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 It would be nice if the
114 (during macroexpansion)
115 said what macroexpansion was at fault, e.g.
117 (during macroexpansion of IN-PACKAGE,
118 during macroexpansion of DEFFOO)
121 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
122 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
123 (Also, when this is fixed, we can enable the code in PROCLAIM which
124 checks for incompatible FTYPE redeclarations.)
127 (I *think* this is a bug. It certainly seems like strange behavior. But
128 the ANSI spec is scary, dark, and deep.. -- WHN)
129 (FORMAT NIL "~,1G" 1.4) => "1. "
130 (FORMAT NIL "~3,1G" 1.4) => "1. "
133 Sometimes (SB-EXT:QUIT) fails with
134 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
135 Process inferior-lisp exited abnormally with code 1
136 I haven't noticed a repeatable case of this yet.
139 The printer doesn't report closures very well. This is true in
143 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
144 It would be nice to make closures have a settable name slot,
145 and make things like DEFSTRUCT and FLET, which create closures,
146 set helpful values into this slot.
149 And as long as we're wishing, it would be awfully nice if INSPECT could
150 also report on closures, telling about the values of the bound variables.
153 The compiler assumes that any time a function of declared FTYPE
154 doesn't signal an error, its arguments were of the declared type.
155 E.g. compiling and loading
156 (DECLAIM (OPTIMIZE (SAFETY 3)))
157 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
159 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
161 (COND ((> (FACTORIAL X) 1.0E6)
162 (FORMAT T "too big~%"))
164 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
166 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
169 will cause the INTEGERP case to be selected, giving bogus output a la
171 This violates the "declarations are assertions" principle.
172 According to the ANSI spec, in the section "System Class FUNCTION",
173 this is a case of "lying to the compiler", but the lying is done
174 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
175 not by the unexpectedly general definition of FACTORIAL. In any case,
176 "declarations are assertions" means that lying to the compiler should
177 cause an error to be signalled, and should not cause a bogus
178 result to be returned. Thus, the compiler should not assume
179 that arbitrary functions check their argument types. (It might
180 make sense to add another flag (CHECKED?) to DEFKNOWN to
181 identify functions which *do* check their argument types.)
182 (Also, verify that the compiler handles declared function
183 return types as assertions.)
186 The definitions of SIGCONTEXT-FLOAT-REGISTER and
187 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
188 supported on FreeBSD because the floating point state is not saved,
189 but at least as of FreeBSD 4.0, the floating point state *is* saved,
190 so they could be supported after all. Very likely
191 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
194 a slew of floating-point-related errors reported by Peter Van Eynde
196 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT on the x86 is
197 bogus, and should probably be 1.4012985e-45. In SBCL,
198 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
199 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
200 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
201 and LEAST-NEGATIVE-LONG-FLOAT.
202 c: Many expressions generate floating infinity on x86/Linux:
207 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
208 on x86/Linux generates the infinities instead. That might or
209 might not be conforming behavior, but it's also inconsistent,
210 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
211 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
212 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
214 d: (in section12.erg) various forms a la
215 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
216 don't give the right behavior.
219 type safety errors reported by Peter Van Eynde July 25, 2000:
220 c: (COERCE 'AND 'FUNCTION) returns something related to
221 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
222 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
223 not a binary input stream, but instead cheerfully reads from
224 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
227 The debugger LIST-LOCATIONS command doesn't work properly.
230 Compiling and loading
231 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
233 then requesting a BACKTRACE at the debugger prompt gives no information
234 about where in the user program the problem occurred.
237 Using the pretty-printer from the command prompt gives funny
238 results, apparently because the pretty-printer doesn't know
239 about user's command input, including the user's carriage return
240 that the user, and therefore the pretty-printer thinks that
241 the new output block should start indented 2 or more characters
242 rightward of the correct location.
245 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
246 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
247 crashes SBCL. In general tracing anything which is used in the
248 implementation of TRACE is likely to have the same problem.
251 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
252 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
253 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
254 WITH-OUTPUT-TO-STRING.
257 ANSI says in one place that type declarations can be abbreviated even
258 when the type name is not a symbol, e.g.
