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 (I *think* this is a bug. It certainly seems like strange behavior. But
122 the ANSI spec is scary, dark, and deep.. -- WHN)
123 (FORMAT NIL "~,1G" 1.4) => "1. "
124 (FORMAT NIL "~3,1G" 1.4) => "1. "
127 Sometimes (SB-EXT:QUIT) fails with
128 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
129 Process inferior-lisp exited abnormally with code 1
130 I haven't noticed a repeatable case of this yet.
133 The printer doesn't report closures very well. This is true in
137 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
138 It would be nice to make closures have a settable name slot,
139 and make things like DEFSTRUCT and FLET, which create closures,
140 set helpful values into this slot.
143 And as long as we're wishing, it would be awfully nice if INSPECT could
144 also report on closures, telling about the values of the bound variables.
147 The compiler assumes that any time a function of declared FTYPE
148 doesn't signal an error, its arguments were of the declared type.
149 E.g. compiling and loading
150 (DECLAIM (OPTIMIZE (SAFETY 3)))
151 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
153 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
155 (COND ((> (FACTORIAL X) 1.0E6)
156 (FORMAT T "too big~%"))
158 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
160 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
163 will cause the INTEGERP case to be selected, giving bogus output a la
165 This violates the "declarations are assertions" principle.
166 According to the ANSI spec, in the section "System Class FUNCTION",
167 this is a case of "lying to the compiler", but the lying is done
168 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
169 not by the unexpectedly general definition of FACTORIAL. In any case,
170 "declarations are assertions" means that lying to the compiler should
171 cause an error to be signalled, and should not cause a bogus
172 result to be returned. Thus, the compiler should not assume
173 that arbitrary functions check their argument types. (It might
174 make sense to add another flag (CHECKED?) to DEFKNOWN to
175 identify functions which *do* check their argument types.)
176 (Also, verify that the compiler handles declared function
177 return types as assertions.)
180 The definitions of SIGCONTEXT-FLOAT-REGISTER and
181 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
182 supported on FreeBSD because the floating point state is not saved,
183 but at least as of FreeBSD 4.0, the floating point state *is* saved,
184 so they could be supported after all. Very likely
185 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
188 a slew of floating-point-related errors reported by Peter Van Eynde
190 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT on the x86 is
191 bogus, and should probably be 1.4012985e-45. In SBCL,
192 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
193 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
194 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
195 and LEAST-NEGATIVE-LONG-FLOAT.
196 c: Many expressions generate floating infinity on x86/Linux:
201 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
202 on x86/Linux generates the infinities instead. That might or
203 might not be conforming behavior, but it's also inconsistent,
204 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
205 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
206 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
208 d: (in section12.erg) various forms a la
209 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
210 don't give the right behavior.
213 type safety errors reported by Peter Van Eynde July 25, 2000:
214 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
215 not a binary input stream, but instead cheerfully reads from
216 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
219 The debugger LIST-LOCATIONS command doesn't work properly.
220 (How should it work properly?)
223 Compiling and loading
224 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
226 then requesting a BACKTRACE at the debugger prompt gives no information
227 about where in the user program the problem occurred.
230 Using the pretty-printer from the command prompt gives funny
231 results, apparently because the pretty-printer doesn't know
232 about user's command input, including the user's carriage return
233 that the user, and therefore the pretty-printer thinks that
234 the new output block should start indented 2 or more characters
235 rightward of the correct location.
238 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
239 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
240 crashes SBCL. In general tracing anything which is used in the
241 implementation of TRACE is likely to have the same problem.
244 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
245 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
246 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
247 WITH-OUTPUT-TO-STRING.
250 ANSI says in one place that type declarations can be abbreviated even
251 when the type name is not a symbol, e.g.
252 (DECLAIM ((VECTOR T) *FOOVECTOR*))
253 SBCL doesn't support this. But ANSI says in another place that this
254 isn't allowed. So it's not clear this is a bug after all. (See the
255 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
259 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
260 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
261 an easily guessable temporary filename in a way which might open
262 applications using LOAD-FOREIGN to hijacking by malicious users
263 on the same machine. Incantations for doing this safely are
264 floating around the net in various "how to write secure programs
265 despite Unix" documents, and it would be good to (1) fix this in
266 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
267 files and make it share the same new safe logic.
