3 Bugs can be reported on the help mailing list
4 sbcl-help@lists.sourceforge.net
5 or on the development mailing list
6 sbcl-devel@lists.sourceforge.net
8 Please include enough information in a bug report that someone reading
9 it can reproduce the problem, i.e. don't write
10 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
11 PRINT-OBJECT doesn't seem to work with *PRINT-LENGTH*. Is this a bug?
13 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
14 In sbcl-1.2.3 running under OpenBSD 4.5 on my Alpha box, when
15 I compile and load the file
16 (DEFSTRUCT (FOO (:PRINT-OBJECT (LAMBDA (X Y)
17 (LET ((*PRINT-LENGTH* 4))
20 then at the command line type
22 the program loops endlessly instead of printing the object.
27 There is also some information on bugs in the manual page and
28 in the TODO file. Eventually more such information may move here.
30 The gaps in the number sequence belong to old bug descriptions which
31 have gone away (typically because they were fixed, but sometimes for
32 other reasons, e.g. because they were moved elsewhere).
36 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
37 instead of just punting when a contradictory structure definition
38 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
39 changes their layout, you probably have to rebuild your entire
40 program, even if you know or guess enough about the internals of
41 SBCL to wager that this (undefined in ANSI) operation would be safe.
43 3: "type checking of structure slots"
45 ANSI specifies that a type mismatch in a structure slot
46 initialization value should not cause a warning.
48 This one might not be fixed for a while because while we're big
49 believers in ANSI compatibility and all, (1) there's no obvious
50 simple way to do it (short of disabling all warnings for type
51 mismatches everywhere), and (2) there's a good portable
52 workaround, and (3) by their own reasoning, it looks as though
53 ANSI may have gotten it wrong. ANSI justifies this specification
55 The restriction against issuing a warning for type mismatches
56 between a slot-initform and the corresponding slot's :TYPE
57 option is necessary because a slot-initform must be specified
58 in order to specify slot options; in some cases, no suitable
60 However, in SBCL (as in CMU CL or, for that matter, any compiler
61 which really understands Common Lisp types) a suitable default
62 does exist, in all cases, because the compiler understands the
63 concept of functions which never return (i.e. has return type NIL).
64 Thus, as a portable workaround, you can use a call to some
65 known-never-to-return function as the default. E.g.
67 (BAR (ERROR "missing :BAR argument")
68 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
70 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
71 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
72 (ERROR "missing required argument"))
74 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
75 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
76 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
77 Such code should compile without complaint and work correctly either
78 on SBCL or on any other completely compliant Common Lisp system.
80 b: &AUX argument in a boa-constructor without a default value means
81 "do not initilize this slot" and does not cause type error. But
82 an error may be signalled at read time and it would be good if
88 The "compiling top-level form:" output ought to be condensed.
89 Perhaps any number of such consecutive lines ought to turn into a
90 single "compiling top-level forms:" line.
93 It would be nice if the
95 (during macroexpansion)
96 said what macroexpansion was at fault, e.g.
98 (during macroexpansion of IN-PACKAGE,
99 during macroexpansion of DEFFOO)
102 (I *think* this is a bug. It certainly seems like strange behavior. But
103 the ANSI spec is scary, dark, and deep.. -- WHN)
104 (FORMAT NIL "~,1G" 1.4) => "1. "
105 (FORMAT NIL "~3,1G" 1.4) => "1. "
108 Sometimes (SB-EXT:QUIT) fails with
109 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
110 Process inferior-lisp exited abnormally with code 1
111 I haven't noticed a repeatable case of this yet.
114 The printer doesn't report closures very well. This is true in
118 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
119 It would be nice to make closures have a settable name slot,
120 and make things like DEFSTRUCT and FLET, which create closures,
121 set helpful values into this slot.
124 And as long as we're wishing, it would be awfully nice if INSPECT could
125 also report on closures, telling about the values of the bound variables.
128 The compiler assumes that any time a function of declared FTYPE
129 doesn't signal an error, its arguments were of the declared type.
130 E.g. compiling and loading
131 (DECLAIM (OPTIMIZE (SAFETY 3)))
132 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
134 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
136 (COND ((> (FACTORIAL X) 1.0E6)
137 (FORMAT T "too big~%"))
139 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
141 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
144 will cause the INTEGERP case to be selected, giving bogus output a la
146 This violates the "declarations are assertions" principle.
147 According to the ANSI spec, in the section "System Class FUNCTION",
148 this is a case of "lying to the compiler", but the lying is done
149 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
150 not by the unexpectedly general definition of FACTORIAL. In any case,
151 "declarations are assertions" means that lying to the compiler should
152 cause an error to be signalled, and should not cause a bogus
153 result to be returned. Thus, the compiler should not assume
154 that arbitrary functions check their argument types. (It might
155 make sense to add another flag (CHECKED?) to DEFKNOWN to
156 identify functions which *do* check their argument types.)