259 (DECLAIM ((VECTOR T) *FOOVECTOR*))
260 SBCL doesn't support this. But ANSI says in another place that this
261 isn't allowed. So it's not clear this is a bug after all. (See the
262 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
266 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
267 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
268 an easily guessable temporary filename in a way which might open
269 applications using LOAD-FOREIGN to hijacking by malicious users
270 on the same machine. Incantations for doing this safely are
271 floating around the net in various "how to write secure programs
272 despite Unix" documents, and it would be good to (1) fix this in
273 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
274 files and make it share the same new safe logic.
276 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
277 make the temporary filename less easily guessable)
280 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
281 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
282 the mean of the distribution can be systematically O(0.1%) wrong.
283 Just increasing R-I-E-B is probably not a good solution, since
284 it would decrease efficiency more than is probably necessary. Perhaps
285 using some sort of accept/reject method would be better.
288 Internally the compiler sometimes evaluates
289 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
290 (I stumbled across this when I added an
291 (assert (not (eq type1 *wild-type*)))
292 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
293 in a type context should probably be translated to T, and so it's
294 probably wrong to ask whether it's equal to the T type and then (using
295 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
296 (I haven't tried to investigate this bug enough to guess whether
297 there might be any user-level symptoms.)
299 In fact, the type system is likely to depend on this inequality not
300 holding... * is not equivalent to T in many cases, such as
301 (VECTOR *) /= (VECTOR T).
304 Inconsistencies between derived and declared VALUES return types for
305 DEFUN aren't checked very well. E.g. the logic which successfully
306 catches problems like
307 (declaim (ftype (function (fixnum) float) foo))
309 (declare (type integer x))
310 (values x)) ; wrong return type, detected, gives warning, good!
312 (declaim (ftype (function (t) (values t t)) bar))
314 (values x)) ; wrong number of return values, no warning, bad!
315 The cause of this is seems to be that (1) the internal function
316 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
317 arguments symmetrically, and (2) when the type checking code was
318 written back when when SBCL's code was still CMU CL, the intent
320 (declaim (ftype (function (t) t) bar))
322 (values x x)) ; wrong number of return values; should give warning?
323 not be warned for, because a two-valued return value is considered
324 to be compatible with callers who expects a single value to be
325 returned. That intent is probably not appropriate for modern ANSI
326 Common Lisp, but fixing this might be complicated because of other
327 divergences between auld-style and new-style handling of
328 multiple-VALUES types. (Some issues related to this were discussed
329 on cmucl-imp at some length sometime in 2000.)
332 The facility for dumping a running Lisp image to disk gets confused
333 when run without the PURIFY option, and creates an unnecessarily large
334 core file (apparently representing memory usage up to the previous
335 high-water mark). Moreover, when the file is loaded, it confuses the
336 GC, so that thereafter memory usage can never be reduced below that
340 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
341 CL), out-of-line structure slot setters are horribly inefficient
342 whenever the type of the slot is declared, because out-of-line
343 structure slot setters are implemented as closures to save space,
344 so the compiler doesn't compile the type test into code, but
345 instead just saves the type in a lexical closure and interprets it
347 To exercise the problem, compile and load
348 (cl:in-package :cl-user)
350 (bar (error "missing") :type bar))
353 (loop (setf (foo-bar *foo*) x)))
355 (defvar *bar* (make-bar))
356 (defvar *foo* (make-foo :bar *bar*))
357 (defvar *setf-foo-bar* #'(setf foo-bar))
359 (loop (funcall *setf-foo-bar* x *foo*)))
360 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
361 use BACKTRACE, to see it's spending all essentially all its time
362 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
363 One possible solution would be simply to give up on
364 representing structure slot accessors as functions, and represent
365 them as macroexpansions instead. This can be inconvenient for users,
366 but it's not clear that it's worse than trying to help by expanding
367 into a horribly inefficient implementation.
