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
90 The "compiling top-level form:" output ought to be condensed.
91 Perhaps any number of such consecutive lines ought to turn into a
92 single "compiling top-level forms:" line.
95 It would be nice if the
97 (during macroexpansion)
98 said what macroexpansion was at fault, e.g.
100 (during macroexpansion of IN-PACKAGE,
101 during macroexpansion of DEFFOO)
104 (I *think* this is a bug. It certainly seems like strange behavior. But
105 the ANSI spec is scary, dark, and deep.. -- WHN)
106 (FORMAT NIL "~,1G" 1.4) => "1. "
107 (FORMAT NIL "~3,1G" 1.4) => "1. "
110 Sometimes (SB-EXT:QUIT) fails with
111 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
112 Process inferior-lisp exited abnormally with code 1
113 I haven't noticed a repeatable case of this yet.
116 The printer doesn't report closures very well. This is true in
120 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
121 It would be nice to make closures have a settable name slot,
122 and make things like DEFSTRUCT and FLET, which create closures,
123 set helpful values into this slot.
126 And as long as we're wishing, it would be awfully nice if INSPECT could
127 also report on closures, telling about the values of the bound variables.
130 The compiler assumes that any time a function of declared FTYPE
131 doesn't signal an error, its arguments were of the declared type.
132 E.g. compiling and loading
133 (DECLAIM (OPTIMIZE (SAFETY 3)))
134 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
136 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
138 (COND ((> (FACTORIAL X) 1.0E6)
139 (FORMAT T "too big~%"))
141 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
143 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
146 will cause the INTEGERP case to be selected, giving bogus output a la
148 This violates the "declarations are assertions" principle.
149 According to the ANSI spec, in the section "System Class FUNCTION",
150 this is a case of "lying to the compiler", but the lying is done
151 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
152 not by the unexpectedly general definition of FACTORIAL. In any case,
153 "declarations are assertions" means that lying to the compiler should
154 cause an error to be signalled, and should not cause a bogus
155 result to be returned. Thus, the compiler should not assume
156 that arbitrary functions check their argument types. (It might
157 make sense to add another flag (CHECKED?) to DEFKNOWN to
158 identify functions which *do* check their argument types.)
159 (Also, verify that the compiler handles declared function
160 return types as assertions.)
163 The definitions of SIGCONTEXT-FLOAT-REGISTER and
164 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
165 supported on FreeBSD because the floating point state is not saved,
166 but at least as of FreeBSD 4.0, the floating point state *is* saved,
167 so they could be supported after all. Very likely
168 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
171 a slew of floating-point-related errors reported by Peter Van Eynde
173 c: Many expressions generate floating infinity on x86/Linux:
178 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
179 on x86/Linux generates the infinities instead. That might or
180 might not be conforming behavior, but it's also inconsistent,
181 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
182 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
183 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
185 d: (in section12.erg) various forms a la
186 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
187 don't give the right behavior.
190 type safety errors reported by Peter Van Eynde July 25, 2000:
191 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
192 not a binary input stream, but instead cheerfully reads from
193 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
196 The debugger LIST-LOCATIONS command doesn't work properly.
197 (How should it work properly?)
200 Compiling and loading
201 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
203 then requesting a BACKTRACE at the debugger prompt gives no information
204 about where in the user program the problem occurred.
207 Using the pretty-printer from the command prompt gives funny
208 results, apparently because the pretty-printer doesn't know
209 about user's command input, including the user's carriage return
210 that the user, and therefore the pretty-printer thinks that
211 the new output block should start indented 2 or more characters
212 rightward of the correct location.
215 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
216 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
217 crashes SBCL. In general tracing anything which is used in the
218 implementation of TRACE is likely to have the same problem.
221 ANSI says in one place that type declarations can be abbreviated even
222 when the type name is not a symbol, e.g.
223 (DECLAIM ((VECTOR T) *FOOVECTOR*))
224 SBCL doesn't support this. But ANSI says in another place that this
225 isn't allowed. So it's not clear this is a bug after all. (See the
226 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
230 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
231 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
232 an easily guessable temporary filename in a way which might open
233 applications using LOAD-FOREIGN to hijacking by malicious users
234 on the same machine. Incantations for doing this safely are
235 floating around the net in various "how to write secure programs
236 despite Unix" documents, and it would be good to (1) fix this in
237 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
238 files and make it share the same new safe logic.
240 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
241 make the temporary filename less easily guessable)
244 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
245 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
246 the mean of the distribution can be systematically O(0.1%) wrong.