157 (Also, verify that the compiler handles declared function
158 return types as assertions.)
161 The definitions of SIGCONTEXT-FLOAT-REGISTER and
162 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
163 supported on FreeBSD because the floating point state is not saved,
164 but at least as of FreeBSD 4.0, the floating point state *is* saved,
165 so they could be supported after all. Very likely
166 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
169 a slew of floating-point-related errors reported by Peter Van Eynde
171 c: Many expressions generate floating infinity on x86/Linux:
176 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
177 on x86/Linux generates the infinities instead. That might or
178 might not be conforming behavior, but it's also inconsistent,
179 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
180 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
181 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
183 d: (in section12.erg) various forms a la
184 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
185 don't give the right behavior.
188 The debugger LIST-LOCATIONS command doesn't work properly.
189 (How should it work properly?)
192 Compiling and loading
193 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
195 then requesting a BACKTRACE at the debugger prompt gives no information
196 about where in the user program the problem occurred.
199 Using the pretty-printer from the command prompt gives funny
200 results, apparently because the pretty-printer doesn't know
201 about user's command input, including the user's carriage return
202 that the user, and therefore the pretty-printer thinks that
203 the new output block should start indented 2 or more characters
204 rightward of the correct location.
207 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
208 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
209 crashes SBCL. In general tracing anything which is used in the
210 implementation of TRACE is likely to have the same problem.
213 ANSI says in one place that type declarations can be abbreviated even
214 when the type name is not a symbol, e.g.
215 (DECLAIM ((VECTOR T) *FOOVECTOR*))
216 SBCL doesn't support this. But ANSI says in another place that this
217 isn't allowed. So it's not clear this is a bug after all. (See the
218 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
222 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
223 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
224 an easily guessable temporary filename in a way which might open
225 applications using LOAD-FOREIGN to hijacking by malicious users
226 on the same machine. Incantations for doing this safely are
227 floating around the net in various "how to write secure programs
228 despite Unix" documents, and it would be good to (1) fix this in
229 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
230 files and make it share the same new safe logic.
232 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
233 make the temporary filename less easily guessable)
236 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
237 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
238 the mean of the distribution can be systematically O(0.1%) wrong.
239 Just increasing R-I-E-B is probably not a good solution, since
240 it would decrease efficiency more than is probably necessary. Perhaps
241 using some sort of accept/reject method would be better.
244 Internally the compiler sometimes evaluates
245 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
246 (I stumbled across this when I added an
247 (assert (not (eq type1 *wild-type*)))
248 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
249 in a type context should probably be translated to T, and so it's
250 probably wrong to ask whether it's equal to the T type and then (using
251 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
252 (I haven't tried to investigate this bug enough to guess whether
253 there might be any user-level symptoms.)
255 In fact, the type system is likely to depend on this inequality not
256 holding... * is not equivalent to T in many cases, such as
257 (VECTOR *) /= (VECTOR T).
260 The facility for dumping a running Lisp image to disk gets confused
261 when run without the PURIFY option, and creates an unnecessarily large
262 core file (apparently representing memory usage up to the previous
263 high-water mark). Moreover, when the file is loaded, it confuses the
264 GC, so that thereafter memory usage can never be reduced below that
267 (As of 0.8.7.3 it's likely that the latter half of this bug is fixed.
268 The interaction between gencgc and the variables used by
269 save-lisp-and-die is still nonoptimal, though, so no respite from
273 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
274 CL), out-of-line structure slot setters are horribly inefficient
275 whenever the type of the slot is declared, because out-of-line
276 structure slot setters are implemented as closures to save space,
277 so the compiler doesn't compile the type test into code, but
278 instead just saves the type in a lexical closure and interprets it
280 To exercise the problem, compile and load
281 (cl:in-package :cl-user)
283 (bar (error "missing") :type bar))
286 (loop (setf (foo-bar *foo*) x)))
288 (defvar *bar* (make-bar))
289 (defvar *foo* (make-foo :bar *bar*))
290 (defvar *setf-foo-bar* #'(setf foo-bar))
292 (loop (funcall *setf-foo-bar* x *foo*)))
293 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
294 use BACKTRACE, to see it's spending all essentially all its time
295 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
296 One possible solution would be simply to give up on
297 representing structure slot accessors as functions, and represent
298 them as macroexpansions instead. This can be inconvenient for users,
299 but it's not clear that it's worse than trying to help by expanding
300 into a horribly inefficient implementation.