368 As a workaround for the problem, #'(SETF FOO) expressions
369 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
370 (defmacro efficient-setf-function (place-function-name)
371 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
372 ;; a workaround for the problem, encouraging the
373 ;; inline expansion of the structure accessor, so
374 ;; that the compiler can optimize its type test
375 (let ((new-value (gensym "NEW-VALUE-"))
376 (structure-value (gensym "STRUCTURE-VALUE-")))
377 `(lambda (,new-value ,structure-value)
378 (setf (,place-function-name ,structure-value)
380 ;; no problem, can just use the ordinary expansion
381 `(function (setf ,place-function-name))))
384 There's apparently a bug in CEILING optimization which caused
385 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
386 applied the patches to SBCL and they didn't seem to cause problems
387 (as reported sbcl-devel 2001-05-04). However, since the patches
388 modify nontrivial code which was apparently written incorrectly
389 the first time around, until regression tests are written I'm not
390 comfortable merging the patches in the CVS version of SBCL.
393 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
394 a clean, just-started SBCL system. And it seems to be right:
395 (ROOM T) can bring a small computer to its knees for a *long*
396 time trying to GC afterwards. Surely there's some more economical
397 way to implement (ROOM T).
400 When the compiler inline expands functions, it may be that different
401 kinds of return values are generated from different code branches.
402 E.g. an inline expansion of POSITION generates integer results
403 from one branch, and NIL results from another. When that inline
404 expansion is used in a context where only one of those results
407 (aref *a1* (position x *a2*)))
408 and the compiler can't prove that the unacceptable branch is
409 never taken, then bogus type mismatch warnings can be generated.
410 If you need to suppress the type mismatch warnings, you can
411 suppress the inline expansion,
413 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
414 (aref *a1* (position x *a2*)))
415 or, sometimes, suppress them by declaring the result to be of an
418 (aref *a1* (the integer (position x *a2*))))
420 This is not a new compiler problem in 0.7.0, but the new compiler
421 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
422 more conspicuous. If you don't need performance from these functions,
423 and the bogus warnings are a nuisance for you, you can return to
424 your pre-0.7.0 state of grace with
425 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
428 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
429 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
430 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
431 when of course it should be NIL. (He says it only fails for X86,
432 not SPARC; dunno about Alpha.)
434 Also, "the same problem exists for LONG-FLOAT-EPSILON,
435 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
436 for the -negative- the + is replaced by a - in the test)."
438 Raymond Toy comments that this is tricky on the X86 since its FPU
439 uses 80-bit precision internally.
442 Even in sbcl-0.pre7.x, which is supposed to be free of the old
443 non-ANSI behavior of treating the function return type inferred
444 from the current function definition as a declaration of the
445 return type from any function of that name, the return type of NIL
446 is attached to FOO in 120a above, and used to optimize code which
450 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
451 the entire lexical environment at the point of MACROLET available
452 in the bodies of the macroexpander functions. In particular, it
453 allows the function bodies (which run at compile time) to try to
454 access lexical variables (which are only defined at runtime).
455 It doesn't even issue a warning, which is bad.
457 The SBCL behavior arguably conforms to the ANSI spec (since the
458 spec says that the behavior is undefined, ergo anything conforms).
459 However, it would be better to issue a compile-time error.
460 Unfortunately I (WHN) don't see any simple way to detect this
461 condition in order to issue such an error, so for the meantime
462 SBCL just does this weird broken "conforming" thing.
464 The ANSI standard says, in the definition of the special operator
466 The macro-expansion functions defined by MACROLET are defined
467 in the lexical environment in which the MACROLET form appears.
468 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
469 the local macro definitions in a MACROLET, but the consequences
470 are undefined if the local macro definitions reference any
471 local variable or function bindings that are visible in that
473 Then it seems to contradict itself by giving the example
475 (macrolet ((fudge (z)
476 ;The parameters x and flag are not accessible
477 ; at this point; a reference to flag would be to
478 ; the global variable of that name.
479 ` (if flag (* ,z ,z) ,z)))
480 ;The parameters x and flag are accessible here.
484 The comment "a reference to flag would be to the global variable
485 of the same name" sounds like good behavior for the system to have.