269 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
270 make the temporary filename less easily guessable)
273 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
274 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
275 the mean of the distribution can be systematically O(0.1%) wrong.
276 Just increasing R-I-E-B is probably not a good solution, since
277 it would decrease efficiency more than is probably necessary. Perhaps
278 using some sort of accept/reject method would be better.
281 Internally the compiler sometimes evaluates
282 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
283 (I stumbled across this when I added an
284 (assert (not (eq type1 *wild-type*)))
285 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
286 in a type context should probably be translated to T, and so it's
287 probably wrong to ask whether it's equal to the T type and then (using
288 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
289 (I haven't tried to investigate this bug enough to guess whether
290 there might be any user-level symptoms.)
292 In fact, the type system is likely to depend on this inequality not
293 holding... * is not equivalent to T in many cases, such as
294 (VECTOR *) /= (VECTOR T).
297 Inconsistencies between derived and declared VALUES return types for
298 DEFUN aren't checked very well. E.g. the logic which successfully
299 catches problems like
300 (declaim (ftype (function (fixnum) float) foo))
302 (declare (type integer x))
303 (values x)) ; wrong return type, detected, gives warning, good!
305 (declaim (ftype (function (t) (values t t)) bar))
307 (values x)) ; wrong number of return values, no warning, bad!
308 The cause of this is seems to be that (1) the internal function
309 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
310 arguments symmetrically, and (2) when the type checking code was
311 written back when when SBCL's code was still CMU CL, the intent
313 (declaim (ftype (function (t) t) bar))
315 (values x x)) ; wrong number of return values; should give warning?
316 not be warned for, because a two-valued return value is considered
317 to be compatible with callers who expects a single value to be
318 returned. That intent is probably not appropriate for modern ANSI
319 Common Lisp, but fixing this might be complicated because of other
320 divergences between auld-style and new-style handling of
321 multiple-VALUES types. (Some issues related to this were discussed
322 on cmucl-imp at some length sometime in 2000.)
325 The facility for dumping a running Lisp image to disk gets confused
326 when run without the PURIFY option, and creates an unnecessarily large
327 core file (apparently representing memory usage up to the previous
328 high-water mark). Moreover, when the file is loaded, it confuses the
329 GC, so that thereafter memory usage can never be reduced below that
333 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
334 CL), out-of-line structure slot setters are horribly inefficient
335 whenever the type of the slot is declared, because out-of-line
336 structure slot setters are implemented as closures to save space,
337 so the compiler doesn't compile the type test into code, but
338 instead just saves the type in a lexical closure and interprets it
340 To exercise the problem, compile and load
341 (cl:in-package :cl-user)
343 (bar (error "missing") :type bar))
346 (loop (setf (foo-bar *foo*) x)))
348 (defvar *bar* (make-bar))
349 (defvar *foo* (make-foo :bar *bar*))
350 (defvar *setf-foo-bar* #'(setf foo-bar))
352 (loop (funcall *setf-foo-bar* x *foo*)))
353 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
354 use BACKTRACE, to see it's spending all essentially all its time
355 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
356 One possible solution would be simply to give up on
357 representing structure slot accessors as functions, and represent
358 them as macroexpansions instead. This can be inconvenient for users,
359 but it's not clear that it's worse than trying to help by expanding
360 into a horribly inefficient implementation.