301 As a workaround for the problem, #'(SETF FOO) expressions
302 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
303 (defmacro efficient-setf-function (place-function-name)
304 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
305 ;; a workaround for the problem, encouraging the
306 ;; inline expansion of the structure accessor, so
307 ;; that the compiler can optimize its type test
308 (let ((new-value (gensym "NEW-VALUE-"))
309 (structure-value (gensym "STRUCTURE-VALUE-")))
310 `(lambda (,new-value ,structure-value)
311 (setf (,place-function-name ,structure-value)
313 ;; no problem, can just use the ordinary expansion
314 `(function (setf ,place-function-name))))
317 There's apparently a bug in CEILING optimization which caused
318 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
319 applied the patches to SBCL and they didn't seem to cause problems
320 (as reported sbcl-devel 2001-05-04). However, since the patches
321 modify nontrivial code which was apparently written incorrectly
322 the first time around, until regression tests are written I'm not
323 comfortable merging the patches in the CVS version of SBCL.
326 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
327 a clean, just-started SBCL system. And it seems to be right:
328 (ROOM T) can bring a small computer to its knees for a *long*
329 time trying to GC afterwards. Surely there's some more economical
330 way to implement (ROOM T).
332 Daniel Barlow doesn't know what fixed this, but observes that it
333 doesn't seem to be the case in 0.8.7.3 any more. Instead, (ROOM T)
334 in a fresh SBCL causes
336 debugger invoked on a SB-INT:BUG in thread 5911:
337 failed AVER: "(SAP= CURRENT END)"
339 unless a GC has happened beforehand.
342 When the compiler inline expands functions, it may be that different
343 kinds of return values are generated from different code branches.
344 E.g. an inline expansion of POSITION generates integer results
345 from one branch, and NIL results from another. When that inline
346 expansion is used in a context where only one of those results
349 (aref *a1* (position x *a2*)))
350 and the compiler can't prove that the unacceptable branch is
351 never taken, then bogus type mismatch warnings can be generated.
352 If you need to suppress the type mismatch warnings, you can
353 suppress the inline expansion,
355 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
356 (aref *a1* (position x *a2*)))
357 or, sometimes, suppress them by declaring the result to be of an
360 (aref *a1* (the integer (position x *a2*))))
362 This is not a new compiler problem in 0.7.0, but the new compiler
363 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
364 more conspicuous. If you don't need performance from these functions,
365 and the bogus warnings are a nuisance for you, you can return to
366 your pre-0.7.0 state of grace with
367 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
372 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
373 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
374 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
375 when of course it should be NIL. (He says it only fails for X86,
376 not SPARC; dunno about Alpha.)
378 Also, "the same problem exists for LONG-FLOAT-EPSILON,
379 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
380 for the -negative- the + is replaced by a - in the test)."
382 Raymond Toy comments that this is tricky on the X86 since its FPU
383 uses 80-bit precision internally.
386 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
387 the entire lexical environment at the point of MACROLET available
388 in the bodies of the macroexpander functions. In particular, it
389 allows the function bodies (which run at compile time) to try to
390 access lexical variables (which are only defined at runtime).
391 It doesn't even issue a warning, which is bad.
393 The SBCL behavior arguably conforms to the ANSI spec (since the
394 spec says that the behavior is undefined, ergo anything conforms).
395 However, it would be better to issue a compile-time error.
396 Unfortunately I (WHN) don't see any simple way to detect this
397 condition in order to issue such an error, so for the meantime
398 SBCL just does this weird broken "conforming" thing.
400 The ANSI standard says, in the definition of the special operator
402 The macro-expansion functions defined by MACROLET are defined
403 in the lexical environment in which the MACROLET form appears.
404 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
405 the local macro definitions in a MACROLET, but the consequences
406 are undefined if the local macro definitions reference any
407 local variable or function bindings that are visible in that
409 Then it seems to contradict itself by giving the example
411 (macrolet ((fudge (z)
412 ;The parameters x and flag are not accessible
413 ; at this point; a reference to flag would be to
414 ; the global variable of that name.
415 ` (if flag (* ,z ,z) ,z)))
416 ;The parameters x and flag are accessible here.
420 The comment "a reference to flag would be to the global variable
421 of the same name" sounds like good behavior for the system to have.
422 but actual specification quoted above says that the actual behavior
425 (Since 0.7.8.23 macroexpanders are defined in a restricted version
426 of the lexical environment, containing no lexical variables and
427 functions, which seems to conform to ANSI and CLtL2, but signalling
428 a STYLE-WARNING for references to variables similar to locals might
432 (as reported by Gabe Garza on cmucl-help 2001-09-21)
434 (defun test-pred (x y)
438 (func (lambda () x)))
439 (print (eq func func))
440 (print (test-pred func func))
441 (delete func (list func))))
442 Now calling (TEST-CASE) gives output
445 (#<FUNCTION {500A9EF9}>)
446 Evidently Python thinks of the lambda as a code transformation so
447 much that it forgets that it's also an object.