486 but actual specification quoted above says that the actual behavior
489 (Since 0.7.8.23 macroexpanders are defined in a restricted version
490 of the lexical environment, containing no lexical variables and
491 functions, which seems to conform to ANSI and CLtL2, but signalling
492 a STYLE-WARNING for references to variables similar to locals might
496 (as reported by Gabe Garza on cmucl-help 2001-09-21)
498 (defun test-pred (x y)
502 (func (lambda () x)))
503 (print (eq func func))
504 (print (test-pred func func))
505 (delete func (list func))))
506 Now calling (TEST-CASE) gives output
509 (#<FUNCTION {500A9EF9}>)
510 Evidently Python thinks of the lambda as a code transformation so
511 much that it forgets that it's also an object.
514 Ideally, uninterning a symbol would allow it, and its associated
515 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
516 at least as of sbcl-0.7.0, this isn't the case. Information about
517 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
518 essentially in ordinary (non-weak) hash tables keyed by symbols.
519 Thus, once a system has an entry in this system, it tends to live
520 forever, even when it is uninterned and all other references to it
523 141: "pretty printing and backquote"
526 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
529 * (write '`(, .ala.) :readably t :pretty t)
532 (note the space between the comma and the point)
535 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
536 prominent SourceForge web/db bug tracking system, which is
537 unfortunately not a reliable way to get a timely response from
538 the SBCL maintainers)
539 In the course of trying to build a test case for an
540 application error, I encountered this behavior:
541 If you start up sbcl, and then lay on CTRL-C for a
542 minute or two, the lisp process will eventually say:
543 %PRIMITIVE HALT called; the party is over.
544 and throw you into the monitor. If I start up lisp,
545 attach to the process with strace, and then do the same
546 (abusive) thing, I get instead:
547 access failure in heap page not marked as write-protected
548 and the monitor again. I don't know enough to have the
549 faintest idea of what is going on here.
550 This is with sbcl 6.12, uname -a reports:
551 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
552 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
553 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
554 it took more than two minutes (but less than five) for me.
557 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
558 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
559 COMPLEX implementation didn't deal with this, and hasn't been
560 upgraded to do so. (This doesn't seem to be a high priority
561 conformance problem, since seems hard to construct useful code
565 Floating point errors are reported poorly. E.g. on x86 OpenBSD
568 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
569 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
570 No traps are enabled? How can this be?
571 It should be possible to be much more specific (overflow, division
572 by zero, etc.) and of course the "How can this be?" should be fixable.
574 See also bugs #45.c and #183
577 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
578 (in-package :cl-user)
581 (defstruct foo bar bletch)
583 (labels ((kidify1 (kid)
591 (declare (inline kid-frob))
594 (the simple-vector (foo-bar perd)))))
596 debugger invoked on condition of type TYPE-ERROR:
597 The value NIL is not of type SB-C::NODE.
598 The location of this failure has moved around as various related
599 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
600 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
602 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
603 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
604 of KIDIFY1, which already does not exist. So it gives up on
605 expansion, leaving garbage consisting of infinished blocks of the
606 partially converted function.)
608 (due to reordering of the compiler this example is compiled
609 successfully by 0.7.14, but the bug probably remains)
612 (reported by Robert E. Brown 2002-04-16)
613 When a function is called with too few arguments, causing the
614 debugger to be entered, the uninitialized slots in the bad call frame
615 seem to cause GCish problems, being interpreted as tagged data even
616 though they're not. In particular, executing ROOM in the
617 debugger at that point causes AVER failures:
620 * (lisp-implementation-version)
626 failed AVER: "(SAP= CURRENT END)"
627 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
628 isn't too surprising since there are many differences in stack
629 implementation and GC conservatism between the X86 and other ports.)
632 In sbcl-0.7.3.11, compiling the (illegal) code
633 (in-package :cl-user)
634 (defmethod prove ((uustk uustk))
637 gives the (not terribly clear) error message
639 ; (during macroexpansion of (DEFMETHOD PROVE ...))
