of the compiler this example is compiled
638 successfully by 0.7.14, but the bug probably remains)
641 (reported by Robert E. Brown 2002-04-16)
642 When a function is called with too few arguments, causing the
643 debugger to be entered, the uninitialized slots in the bad call frame
644 seem to cause GCish problems, being interpreted as tagged data even
645 though they're not. In particular, executing ROOM in the
646 debugger at that point causes AVER failures:
649 * (lisp-implementation-version)
655 failed AVER: "(SAP= CURRENT END)"
656 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
657 isn't too surprising since there are many differences in stack
658 implementation and GC conservatism between the X86 and other ports.)
661 In sbcl-0.7.3.11, compiling the (illegal) code
662 (in-package :cl-user)
663 (defmethod prove ((uustk uustk))
666 gives the (not terribly clear) error message
668 ; (during macroexpansion of (DEFMETHOD PROVE ...))
669 ; can't get template for (FROB NIL NIL)
670 The problem seems to be that the code walker used by the DEFMETHOD
671 macro is unhappy with the illegal syntax in the method body, and
672 is giving an unclear error message.
675 The compiler sometimes tries to constant-fold expressions before
676 it checks to see whether they can be reached. This can lead to
677 bogus warnings about errors in the constant folding, e.g. in code
680 (WRITE-STRING (> X 0) "+" "0"))
681 compiled in a context where the compiler can prove that X is NIL,
682 and the compiler complains that (> X 0) causes a type error because
683 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
684 caused a full WARNING, which made the bug really annoying because then
685 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
686 code. Since then the warning has been downgraded to STYLE-WARNING,
687 so it's still a bug but at least it's a little less annoying.
689 183: "IEEE floating point issues"
690 Even where floating point handling is being dealt with relatively
691 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
692 accrued-exceptions and current-exceptions part of the fp control
693 word don't seem to bear much relation to reality. E.g. on
697 debugger invoked on condition of type DIVISION-BY-ZERO:
698 arithmetic error DIVISION-BY-ZERO signalled
699 0] (sb-vm::get-floating-point-modes)
701 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
702 :ROUNDING-MODE :NEAREST
703 :CURRENT-EXCEPTIONS NIL
704 :ACCRUED-EXCEPTIONS (:INEXACT)
707 * (sb-vm::get-floating-point-modes)
708 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
709 :ROUNDING-MODE :NEAREST
710 :CURRENT-EXCEPTIONS (:INEXACT)
711 :ACCRUED-EXCEPTIONS (:INEXACT)
714 188: "compiler performance fiasco involving type inference and UNION-TYPE"
715 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
716 bug could be fixed properly, so you won't see the bug unless you restore
717 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
721 (declare (optimize (safety 3)))
722 (declare (optimize (compilation-speed 2)))
723 (declare (optimize (speed 1) (debug 1) (space 1)))
724 (let ((fn "if-this-file-exists-the-universe-is-strange"))
725 (load fn :if-does-not-exist nil)
726 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
727 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
728 (load (concatenate 'string fn ".misc-garbage")
729 :if-does-not-exist nil)))))
731 134.552 seconds of real time
732 133.35156 seconds of user run time
733 0.03125 seconds of system run time
734 [Run times include 2.787 seconds GC run time.]
736 246883368 bytes consed.
737 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
738 thinking about type relationships involving types like
740 (OR (INTEGER 576 576)
751 In recent SBCL the following example also illustrates this bug:
756 (declare (optimize (safety 3)))
757 (declare (optimize (compilation-speed 2)))
758 (declare (optimize (speed 1) (debug 1) (space 1)))
760 (declare (type (integer 0) start))
761 (print (incf start 22))
762 (print (incf start 26))
763 (print (incf start 28)))
765 (declare (type (integer 0) start))
766 (print (incf start 22))
767 (print (incf start 26)))
769 (declare (type (integer 0) start))
770 (print (incf start 22))
771 (print (incf start 26))))))
773 190: "PPC/Linux pipe? buffer? bug"
774 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
775 on the PPC/Linux platform, waiting for a zombie env process. This
776 is a classic symptom of buffer filling and deadlock, but it seems
777 only sporadically reproducible.
779 191: "Miscellaneous PCL deficiencies"
780 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
781 a. DEFCLASS does not inform the compiler about generated
782 functions. Compiling a file with
786 (WITH-SLOTS (A-CLASS-X) A
788 results in a STYLE-WARNING:
790 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
792 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
793 Mai points out that the declamation of functions is in fact
794 incorrect in some cases (most notably for structure
795 classes). This means that at present erroneous attempts to use
796 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
797 won't get the corresponding STYLE-WARNING.
798 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
799 lists and &KEY arguments) do not signal errors when they should.