450 Ideally, uninterning a symbol would allow it, and its associated
451 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
452 at least as of sbcl-0.7.0, this isn't the case. Information about
453 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
454 essentially in ordinary (non-weak) hash tables keyed by symbols.
455 Thus, once a system has an entry in this system, it tends to live
456 forever, even when it is uninterned and all other references to it
459 141: "pretty printing and backquote"
462 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
464 c. (reported by Paul F. Dietz)
466 `(LAMBDA (SB-IMPL::BACKQ-COMMA X))
469 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
470 prominent SourceForge web/db bug tracking system, which is
471 unfortunately not a reliable way to get a timely response from
472 the SBCL maintainers)
473 In the course of trying to build a test case for an
474 application error, I encountered this behavior:
475 If you start up sbcl, and then lay on CTRL-C for a
476 minute or two, the lisp process will eventually say:
477 %PRIMITIVE HALT called; the party is over.
478 and throw you into the monitor. If I start up lisp,
479 attach to the process with strace, and then do the same
480 (abusive) thing, I get instead:
481 access failure in heap page not marked as write-protected
482 and the monitor again. I don't know enough to have the
483 faintest idea of what is going on here.
484 This is with sbcl 6.12, uname -a reports:
485 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
486 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
487 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
488 it took more than two minutes (but less than five) for me.
492 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
493 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
494 COMPLEX implementation didn't deal with this, and hasn't been
495 upgraded to do so. (This doesn't seem to be a high priority
496 conformance problem, since seems hard to construct useful code
499 b. (fixed in 0.8.3.43)
502 Floating point errors are reported poorly. E.g. on x86 OpenBSD
505 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
506 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
507 No traps are enabled? How can this be?
508 It should be possible to be much more specific (overflow, division
509 by zero, etc.) and of course the "How can this be?" should be fixable.
511 See also bugs #45.c and #183
514 (reported by Robert E. Brown 2002-04-16)
515 When a function is called with too few arguments, causing the
516 debugger to be entered, the uninitialized slots in the bad call frame
517 seem to cause GCish problems, being interpreted as tagged data even
518 though they're not. In particular, executing ROOM in the
519 debugger at that point causes AVER failures:
522 * (lisp-implementation-version)
528 failed AVER: "(SAP= CURRENT END)"
529 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
530 isn't too surprising since there are many differences in stack
531 implementation and GC conservatism between the X86 and other ports.)
533 This is probably the same bug as 216
536 In sbcl-0.7.3.11, compiling the (illegal) code
537 (in-package :cl-user)
538 (defmethod prove ((uustk uustk))
541 gives the (not terribly clear) error message
543 ; (during macroexpansion of (DEFMETHOD PROVE ...))
544 ; can't get template for (FROB NIL NIL)
545 The problem seems to be that the code walker used by the DEFMETHOD
546 macro is unhappy with the illegal syntax in the method body, and
547 is giving an unclear error message.
550 The compiler sometimes tries to constant-fold expressions before
551 it checks to see whether they can be reached. This can lead to
552 bogus warnings about errors in the constant folding, e.g. in code
555 (WRITE-STRING (> X 0) "+" "0"))
556 compiled in a context where the compiler can prove that X is NIL,
557 and the compiler complains that (> X 0) causes a type error because
558 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
559 caused a full WARNING, which made the bug really annoying because then
560 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
561 code. Since then the warning has been downgraded to STYLE-WARNING,
562 so it's still a bug but at least it's a little less annoying.
564 183: "IEEE floating point issues"
565 Even where floating point handling is being dealt with relatively
566 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
567 accrued-exceptions and current-exceptions part of the fp control
568 word don't seem to bear much relation to reality. E.g. on
572 debugger invoked on condition of type DIVISION-BY-ZERO:
573 arithmetic error DIVISION-BY-ZERO signalled
574 0] (sb-vm::get-floating-point-modes)
576 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
577 :ROUNDING-MODE :NEAREST
578 :CURRENT-EXCEPTIONS NIL
579 :ACCRUED-EXCEPTIONS (:INEXACT)
582 * (sb-vm::get-floating-point-modes)
583 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
584 :ROUNDING-MODE :NEAREST
585 :CURRENT-EXCEPTIONS (:INEXACT)
586 :ACCRUED-EXCEPTIONS (:INEXACT)
589 188: "compiler performance fiasco involving type inference and UNION-TYPE"
593 (declare (optimize (safety 3)))
594 (declare (optimize (compilation-speed 2)))
595 (declare (optimize (speed 1) (debug 1) (space 1)))
597 (declare (type (integer 0) start))
598 (print (incf start 22))
599 (print (incf start 26))
600 (print (incf start 28)))
602 (declare (type (integer 0) start))
603 (print (incf start 22))
604 (print (incf start 26)))
606 (declare (type (integer 0) start))
607 (print (incf start 22))
608 (print (incf start 26))))))
610 This example could be solved with clever enough constraint
611 propagation or with SSA, but consider
616 The careful type of X is {2k} :-(. Is it really important to be
617 able to work with unions of many intervals?