640 ; can't get template for (FROB NIL NIL)
641 The problem seems to be that the code walker used by the DEFMETHOD
642 macro is unhappy with the illegal syntax in the method body, and
643 is giving an unclear error message.
646 The compiler sometimes tries to constant-fold expressions before
647 it checks to see whether they can be reached. This can lead to
648 bogus warnings about errors in the constant folding, e.g. in code
651 (WRITE-STRING (> X 0) "+" "0"))
652 compiled in a context where the compiler can prove that X is NIL,
653 and the compiler complains that (> X 0) causes a type error because
654 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
655 caused a full WARNING, which made the bug really annoying because then
656 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
657 code. Since then the warning has been downgraded to STYLE-WARNING,
658 so it's still a bug but at least it's a little less annoying.
660 183: "IEEE floating point issues"
661 Even where floating point handling is being dealt with relatively
662 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
663 accrued-exceptions and current-exceptions part of the fp control
664 word don't seem to bear much relation to reality. E.g. on
668 debugger invoked on condition of type DIVISION-BY-ZERO:
669 arithmetic error DIVISION-BY-ZERO signalled
670 0] (sb-vm::get-floating-point-modes)
672 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
673 :ROUNDING-MODE :NEAREST
674 :CURRENT-EXCEPTIONS NIL
675 :ACCRUED-EXCEPTIONS (:INEXACT)
678 * (sb-vm::get-floating-point-modes)
679 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
680 :ROUNDING-MODE :NEAREST
681 :CURRENT-EXCEPTIONS (:INEXACT)
682 :ACCRUED-EXCEPTIONS (:INEXACT)
685 188: "compiler performance fiasco involving type inference and UNION-TYPE"
689 (declare (optimize (safety 3)))
690 (declare (optimize (compilation-speed 2)))
691 (declare (optimize (speed 1) (debug 1) (space 1)))
693 (declare (type (integer 0) start))
694 (print (incf start 22))
695 (print (incf start 26))
696 (print (incf start 28)))
698 (declare (type (integer 0) start))
699 (print (incf start 22))
700 (print (incf start 26)))
702 (declare (type (integer 0) start))
703 (print (incf start 22))
704 (print (incf start 26))))))
706 190: "PPC/Linux pipe? buffer? bug"
707 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
708 on the PPC/Linux platform, waiting for a zombie env process. This
709 is a classic symptom of buffer filling and deadlock, but it seems
710 only sporadically reproducible.
712 191: "Miscellaneous PCL deficiencies"
713 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
714 a. DEFCLASS does not inform the compiler about generated
715 functions. Compiling a file with
719 (WITH-SLOTS (A-CLASS-X) A
721 results in a STYLE-WARNING:
723 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
725 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
726 Mai points out that the declamation of functions is in fact
727 incorrect in some cases (most notably for structure
728 classes). This means that at present erroneous attempts to use
729 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
730 won't get the corresponding STYLE-WARNING.
731 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
732 lists and &KEY arguments) do not signal errors when they should.
735 201: "Incautious type inference from compound types"
736 a. (reported by APD sbcl-devel 2002-09-17)
738 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
740 (FORMAT NIL "~S IS ~S, Y = ~S"
747 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
751 (declare (type (array * (4 4)) x))
753 (setq x (make-array '(4 4)))
754 (adjust-array y '(3 5))
755 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
757 * (foo (make-array '(4 4) :adjustable t))
760 205: "environment issues in cross compiler"
761 (These bugs have no impact on user code, but should be fixed or
763 a. Macroexpanders introduced with MACROLET are defined in the null
765 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
766 the null lexical environment.
767 c. The cross-compiler cannot inline functions defined in a non-null
770 206: ":SB-FLUID feature broken"
771 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
772 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
775 207: "poorly distributed SXHASH results for compound data"
776 SBCL's SXHASH could probably try a little harder. ANSI: "the
777 intent is that an implementation should make a good-faith
778 effort to produce hash-codes that are well distributed
779 within the range of non-negative fixnums". But
780 (let ((hits (make-hash-table)))
783 (let* ((ij (cons i j))
784 (newlist (push ij (gethash (sxhash ij) hits))))
786 (format t "~&collision: ~S~%" newlist))))))
787 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
788 would be an obvious way of fix. Maybe it would be acceptably efficient
789 to redo MIX using a lookup into a 256-entry s-box containing
790 29-bit pseudorandom numbers?