802 201: "Incautious type inference from compound CONS types"
803 (reported by APD sbcl-devel 2002-09-17)
805 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
807 (FORMAT NIL "~S IS ~S, Y = ~S"
814 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
816 205: "environment issues in cross compiler"
817 (These bugs have no impact on user code, but should be fixed or
819 a. Macroexpanders introduced with MACROLET are defined in the null
821 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
822 the null lexical environment.
823 c. The cross-compiler cannot inline functions defined in a non-null
826 206: ":SB-FLUID feature broken"
827 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
828 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
831 207: "poorly distributed SXHASH results for compound data"
832 SBCL's SXHASH could probably try a little harder. ANSI: "the
833 intent is that an implementation should make a good-faith
834 effort to produce hash-codes that are well distributed
835 within the range of non-negative fixnums". But
836 (let ((hits (make-hash-table)))
839 (let* ((ij (cons i j))
840 (newlist (push ij (gethash (sxhash ij) hits))))
842 (format t "~&collision: ~S~%" newlist))))))
843 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
844 would be an obvious way of fix. Maybe it would be acceptably efficient
845 to redo MIX using a lookup into a 256-entry s-box containing
846 29-bit pseudorandom numbers?
848 211: "keywords processing"
849 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
850 number of keyword arguments.
853 (flet ((foo (&key y) (list y)))
854 (list (foo :y 1 :y 2)))
856 issues confusing message
861 ; caught STYLE-WARNING:
862 ; The variable #:G15 is defined but never used.
864 212: "Sequence functions and circular arguments"
865 COERCE, MERGE and CONCATENATE go into an infinite loop when given
866 circular arguments; it would be good for the user if they could be
867 given an error instead (ANSI 17.1.1 allows this behaviour on the part
868 of the implementation, as conforming code cannot give non-proper
869 sequences to these functions. MAP also has this problem (and
870 solution), though arguably the convenience of being able to do
871 (MAP 'LIST '+ FOO '#1=(1 . #1#))
872 might be classed as more important (though signalling an error when
873 all of the arguments are circular is probably desireable).
875 213: "Sequence functions and type checking"
876 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
877 various complicated, though recognizeable, CONS types [e.g.
878 (CONS * (CONS * NULL))
879 which according to ANSI should be recognized] (and, in SAFETY 3
880 code, should return a list of LENGTH 2 or signal an error)
881 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
882 check that it will return a sequence of the given type. Fixing
883 it along the same lines as the others (cf. work done around
884 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
885 entirely straightforward.
886 c. All of these functions will silently accept a type of the form
888 whether or not the return value is of this type. This is
889 probably permitted by ANSI (see "Exceptional Situations" under
890 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
891 know about this escape clause, so code of the form
892 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
893 can erroneously return T.
896 SBCL 0.6.12.43 fails to compile
899 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
900 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
903 or a more simple example:
906 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
907 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
909 215: ":TEST-NOT handling by functions"
910 a. FIND and POSITION currently signal errors when given non-NIL for
911 both their :TEST and (deprecated) :TEST-NOT arguments, but by
912 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
913 means that the effect is "unpredictable but harmless". It's not
914 clear what that actually means; it may preclude conforming
915 implementations from signalling errors.
916 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
917 when conflict occurs. As a quality of implementation issue, it
918 might be preferable to treat :TEST and :TEST-NOT as being in some
919 sense the same &KEY, and effectively take the first test function in
921 c. Again, a quality of implementation issue: it would be good to issue a
922 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
923 WARNING for calls with both :TEST and :TEST-NOT; possibly this
924 latter should be WARNed about at execute-time too.
926 216: "debugger confused by frames with invalid number of arguments"
927 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
928 leaves the system confused, enough so that (QUIT) no longer works.
929 It's as though the process of working with the uninitialized slot in
930 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
931 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
933 217: "Bad type operations with FUNCTION types"
936 * (values-type-union (specifier-type '(function (base-char)))
937 (specifier-type '(function (integer))))
939 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
941 It causes insertion of wrong type assertions into generated
945 (let ((f (etypecase x
946 (character #'write-char)
947 (integer #'write-byte))))
950 (character (write-char x s))
951 (integer (write-byte x s)))))
953 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
955 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
956 produce invalid code, but type checking is not accurate. Similar
957 problems exist with VALUES-TYPE-INTERSECTION.)
960 Sbcl 0.7.9 fails to compile
962 (multiple-value-call #'list
963 (the integer (helper))
966 Type check for INTEGER, the result of which serves as the first
967 argument of M-V-C, is inserted after evaluation of NIL. So arguments
968 of M-V-C are pushed in the wrong order. As a temporary workaround
969 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
970 would be to put the check between evaluation of arguments, but it
971 could be tricky to check result types of PROG1, IF etc.