619 191: "Miscellaneous PCL deficiencies"
620 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
621 a. DEFCLASS does not inform the compiler about generated
622 functions. Compiling a file with
626 (WITH-SLOTS (A-CLASS-X) A
628 results in a STYLE-WARNING:
630 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
632 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
633 Mai points out that the declamation of functions is in fact
634 incorrect in some cases (most notably for structure
635 classes). This means that at present erroneous attempts to use
636 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
637 won't get the corresponding STYLE-WARNING.
638 c. (fixed in 0.8.4.23)
640 201: "Incautious type inference from compound types"
641 a. (reported by APD sbcl-devel 2002-09-17)
643 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
645 (FORMAT NIL "~S IS ~S, Y = ~S"
652 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
656 (declare (type (array * (4 4)) x))
658 (setq x (make-array '(4 4)))
659 (adjust-array y '(3 5))
660 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
662 * (foo (make-array '(4 4) :adjustable t))
665 205: "environment issues in cross compiler"
666 (These bugs have no impact on user code, but should be fixed or
668 a. Macroexpanders introduced with MACROLET are defined in the null
670 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
671 the null lexical environment.
672 c. The cross-compiler cannot inline functions defined in a non-null
675 206: ":SB-FLUID feature broken"
676 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
677 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
680 207: "poorly distributed SXHASH results for compound data"
681 SBCL's SXHASH could probably try a little harder. ANSI: "the
682 intent is that an implementation should make a good-faith
683 effort to produce hash-codes that are well distributed
684 within the range of non-negative fixnums". But
685 (let ((hits (make-hash-table)))
688 (let* ((ij (cons i j))
689 (newlist (push ij (gethash (sxhash ij) hits))))
691 (format t "~&collision: ~S~%" newlist))))))
692 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
693 would be an obvious way of fix. Maybe it would be acceptably efficient
694 to redo MIX using a lookup into a 256-entry s-box containing
695 29-bit pseudorandom numbers?
697 211: "keywords processing"
698 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
699 number of keyword arguments.
702 (flet ((foo (&key y) (list y)))
703 (list (foo :y 1 :y 2)))
705 issues confusing message
710 ; caught STYLE-WARNING:
711 ; The variable #:G15 is defined but never used.
713 212: "Sequence functions and circular arguments"
714 COERCE, MERGE and CONCATENATE go into an infinite loop when given
715 circular arguments; it would be good for the user if they could be
716 given an error instead (ANSI 17.1.1 allows this behaviour on the part
717 of the implementation, as conforming code cannot give non-proper
718 sequences to these functions. MAP also has this problem (and
719 solution), though arguably the convenience of being able to do
720 (MAP 'LIST '+ FOO '#1=(1 . #1#))
721 might be classed as more important (though signalling an error when
722 all of the arguments are circular is probably desireable).
724 213: "Sequence functions and type checking"
725 a. (fixed in 0.8.4.36)
726 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
727 check that it will return a sequence of the given type. Fixing
728 it along the same lines as the others (cf. work done around
729 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
730 entirely straightforward.
731 c. All of these functions will silently accept a type of the form
733 whether or not the return value is of this type. This is
734 probably permitted by ANSI (see "Exceptional Situations" under
735 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
736 know about this escape clause, so code of the form
737 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
738 can erroneously return T.
740 215: ":TEST-NOT handling by functions"
741 a. FIND and POSITION currently signal errors when given non-NIL for
742 both their :TEST and (deprecated) :TEST-NOT arguments, but by
743 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
744 means that the effect is "unpredictable but harmless". It's not
745 clear what that actually means; it may preclude conforming
746 implementations from signalling errors.
747 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
748 when conflict occurs. As a quality of implementation issue, it
749 might be preferable to treat :TEST and :TEST-NOT as being in some
750 sense the same &KEY, and effectively take the first test function in
752 c. Again, a quality of implementation issue: it would be good to issue a
753 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
754 WARNING for calls with both :TEST and :TEST-NOT; possibly this
755 latter should be WARNed about at execute-time too.
757 216: "debugger confused by frames with invalid number of arguments"
758 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
759 leaves the system confused, enough so that (QUIT) no longer works.
760 It's as though the process of working with the uninitialized slot in
761 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
762 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
764 This is probably the same bug as 162
766 217: "Bad type operations with FUNCTION types"
769 * (values-type-union (specifier-type '(function (base-char)))
770 (specifier-type '(function (integer))))
772 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
774 It causes insertion of wrong type assertions into generated
778 (let ((f (etypecase x
779 (character #'write-char)
780 (integer #'write-byte))))
783 (character (write-char x s))
784 (integer (write-byte x s)))))
786 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
788 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
789 produce invalid code, but type checking is not accurate.)