792 211: "keywords processing"
793 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
794 number of keyword arguments.
797 (flet ((foo (&key y) (list y)))
798 (list (foo :y 1 :y 2)))
800 issues confusing message
805 ; caught STYLE-WARNING:
806 ; The variable #:G15 is defined but never used.
808 212: "Sequence functions and circular arguments"
809 COERCE, MERGE and CONCATENATE go into an infinite loop when given
810 circular arguments; it would be good for the user if they could be
811 given an error instead (ANSI 17.1.1 allows this behaviour on the part
812 of the implementation, as conforming code cannot give non-proper
813 sequences to these functions. MAP also has this problem (and
814 solution), though arguably the convenience of being able to do
815 (MAP 'LIST '+ FOO '#1=(1 . #1#))
816 might be classed as more important (though signalling an error when
817 all of the arguments are circular is probably desireable).
819 213: "Sequence functions and type checking"
820 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
821 various complicated, though recognizeable, CONS types [e.g.
822 (CONS * (CONS * NULL))
823 which according to ANSI should be recognized] (and, in SAFETY 3
824 code, should return a list of LENGTH 2 or signal an error)
825 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
826 check that it will return a sequence of the given type. Fixing
827 it along the same lines as the others (cf. work done around
828 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
829 entirely straightforward.
830 c. All of these functions will silently accept a type of the form
832 whether or not the return value is of this type. This is
833 probably permitted by ANSI (see "Exceptional Situations" under
834 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
835 know about this escape clause, so code of the form
836 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
837 can erroneously return T.
840 SBCL 0.6.12.43 fails to compile
843 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
844 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
847 or a more simple example:
850 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
851 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
853 215: ":TEST-NOT handling by functions"
854 a. FIND and POSITION currently signal errors when given non-NIL for
855 both their :TEST and (deprecated) :TEST-NOT arguments, but by
856 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
857 means that the effect is "unpredictable but harmless". It's not
858 clear what that actually means; it may preclude conforming
859 implementations from signalling errors.
860 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
861 when conflict occurs. As a quality of implementation issue, it
862 might be preferable to treat :TEST and :TEST-NOT as being in some
863 sense the same &KEY, and effectively take the first test function in
865 c. Again, a quality of implementation issue: it would be good to issue a
866 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
867 WARNING for calls with both :TEST and :TEST-NOT; possibly this
868 latter should be WARNed about at execute-time too.
870 216: "debugger confused by frames with invalid number of arguments"
871 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
872 leaves the system confused, enough so that (QUIT) no longer works.
873 It's as though the process of working with the uninitialized slot in
874 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
875 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
877 217: "Bad type operations with FUNCTION types"
880 * (values-type-union (specifier-type '(function (base-char)))
881 (specifier-type '(function (integer))))
883 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
885 It causes insertion of wrong type assertions into generated
889 (let ((f (etypecase x
890 (character #'write-char)
891 (integer #'write-byte))))
894 (character (write-char x s))
895 (integer (write-byte x s)))))
897 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
899 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
900 produce invalid code, but type checking is not accurate. Similar
901 problems exist with VALUES-TYPE-INTERSECTION.)
904 Sbcl 0.7.9 fails to compile
906 (multiple-value-call #'list
907 (the integer (helper))
910 Type check for INTEGER, the result of which serves as the first
911 argument of M-V-C, is inserted after evaluation of NIL. So arguments
912 of M-V-C are pushed in the wrong order. As a temporary workaround
913 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
914 would be to put the check between evaluation of arguments, but it
915 could be tricky to check result types of PROG1, IF etc.