361 As a workaround for the problem, #'(SETF FOO) expressions
362 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
363 (defmacro efficient-setf-function (place-function-name)
364 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
365 ;; a workaround for the problem, encouraging the
366 ;; inline expansion of the structure accessor, so
367 ;; that the compiler can optimize its type test
368 (let ((new-value (gensym "NEW-VALUE-"))
369 (structure-value (gensym "STRUCTURE-VALUE-")))
370 `(lambda (,new-value ,structure-value)
371 (setf (,place-function-name ,structure-value)
373 ;; no problem, can just use the ordinary expansion
374 `(function (setf ,place-function-name))))
377 There's apparently a bug in CEILING optimization which caused
378 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
379 applied the patches to SBCL and they didn't seem to cause problems
380 (as reported sbcl-devel 2001-05-04). However, since the patches
381 modify nontrivial code which was apparently written incorrectly
382 the first time around, until regression tests are written I'm not
383 comfortable merging the patches in the CVS version of SBCL.
386 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
387 a clean, just-started SBCL system. And it seems to be right:
388 (ROOM T) can bring a small computer to its knees for a *long*
389 time trying to GC afterwards. Surely there's some more economical
390 way to implement (ROOM T).
393 When the compiler inline expands functions, it may be that different
394 kinds of return values are generated from different code branches.
395 E.g. an inline expansion of POSITION generates integer results
396 from one branch, and NIL results from another. When that inline
397 expansion is used in a context where only one of those results
400 (aref *a1* (position x *a2*)))
401 and the compiler can't prove that the unacceptable branch is
402 never taken, then bogus type mismatch warnings can be generated.
403 If you need to suppress the type mismatch warnings, you can
404 suppress the inline expansion,
406 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
407 (aref *a1* (position x *a2*)))
408 or, sometimes, suppress them by declaring the result to be of an
411 (aref *a1* (the integer (position x *a2*))))
413 This is not a new compiler problem in 0.7.0, but the new compiler
414 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
415 more conspicuous. If you don't need performance from these functions,
416 and the bogus warnings are a nuisance for you, you can return to
417 your pre-0.7.0 state of grace with
418 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
421 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
422 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
423 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
424 when of course it should be NIL. (He says it only fails for X86,
425 not SPARC; dunno about Alpha.)
427 Also, "the same problem exists for LONG-FLOAT-EPSILON,
428 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
429 for the -negative- the + is replaced by a - in the test)."
431 Raymond Toy comments that this is tricky on the X86 since its FPU
432 uses 80-bit precision internally.
435 Even in sbcl-0.pre7.x, which is supposed to be free of the old
436 non-ANSI behavior of treating the function return type inferred
437 from the current function definition as a declaration of the
438 return type from any function of that name, the return type of NIL
439 is attached to FOO in 120a above, and used to optimize code which
443 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
444 the entire lexical environment at the point of MACROLET available
445 in the bodies of the macroexpander functions. In particular, it
446 allows the function bodies (which run at compile time) to try to
447 access lexical variables (which are only defined at runtime).
448 It doesn't even issue a warning, which is bad.
450 The SBCL behavior arguably conforms to the ANSI spec (since the
451 spec says that the behavior is undefined, ergo anything conforms).
452 However, it would be better to issue a compile-time error.
453 Unfortunately I (WHN) don't see any simple way to detect this
454 condition in order to issue such an error, so for the meantime
455 SBCL just does this weird broken "conforming" thing.
457 The ANSI standard says, in the definition of the special operator
459 The macro-expansion functions defined by MACROLET are defined
460 in the lexical environment in which the MACROLET form appears.
461 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
462 the local macro definitions in a MACROLET, but the consequences
463 are undefined if the local macro definitions reference any
464 local variable or function bindings that are visible in that
466 Then it seems to contradict itself by giving the example
468 (macrolet ((fudge (z)
469 ;The parameters x and flag are not accessible
470 ; at this point; a reference to flag would be to
471 ; the global variable of that name.
472 ` (if flag (* ,z ,z) ,z)))
473 ;The parameters x and flag are accessible here.
477 The comment "a reference to flag would be to the global variable
478 of the same name" sounds like good behavior for the system to have.