791 233: bugs in constraint propagation
793 (declaim (optimize (speed 2) (safety 3)))
795 (if (typep (prog1 x (setq x y)) 'double-float)
798 (foo 1d0 5) => segmentation violation
800 235: "type system and inline expansion"
802 (declaim (ftype (function (cons) number) acc))
803 (declaim (inline acc))
805 (the number (car c)))
808 (values (locally (declare (optimize (safety 0)))
810 (locally (declare (optimize (safety 3)))
813 (foo '(nil) '(t)) => NIL, T.
815 237: "Environment arguments to type functions"
816 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
817 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
818 argument, but they ignore it completely. This is almost
819 certainly not correct.
820 b. Also, the compiler's optimizers for TYPEP have not been informed
821 about the new argument; consequently, they will not transform
822 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
823 just as optimizeable as (TYPEP 1 'INTEGER).
825 238: "REPL compiler overenthusiasm for CLOS code"
827 * (defclass foo () ())
828 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
829 causes approximately 100 lines of code deletion notes. Some
830 discussion on this issue happened under the title 'Three "interesting"
831 bugs in PCL', resulting in a fix for this oververbosity from the
832 compiler proper; however, the problem persists in the interactor
833 because the notion of original source is not preserved: for the
834 compiler, the original source of the above expression is (DEFMETHOD
835 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
836 compiler gets its hands on the code needing compilation from the REPL,
837 it has been macroexpanded several times.
839 A symptom of the same underlying problem, reported by Tony Martinez:
841 (with-input-from-string (*query-io* " no")
843 (simple-type-error () 'error))
845 ; (SB-KERNEL:FLOAT-WAIT)
847 ; note: deleting unreachable code
848 ; compilation unit finished
851 241: "DEFCLASS mysteriously remembers uninterned accessor names."
852 (from tonyms on #lisp IRC 2003-02-25)
853 In sbcl-0.7.12.55, typing
854 (defclass foo () ((bar :accessor foo-bar)))
857 (defclass foo () ((bar :accessor foo-bar)))
858 gives the error message
859 "#:FOO-BAR already names an ordinary function or a macro."
860 So it's somehow checking the uninterned old accessor name instead
861 of the new requested accessor name, which seems broken to me (WHN).
863 242: "WRITE-SEQUENCE suboptimality"
864 (observed from clx performance)
865 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
866 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
867 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
868 rather than writing the sequence in one go, leading to severe
869 performance degradation.
871 243: "STYLE-WARNING overenthusiasm for unused variables"
872 (observed from clx compilation)
873 In sbcl-0.7.14, in the presence of the macros
874 (DEFMACRO FOO (X) `(BAR ,X))
875 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
876 somewhat surprising style warnings are emitted for
877 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
879 ; (LAMBDA (Y) (FOO Y))
881 ; caught STYLE-WARNING:
882 ; The variable Y is defined but never used.
884 245: bugs in disassembler
885 a. On X86 an immediate operand for IMUL is printed incorrectly.
886 b. On X86 operand size prefix is not recognized.
889 (defun foo (&key (a :x))
893 does not cause a warning. (BTW: old SBCL issued a warning, but for a
894 function, which was never called!)
897 Compiler does not emit warnings for
899 a. (lambda () (svref (make-array 8 :adjustable t) 1))
902 (list (let ((y (the real x)))
903 (unless (floatp y) (error ""))
908 (declare (optimize (debug 0)))
909 (declare (type vector x))
910 (list (fill-pointer x)
914 Complex array type does not have corresponding type specifier.
916 This is a problem because the compiler emits optimization notes when
917 you use a non-simple array, and without a type specifier for hairy
918 array types, there's no good way to tell it you're doing it
919 intentionally so that it should shut up and just compile the code.
921 Another problem is confusing error message "asserted type ARRAY
922 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
923 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
925 The last problem is that when type assertions are converted to type
926 checks, types are represented with type specifiers, so we could lose
927 complex attribute. (Now this is probably not important, because
928 currently checks for complex arrays seem to be performed by
932 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
933 warning. Analogous cases with the index and length being equal and
934 greater than 0 are warned for; the problem here seems to be that the
935 type required for an array reference of this type is (INTEGER 0 (0))
936 which is canonicalized to NIL.
941 (t1 (specifier-type s)))
942 (eval `(defstruct ,s))
943 (type= t1 (specifier-type s)))
948 b. The same for CSUBTYPEP.