917 233: bugs in constraint propagation
920 (declare (optimize (speed 2) (safety 3)))
926 (quux y (+ y 2d0) (* y 3d0)))))
927 (foo 4) => segmentation violation
929 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
933 (declaim (optimize (speed 2) (safety 3)))
935 (if (typep (prog1 x (setq x y)) 'double-float)
938 (foo 1d0 5) => segmentation violation
940 235: "type system and inline expansion"
942 (declaim (ftype (function (cons) number) acc))
943 (declaim (inline acc))
945 (the number (car c)))
948 (values (locally (declare (optimize (safety 0)))
950 (locally (declare (optimize (safety 3)))
953 (foo '(nil) '(t)) => NIL, T.
955 237: "Environment arguments to type functions"
956 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
957 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
958 argument, but they ignore it completely. This is almost
959 certainly not correct.
960 b. Also, the compiler's optimizers for TYPEP have not been informed
961 about the new argument; consequently, they will not transform
962 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
963 just as optimizeable as (TYPEP 1 'INTEGER).
965 238: "REPL compiler overenthusiasm for CLOS code"
967 * (defclass foo () ())
968 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
969 causes approximately 100 lines of code deletion notes. Some
970 discussion on this issue happened under the title 'Three "interesting"
971 bugs in PCL', resulting in a fix for this oververbosity from the
972 compiler proper; however, the problem persists in the interactor
973 because the notion of original source is not preserved: for the
974 compiler, the original source of the above expression is (DEFMETHOD
975 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
976 compiler gets its hands on the code needing compilation from the REPL,
977 it has been macroexpanded several times.
979 A symptom of the same underlying problem, reported by Tony Martinez:
981 (with-input-from-string (*query-io* " no")
983 (simple-type-error () 'error))
985 ; (SB-KERNEL:FLOAT-WAIT)
987 ; note: deleting unreachable code
988 ; compilation unit finished
991 241: "DEFCLASS mysteriously remembers uninterned accessor names."
992 (from tonyms on #lisp IRC 2003-02-25)
993 In sbcl-0.7.12.55, typing
994 (defclass foo () ((bar :accessor foo-bar)))
997 (defclass foo () ((bar :accessor foo-bar)))
998 gives the error message
999 "#:FOO-BAR already names an ordinary function or a macro."
1000 So it's somehow checking the uninterned old accessor name instead
1001 of the new requested accessor name, which seems broken to me (WHN).
1003 242: "WRITE-SEQUENCE suboptimality"
1004 (observed from clx performance)
1005 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
1006 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
1007 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
1008 rather than writing the sequence in one go, leading to severe
1009 performance degradation.
1011 243: "STYLE-WARNING overenthusiasm for unused variables"
1012 (observed from clx compilation)
1013 In sbcl-0.7.14, in the presence of the macros
1014 (DEFMACRO FOO (X) `(BAR ,X))
1015 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
1016 somewhat surprising style warnings are emitted for
1017 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
1019 ; (LAMBDA (Y) (FOO Y))
1021 ; caught STYLE-WARNING:
1022 ; The variable Y is defined but never used.
1024 245: bugs in disassembler
1025 a. On X86 an immediate operand for IMUL is printed incorrectly.
1026 b. On X86 operand size prefix is not recognized.
1028 248: "reporting errors in type specifier syntax"
1029 (TYPEP 1 '(SYMBOL NIL)) says something about "unknown type
1033 (defun foo (&key (a :x))
1034 (declare (fixnum a))
1037 does not cause a warning. (BTW: old SBCL issued a warning, but for a
1038 function, which was never called!)
1041 Compiler does not emit warnings for
1043 a. (lambda () (svref (make-array 8 :adjustable t) 1))
1046 (list (let ((y (the real x)))
1047 (unless (floatp y) (error ""))
1049 (integer-length x)))
1052 (declare (optimize (debug 0)))
1053 (declare (type vector x))
1054 (list (fill-pointer x)
1058 Complex array type does not have corresponding type specifier.
1060 DEFUNCT CATEGORIES OF BUGS
1062 These labels were used for bugs related to the old IR1 interpreter.
1063 The # values reached 6 before the category was closed down.