479 but actual specification quoted above says that the actual behavior
482 (Since 0.7.8.23 macroexpanders are defined in a restricted version
483 of the lexical environment, containing no lexical variables and
484 functions, which seems to conform to ANSI and CLtL2, but signalling
485 a STYLE-WARNING for references to variables similar to locals might
489 (as reported by Gabe Garza on cmucl-help 2001-09-21)
491 (defun test-pred (x y)
495 (func (lambda () x)))
496 (print (eq func func))
497 (print (test-pred func func))
498 (delete func (list func))))
499 Now calling (TEST-CASE) gives output
502 (#<FUNCTION {500A9EF9}>)
503 Evidently Python thinks of the lambda as a code transformation so
504 much that it forgets that it's also an object.
507 Ideally, uninterning a symbol would allow it, and its associated
508 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
509 at least as of sbcl-0.7.0, this isn't the case. Information about
510 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
511 essentially in ordinary (non-weak) hash tables keyed by symbols.
512 Thus, once a system has an entry in this system, it tends to live
513 forever, even when it is uninterned and all other references to it
516 141: "pretty printing and backquote"
519 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
522 * (write '`(, .ala.) :readably t :pretty t)
525 (note the space between the comma and the point)
528 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
529 prominent SourceForge web/db bug tracking system, which is
530 unfortunately not a reliable way to get a timely response from
531 the SBCL maintainers)
532 In the course of trying to build a test case for an
533 application error, I encountered this behavior:
534 If you start up sbcl, and then lay on CTRL-C for a
535 minute or two, the lisp process will eventually say:
536 %PRIMITIVE HALT called; the party is over.
537 and throw you into the monitor. If I start up lisp,
538 attach to the process with strace, and then do the same
539 (abusive) thing, I get instead:
540 access failure in heap page not marked as write-protected
541 and the monitor again. I don't know enough to have the
542 faintest idea of what is going on here.
543 This is with sbcl 6.12, uname -a reports:
544 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
545 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
546 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
547 it took more than two minutes (but less than five) for me.
550 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
551 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
552 COMPLEX implementation didn't deal with this, and hasn't been
553 upgraded to do so. (This doesn't seem to be a high priority
554 conformance problem, since seems hard to construct useful code
558 Floating point errors are reported poorly. E.g. on x86 OpenBSD
561 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
562 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
563 No traps are enabled? How can this be?
564 It should be possible to be much more specific (overflow, division
565 by zero, etc.) and of course the "How can this be?" should be fixable.
567 See also bugs #45.c and #183
570 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
571 (in-package :cl-user)
574 (defstruct foo bar bletch)
576 (labels ((kidify1 (kid)
584 (declare (inline kid-frob))
587 (the simple-vector (foo-bar perd)))))
589 debugger invoked on condition of type TYPE-ERROR:
590 The value NIL is not of type SB-C::NODE.
591 The location of this failure has moved around as various related
592 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
593 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
595 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
596 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
597 of KIDIFY1, which already does not exist. So it gives up on
598 expansion, leaving garbage consisting of infinished blocks of the
599 partially converted function.)
601 (due to reordering of the compiler this example is compiled
602 successfully by 0.7.14, but the bug probably remains)
605 (reported by Robert E. Brown 2002-04-16)
606 When a function is called with too few arguments, causing the
607 debugger to be entered, the uninitialized slots in the bad call frame
608 seem to cause GCish problems, being interpreted as tagged data even
609 though they're not. In particular, executing ROOM in the
610 debugger at that point causes AVER failures:
613 * (lisp-implementation-version)
619 failed AVER: "(SAP= CURRENT END)"
620 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
621 isn't too surprising since there are many differences in stack
622 implementation and GC conservatism between the X86 and other ports.)
625 In sbcl-0.7.3.11, compiling the (illegal) code
626 (in-package :cl-user)
627 (defmethod prove ((uustk uustk))
630 gives the (not terribly clear) error message
632 ; (during macroexpansion of (DEFMETHOD PROVE ...))
633 ; can't get template for (FROB NIL NIL)
634 The problem seems to be that the code walker used by the DEFMETHOD
635 macro is unhappy with the illegal syntax in the method body, and
636 is giving an unclear error message.