950 262: "yet another bug in inline expansion of local functions"
954 (declare (integer x y))
957 (declare (integer u))
958 (if (> (1+ (the unsigned-byte u)) 0)
960 (return (+ 38 (cos (/ u 78)))))))
961 (declare (inline xyz))
963 (* (funcall (eval #'xyz) x)
965 (funcall (if (> x 5) #'xyz #'identity)
970 Urgh... It's time to write IR1-copier.
973 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
974 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
975 looks reasonable, if not obviously correct; however, it caused the
976 PPC/Linux port to segfault during warm-init while loading
977 src/pcl/std-class.fasl. A workaround patch was made, but it would
978 be nice to understand why the first patch caused problems, and to
979 fix the cause if possible.
981 268: "wrong free declaration scope"
982 The following code must signal type error:
984 (locally (declare (optimize (safety 3)))
985 (flet ((foo (x &optional (y (car x)))
986 (declare (optimize (safety 0)))
988 (funcall (eval #'foo) 1)))
991 SCALE-FLOAT should accept any integer for its second argument.
994 In the following function constraint propagator optimizes nothing:
997 (declare (integer x))
998 (declare (optimize speed))
1006 Compilation of the following two forms causes "X is unbound" error:
1008 (symbol-macrolet ((x pi))
1009 (macrolet ((foo (y) (+ x y)))
1010 (declaim (inline bar))
1016 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
1019 CLHS says that type declaration of a symbol macro should not affect
1020 its expansion, but in SBCL it does. (If you like magic and want to
1021 fix it, don't forget to change all uses of MACROEXPAND to
1025 The following code (taken from CLOCC) takes a lot of time to compile:
1028 (declare (type (integer 0 #.large-constant) n))
1031 (fixed in 0.8.2.51, but a test case would be good)
1034 (defmethod fee ((x fixnum))
1037 (fee 1) => type error
1044 (declare (optimize speed))
1045 (loop for i of-type (integer 0) from 0 by 2 below 10
1048 uses generic arithmetic.
1050 b. (fixed in 0.8.3.6)
1052 279: type propagation error -- correctly inferred type goes astray?
1053 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
1054 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
1055 &OPTIONAL), not a (INTEGER 1 536870911)
1056 is emitted when compiling this file:
1057 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
1058 (integer #.most-negative-fixnum 0))
1063 (let* (;; Uncomment this for a type mismatch warning indicating
1064 ;; that the type of (FOO X) is correctly understood.
1065 #+nil (fs-foo (float-sign (foo x)))
1066 ;; Uncomment this for a type mismatch warning
1067 ;; indicating that the type of (ABS (FOO X)) is
1068 ;; correctly understood.
1069 #+nil (fs-abs-foo (float-sign (abs (foo x))))
1070 ;; something wrong with this one though
1071 (abs-foo (abs (foo x))))
1072 (declare (type (integer 1 100) abs-foo))
1077 280: bogus WARNING about duplicate function definition
1078 In sbcl-0.8.3 and sbcl-0.8.1.47, if BS.MIN is defined inline,
1080 (declaim (inline bs.min))
1081 (defun bs.min (bases) nil)
1082 before compiling the file below, the compiler warns
1083 Duplicate definition for BS.MIN found in one static
1084 unit (usually a file).
1086 (declaim (special *minus* *plus* *stagnant*))
1087 (defun b.*.min (&optional (x () xp) (y () yp) &rest rest)
1089 (define-compiler-macro b.*.min (&rest rest)
1091 (defun afish-d-rbd (pd)
1093 (b.*.min (foo-d-rbd *stagnant*))
1094 (multiple-value-bind (reduce-fn initial-value)
1096 (list (values #'bs.min 0))
1097 (vector (values #'bs.min *plus*)))
1098 (let ((cv-ks (cv (kpd.ks pd))))
1099 (funcall reduce-fn d-rbds)))))
1100 (defun bfish-d-rbd (pd)
1102 (b.*.min (foo-d-rbd *stagnant*))
1103 (multiple-value-bind (reduce-fn initial-value)
1105 (list (values #'bs.min *minus*))
1106 (vector (values #'bs.min 0)))
1107 (let ((cv-ks (cv (kpd.ks pd))))
1108 (funcall reduce-fn d-rbds)))))
1110 281: COMPUTE-EFFECTIVE-METHOD error signalling.
1111 (slightly obscured by a non-0 default value for
1112 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
1113 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
1114 when it finds a method with invalid qualifiers. However, it
1115 shouldn't signal errors when any such methods are not applicable to
1116 the particular call being evaluated, and certainly it shouldn't when
1117 simply precomputing effective methods that may never be called.