973 233: bugs in constraint propagation
976 (declare (optimize (speed 2) (safety 3)))
982 (quux y (+ y 2d0) (* y 3d0)))))
983 (foo 4) => segmentation violation
985 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
989 (declaim (optimize (speed 2) (safety 3)))
991 (if (typep (prog1 x (setq x y)) 'double-float)
994 (foo 1d0 5) => segmentation violation
996 235: "type system and inline expansion"
998 (declaim (ftype (function (cons) number) acc))
999 (declaim (inline acc))
1001 (the number (car c)))
1004 (values (locally (declare (optimize (safety 0)))
1006 (locally (declare (optimize (safety 3)))
1009 (foo '(nil) '(t)) => NIL, T.
1011 b. (reported by brown on #lisp 2003-01-21)
1014 (declare (optimize (speed 3) (safety 0)))
1015 (declare (notinline mapcar))
1016 (let ((z (mapcar #'car x)))
1019 Without (DECLARE (NOTINLINE MAPCAR)), Python cannot derive that Z is
1022 237: "Environment arguments to type functions"
1023 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
1024 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
1025 argument, but they ignore it completely. This is almost
1026 certainly not correct.
1027 b. Also, the compiler's optimizers for TYPEP have not been informed
1028 about the new argument; consequently, they will not transform
1029 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
1030 just as optimizeable as (TYPEP 1 'INTEGER).
1032 238: "REPL compiler overenthusiasm for CLOS code"
1034 * (defclass foo () ())
1035 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
1036 causes approximately 100 lines of code deletion notes. Some
1037 discussion on this issue happened under the title 'Three "interesting"
1038 bugs in PCL', resulting in a fix for this oververbosity from the
1039 compiler proper; however, the problem persists in the interactor
1040 because the notion of original source is not preserved: for the
1041 compiler, the original source of the above expression is (DEFMETHOD
1042 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
1043 compiler gets its hands on the code needing compilation from the REPL,
1044 it has been macroexpanded several times.
1046 A symptom of the same underlying problem, reported by Tony Martinez:
1048 (with-input-from-string (*query-io* " no")
1050 (simple-type-error () 'error))
1052 ; (SB-KERNEL:FLOAT-WAIT)
1054 ; note: deleting unreachable code
1055 ; compilation unit finished
1058 241: "DEFCLASS mysteriously remembers uninterned accessor names."
1059 (from tonyms on #lisp IRC 2003-02-25)
1060 In sbcl-0.7.12.55, typing
1061 (defclass foo () ((bar :accessor foo-bar)))
1064 (defclass foo () ((bar :accessor foo-bar)))
1065 gives the error message
1066 "#:FOO-BAR already names an ordinary function or a macro."
1067 So it's somehow checking the uninterned old accessor name instead
1068 of the new requested accessor name, which seems broken to me (WHN).
1070 242: "WRITE-SEQUENCE suboptimality"
1071 (observed from clx performance)
1072 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
1073 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
1074 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
1075 rather than writing the sequence in one go, leading to severe
1076 performance degradation.
1078 243: "STYLE-WARNING overenthusiasm for unused variables"
1079 (observed from clx compilation)
1080 In sbcl-0.7.14, in the presence of the macros
1081 (DEFMACRO FOO (X) `(BAR ,X))
1082 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
1083 somewhat surprising style warnings are emitted for
1084 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
1086 ; (LAMBDA (Y) (FOO Y))
1088 ; caught STYLE-WARNING:
1089 ; The variable Y is defined but never used.
1091 245: bugs in disassembler
1092 a. On X86 an immediate operand for IMUL is printed incorrectly.
1093 b. On X86 operand size prefix is not recognized.
1095 246: "NTH-VALUE scaling problem"
1096 NTH-VALUE's current implementation for constant integers scales in
1097 compile-time as O(n^4), as indeed must the optional dispatch
1098 mechanism on which it is implemented. While it is unlikely to
1099 matter in real user code, it's still unpleasant to observe that
1100 (NTH-VALUE 1000 (VALUES-LIST (MAKE-LIST 1001))) takes several hours
1103 248: "reporting errors in type specifier syntax"
1104 (TYPEP 1 '(SYMBOL NIL)) says something about "unknown type
1108 (defun foo (&key (a :x))
1109 (declare (fixnum a))
1112 does not cause a warning. (BTW: old SBCL issued a warning, but for a
1113 function, which was never called!)
1115 253: "type checking is embedded THEs"
1116 Compiler cannot perform type checking in
1118 (let () (list (the fixnum (the unsigned-byte (eval -1)))))
1122 DEFUNCT CATEGORIES OF BUGS
1124 These labels were used for bugs related to the old IR1 interpreter.
1125 The # values reached 6 before the category was closed down.