639 The compiler sometimes tries to constant-fold expressions before
640 it checks to see whether they can be reached. This can lead to
641 bogus warnings about errors in the constant folding, e.g. in code
644 (WRITE-STRING (> X 0) "+" "0"))
645 compiled in a context where the compiler can prove that X is NIL,
646 and the compiler complains that (> X 0) causes a type error because
647 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
648 caused a full WARNING, which made the bug really annoying because then
649 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
650 code. Since then the warning has been downgraded to STYLE-WARNING,
651 so it's still a bug but at least it's a little less annoying.
653 183: "IEEE floating point issues"
654 Even where floating point handling is being dealt with relatively
655 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
656 accrued-exceptions and current-exceptions part of the fp control
657 word don't seem to bear much relation to reality. E.g. on
661 debugger invoked on condition of type DIVISION-BY-ZERO:
662 arithmetic error DIVISION-BY-ZERO signalled
663 0] (sb-vm::get-floating-point-modes)
665 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
666 :ROUNDING-MODE :NEAREST
667 :CURRENT-EXCEPTIONS NIL
668 :ACCRUED-EXCEPTIONS (:INEXACT)
671 * (sb-vm::get-floating-point-modes)
672 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
673 :ROUNDING-MODE :NEAREST
674 :CURRENT-EXCEPTIONS (:INEXACT)
675 :ACCRUED-EXCEPTIONS (:INEXACT)
678 188: "compiler performance fiasco involving type inference and UNION-TYPE"
682 (declare (optimize (safety 3)))
683 (declare (optimize (compilation-speed 2)))
684 (declare (optimize (speed 1) (debug 1) (space 1)))
686 (declare (type (integer 0) start))
687 (print (incf start 22))
688 (print (incf start 26))
689 (print (incf start 28)))
691 (declare (type (integer 0) start))
692 (print (incf start 22))
693 (print (incf start 26)))
695 (declare (type (integer 0) start))
696 (print (incf start 22))
697 (print (incf start 26))))))
699 This example could be solved with clever enough constraint
700 propagation or with SSA, but consider
703 (loop (if (random-boolean)
707 The careful type of X is {2k+5n} :-(. Is it really important to be
708 able to work with unions of many intervals?
710 190: "PPC/Linux pipe? buffer? bug"
711 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
712 on the PPC/Linux platform, waiting for a zombie env process. This
713 is a classic symptom of buffer filling and deadlock, but it seems
714 only sporadically reproducible.
716 191: "Miscellaneous PCL deficiencies"
717 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
718 a. DEFCLASS does not inform the compiler about generated
719 functions. Compiling a file with
723 (WITH-SLOTS (A-CLASS-X) A
725 results in a STYLE-WARNING:
727 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
729 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
730 Mai points out that the declamation of functions is in fact
731 incorrect in some cases (most notably for structure
732 classes). This means that at present erroneous attempts to use
733 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
734 won't get the corresponding STYLE-WARNING.
735 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
736 lists and &KEY arguments) do not signal errors when they should.
738 201: "Incautious type inference from compound types"
739 a. (reported by APD sbcl-devel 2002-09-17)
741 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
743 (FORMAT NIL "~S IS ~S, Y = ~S"
750 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
754 (declare (type (array * (4 4)) x))
756 (setq x (make-array '(4 4)))
757 (adjust-array y '(3 5))
758 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
760 * (foo (make-array '(4 4) :adjustable t))
763 205: "environment issues in cross compiler"
764 (These bugs have no impact on user code, but should be fixed or
766 a. Macroexpanders introduced with MACROLET are defined in the null
768 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
769 the null lexical environment.
770 c. The cross-compiler cannot inline functions defined in a non-null
773 206: ":SB-FLUID feature broken"
774 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
775 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
778 207: "poorly distributed SXHASH results for compound data"
779 SBCL's SXHASH could probably try a little harder. ANSI: "the
780 intent is that an implementation should make a good-faith
781 effort to produce hash-codes that are well distributed
782 within the range of non-negative fixnums". But
783 (let ((hits (make-hash-table)))
786 (let* ((ij (cons i j))
787 (newlist (push ij (gethash (sxhash ij) hits))))
789 (format t "~&collision: ~S~%" newlist))))))
790 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
791 would be an obvious way of fix. Maybe it would be acceptably efficient
792 to redo MIX using a lookup into a 256-entry s-box containing
793 29-bit pseudorandom numbers?