1118 (setf sb-pcl::*max-emf-precompute-methods* 0)
1120 (:method-combination +)
1121 (:method ((x symbol)) 1)
1122 (:method + ((x number)) x))
1123 (foo 1) -> ERROR, but should simply return 1
1125 The issue seems to be that construction of a discriminating function
1126 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
1128 283: Thread safety: libc functions
1129 There are places that we call unsafe-for-threading libc functions
1130 that we should find alternatives for, or put locks around. Known or
1131 strongly suspected problems, as of 0.8.3.10: please update this
1132 bug instead of creating new ones
1134 localtime() - called for timezone calculations in code/time.lisp
1136 284: Thread safety: special variables
1137 There are lots of special variables in SBCL, and I feel sure that at
1138 least some of them are indicative of potentially thread-unsafe
1139 parts of the system. See doc/internals/notes/threading-specials
1141 286: "recursive known functions"
1142 Self-call recognition conflicts with known function
1143 recognition. Currently cross compiler and target COMPILE do not
1144 recognize recursion, and in target compiler it can be disabled. We
1145 can always disable it for known functions with RECURSIVE attribute,
1146 but there remains a possibility of a function with a
1147 (tail)-recursive simplification pass and transforms/VOPs for base
1150 287: PPC/Linux miscompilation or corruption in first GC
1151 When the runtime is compiled with -O3 on certain PPC/Linux machines, a
1152 segmentation fault is reported at the point of first triggered GC,
1153 during the compilation of DEFSTRUCT WRAPPER. As a temporary workaround,
1154 the runtime is no longer compiled with -O3 on PPC/Linux, but it is likely
1155 that this merely obscures, not solves, the underlying problem; as and when
1156 underlying problems are fixed, it would be worth trying again to provoke
1159 288: fundamental cross-compilation issues (from old UGLINESS file)
1160 Using host floating point numbers to represent target floating point
1161 numbers, or host characters to represent target characters, is
1162 theoretically shaky. (The characters are OK as long as the characters
1163 are in the ANSI-guaranteed character set, though, so they aren't a
1164 real problem as long as the sources don't need anything but that;
1165 the floats are a real problem.)
1167 289: "type checking and source-transforms"
1169 (block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
1172 Our policy is to check argument types at the moment of a call. It
1173 disagrees with ANSI, which says that type assertions are put
1174 immediately onto argument expressions, but is easier to implement in
1175 IR1 and is more compatible to type inference, inline expansion,
1176 etc. IR1-transforms automatically keep this policy, but source
1177 transforms for associative functions (such as +), being applied
1178 during IR1-convertion, do not. It may be tolerable for direct calls
1179 (+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
1181 b. Another aspect of this problem is efficiency. [x y + z +]
1182 requires less registers than [x y z + +]. This transformation is
1183 currently performed with source transforms, but it would be good to
1184 also perform it in IR1 optimization phase.
1186 290: Alpha floating point and denormalized traps
1187 In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
1188 hardware or kernel deficiency: the status of the enable/disable
1189 denormalized-float traps bit seems to be ambiguous; by the time we
1190 get to os_restore_fp_control after a trap, denormalized traps seem
1191 to be enabled. Since we don't want a trap every time someone uses a
1192 denormalized float, in general, we mask out that bit when we restore
1193 the control word; however, this clobbers any change the user might
1197 (reported by Adam Warner, sbcl-devel 2003-09-23)
1199 The --load toplevel argument does not perform any sanitization of its
1200 argument. As a result, files with Lisp pathname pattern characters
1201 (#\* or #\?, for instance) or quotation marks can cause the system
1202 to perform arbitrary behaviour.
1205 LOOP with non-constant arithmetic step clauses suffers from overzealous
1206 type constraint: code of the form
1207 (loop for d of-type double-float from 0d0 to 10d0 by x collect d)
1208 compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
1209 (0))). However, an integral value of X should be legal, because
1210 successive adds of integers to double-floats produces double-floats,
1211 so none of the type restrictions in the code is violated.
1213 298: (aka PFD MISC.183)
1217 (multiple-value-call #'bar
1219 (catch 'tag (return-from foo (int)))))
1221 This program violates "unknown values LVAR stack discipline": if INT
1222 returns, values returned by (EXT) must be removed from under that of
1225 300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
1226 argument type only after having read a character. This is caused
1227 with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
1228 exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
1229 less error prone to have EXPLICIT-CHECK be a local declaration,
1230 being put into the definition, instead of an attribute being kept in
1231 a separate file; maybe also put it into SB-EXT?
1233 301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
1234 in NOTE-ASSUMED-TYPES
1235 In sbcl-0.8.7.32, compiling the file
1237 (declare (type integer x))
1238 (declare (type (vector (or hash-table bit)) y))
1241 (declare (type integer x))
1242 (declare (type (simple-array base (2)) y))
1245 failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
1247 302: Undefined type messes up DATA-VECTOR-REF expansion.
1249 (defun dis (s ei x y)
1250 (declare (type (simple-array function (2)) s) (type ei ei))
1251 (funcall (aref s ei) x y))
1252 on sbcl-0.8.7.36/X86/Linux causes a BUG to be signalled:
1253 full call to SB-KERNEL:DATA-VECTOR-REF