795 211: "keywords processing"
796 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
797 number of keyword arguments.
800 (flet ((foo (&key y) (list y)))
801 (list (foo :y 1 :y 2)))
803 issues confusing message
808 ; caught STYLE-WARNING:
809 ; The variable #:G15 is defined but never used.
811 212: "Sequence functions and circular arguments"
812 COERCE, MERGE and CONCATENATE go into an infinite loop when given
813 circular arguments; it would be good for the user if they could be
814 given an error instead (ANSI 17.1.1 allows this behaviour on the part
815 of the implementation, as conforming code cannot give non-proper
816 sequences to these functions. MAP also has this problem (and
817 solution), though arguably the convenience of being able to do
818 (MAP 'LIST '+ FOO '#1=(1 . #1#))
819 might be classed as more important (though signalling an error when
820 all of the arguments are circular is probably desireable).
822 213: "Sequence functions and type checking"
823 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
824 various complicated, though recognizeable, CONS types [e.g.
825 (CONS * (CONS * NULL))
826 which according to ANSI should be recognized] (and, in SAFETY 3
827 code, should return a list of LENGTH 2 or signal an error)
828 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
829 check that it will return a sequence of the given type. Fixing
830 it along the same lines as the others (cf. work done around
831 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
832 entirely straightforward.
833 c. All of these functions will silently accept a type of the form
835 whether or not the return value is of this type. This is
836 probably permitted by ANSI (see "Exceptional Situations" under
837 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
838 know about this escape clause, so code of the form
839 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
840 can erroneously return T.
843 SBCL 0.6.12.43 fails to compile
846 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
847 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
850 or a more simple example:
853 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
854 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
856 215: ":TEST-NOT handling by functions"
857 a. FIND and POSITION currently signal errors when given non-NIL for
858 both their :TEST and (deprecated) :TEST-NOT arguments, but by
859 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
860 means that the effect is "unpredictable but harmless". It's not
861 clear what that actually means; it may preclude conforming
862 implementations from signalling errors.
863 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
864 when conflict occurs. As a quality of implementation issue, it
865 might be preferable to treat :TEST and :TEST-NOT as being in some
866 sense the same &KEY, and effectively take the first test function in
868 c. Again, a quality of implementation issue: it would be good to issue a
869 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
870 WARNING for calls with both :TEST and :TEST-NOT; possibly this
871 latter should be WARNed about at execute-time too.
873 216: "debugger confused by frames with invalid number of arguments"
874 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
875 leaves the system confused, enough so that (QUIT) no longer works.
876 It's as though the process of working with the uninitialized slot in
877 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
878 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
880 217: "Bad type operations with FUNCTION types"
883 * (values-type-union (specifier-type '(function (base-char)))
884 (specifier-type '(function (integer))))
886 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
888 It causes insertion of wrong type assertions into generated
892 (let ((f (etypecase x
893 (character #'write-char)
894 (integer #'write-byte))))
897 (character (write-char x s))
898 (integer (write-byte x s)))))
900 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
902 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
903 produce invalid code, but type checking is not accurate. Similar
904 problems exist with VALUES-TYPE-INTERSECTION.)
907 Sbcl 0.7.9 fails to compile
909 (multiple-value-call #'list
910 (the integer (helper))
913 Type check for INTEGER, the result of which serves as the first
914 argument of M-V-C, is inserted after evaluation of NIL. So arguments
915 of M-V-C are pushed in the wrong order. As a temporary workaround
916 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
917 would be to put the check between evaluation of arguments, but it
918 could be tricky to check result types of PROG1, IF etc.
920 233: bugs in constraint propagation
923 (declare (optimize (speed 2) (safety 3)))
929 (quux y (+ y 2d0) (* y 3d0)))))
930 (foo 4) => segmentation violation
932 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
936 (declaim (optimize (speed 2) (safety 3)))
938 (if (typep (prog1 x (setq x y)) 'double-float)
941 (foo 1d0 5) => segmentation violation
943 235: "type system and inline expansion"
945 (declaim (ftype (function (cons) number) acc))
946 (declaim (inline acc))
948 (the number (car c)))
951 (values (locally (declare (optimize (safety 0)))
953 (locally (declare (optimize (safety 3)))
956 (foo '(nil) '(t)) => NIL, T.
958 237: "Environment arguments to type functions"
959 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
960 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
961 argument, but they ignore it completely. This is almost
962 certainly not correct.
963 b. Also, the compiler's optimizers for TYPEP have not been informed
964 about the new argument; consequently, they will not transform
965 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
966 just as optimizeable as (TYPEP 1 'INTEGER).
968 238: "REPL compiler overenthusiasm for CLOS code"
970 * (defclass foo () ())
971 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
972 causes approximately 100 lines of code deletion notes. Some
973 discussion on this issue happened under the title 'Three "interesting"
974 bugs in PCL', resulting in a fix for this oververbosity from the
975 compiler proper; however, the problem persists in the interactor
976 because the notion of original source is not preserved: for the
977 compiler, the original source of the above expression is (DEFMETHOD
978 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
979 compiler gets its hands on the code needing compilation from the REPL,
980 it has been macroexpanded several times.
982 A symptom of the same underlying problem, reported by Tony Martinez:
984 (with-input-from-string (*query-io* " no")
986 (simple-type-error () 'error))
988 ; (SB-KERNEL:FLOAT-WAIT)
990 ; note: deleting unreachable code
991 ; compilation unit finished
994 241: "DEFCLASS mysteriously remembers uninterned accessor names."
995 (from tonyms on #lisp IRC 2003-02-25)
996 In sbcl-0.7.12.55, typing
997 (defclass foo () ((bar :accessor foo-bar)))
1000 (defclass foo () ((bar :accessor foo-bar)))
1001 gives the error message
1002 "#:FOO-BAR already names an ordinary function or a macro."
1003 So it's somehow checking the uninterned old accessor name instead
1004 of the new requested accessor name, which seems broken to me (WHN).
1006 242: "WRITE-SEQUENCE suboptimality"
1007 (observed from clx performance)
1008 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
1009 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
1010 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
1011 rather than writing the sequence in one go, leading to severe
1012 performance degradation.
1014 243: "STYLE-WARNING overenthusiasm for unused variables"
1015 (observed from clx compilation)
1016 In sbcl-0.7.14, in the presence of the macros
1017 (DEFMACRO FOO (X) `(BAR ,X))
1018 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
1019 somewhat surprising style warnings are emitted for
1020 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
1022 ; (LAMBDA (Y) (FOO Y))
1024 ; caught STYLE-WARNING:
1025 ; The variable Y is defined but never used.
1027 245: bugs in disassembler
1028 a. On X86 an immediate operand for IMUL is printed incorrectly.
1029 b. On X86 operand size prefix is not recognized.
1031 248: "reporting errors in type specifier syntax"
1032 (TYPEP 1 '(SYMBOL NIL)) says something about "unknown type
1036 (defun foo (&key (a :x))
1037 (declare (fixnum a))
1040 does not cause a warning. (BTW: old SBCL issued a warning, but for a
1041 function, which was never called!)
1044 Compiler does not emit warnings for
1046 a. (lambda () (svref (make-array 8 :adjustable t) 1))
1049 (list (let ((y (the real x)))
1050 (unless (floatp y) (error ""))
1052 (integer-length x)))
1055 (declare (optimize (debug 0)))
1056 (declare (type vector x))
1057 (list (fill-pointer x)
1061 Complex array type does not have corresponding type specifier.
1063 DEFUNCT CATEGORIES OF BUGS
1065 These labels were used for bugs related to the old IR1 interpreter.
1066 The # values reached 6 before the category was closed down.