247 Just increasing R-I-E-B is probably not a good solution, since
248 it would decrease efficiency more than is probably necessary. Perhaps
249 using some sort of accept/reject method would be better.
252 Internally the compiler sometimes evaluates
253 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
254 (I stumbled across this when I added an
255 (assert (not (eq type1 *wild-type*)))
256 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
257 in a type context should probably be translated to T, and so it's
258 probably wrong to ask whether it's equal to the T type and then (using
259 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
260 (I haven't tried to investigate this bug enough to guess whether
261 there might be any user-level symptoms.)
263 In fact, the type system is likely to depend on this inequality not
264 holding... * is not equivalent to T in many cases, such as
265 (VECTOR *) /= (VECTOR T).
268 The facility for dumping a running Lisp image to disk gets confused
269 when run without the PURIFY option, and creates an unnecessarily large
270 core file (apparently representing memory usage up to the previous
271 high-water mark). Moreover, when the file is loaded, it confuses the
272 GC, so that thereafter memory usage can never be reduced below that
276 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
277 CL), out-of-line structure slot setters are horribly inefficient
278 whenever the type of the slot is declared, because out-of-line
279 structure slot setters are implemented as closures to save space,
280 so the compiler doesn't compile the type test into code, but
281 instead just saves the type in a lexical closure and interprets it
283 To exercise the problem, compile and load
284 (cl:in-package :cl-user)
286 (bar (error "missing") :type bar))
289 (loop (setf (foo-bar *foo*) x)))
291 (defvar *bar* (make-bar))
292 (defvar *foo* (make-foo :bar *bar*))
293 (defvar *setf-foo-bar* #'(setf foo-bar))
295 (loop (funcall *setf-foo-bar* x *foo*)))
296 then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
297 use BACKTRACE, to see it's spending all essentially all its time
298 in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
299 One possible solution would be simply to give up on
300 representing structure slot accessors as functions, and represent
301 them as macroexpansions instead. This can be inconvenient for users,
302 but it's not clear that it's worse than trying to help by expanding
303 into a horribly inefficient implementation.
304 As a workaround for the problem, #'(SETF FOO) expressions
305 can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
306 (defmacro efficient-setf-function (place-function-name)
307 (or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
308 ;; a workaround for the problem, encouraging the
309 ;; inline expansion of the structure accessor, so
310 ;; that the compiler can optimize its type test
311 (let ((new-value (gensym "NEW-VALUE-"))
312 (structure-value (gensym "STRUCTURE-VALUE-")))
313 `(lambda (,new-value ,structure-value)
314 (setf (,place-function-name ,structure-value)
316 ;; no problem, can just use the ordinary expansion
317 `(function (setf ,place-function-name))))
320 There's apparently a bug in CEILING optimization which caused
321 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
322 applied the patches to SBCL and they didn't seem to cause problems
323 (as reported sbcl-devel 2001-05-04). However, since the patches
324 modify nontrivial code which was apparently written incorrectly
325 the first time around, until regression tests are written I'm not
326 comfortable merging the patches in the CVS version of SBCL.
329 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
330 a clean, just-started SBCL system. And it seems to be right:
331 (ROOM T) can bring a small computer to its knees for a *long*
332 time trying to GC afterwards. Surely there's some more economical
333 way to implement (ROOM T).
336 When the compiler inline expands functions, it may be that different
337 kinds of return values are generated from different code branches.
338 E.g. an inline expansion of POSITION generates integer results
339 from one branch, and NIL results from another. When that inline
340 expansion is used in a context where only one of those results
343 (aref *a1* (position x *a2*)))
344 and the compiler can't prove that the unacceptable branch is
345 never taken, then bogus type mismatch warnings can be generated.
346 If you need to suppress the type mismatch warnings, you can
347 suppress the inline expansion,
349 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
350 (aref *a1* (position x *a2*)))
351 or, sometimes, suppress them by declaring the result to be of an
354 (aref *a1* (the integer (position x *a2*))))
356 This is not a new compiler problem in 0.7.0, but the new compiler
357 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
358 more conspicuous. If you don't need performance from these functions,
359 and the bogus warnings are a nuisance for you, you can return to
360 your pre-0.7.0 state of grace with
361 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
366 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
367 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
368 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
369 when of course it should be NIL. (He says it only fails for X86,
370 not SPARC; dunno about Alpha.)
372 Also, "the same problem exists for LONG-FLOAT-EPSILON,
373 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
374 for the -negative- the + is replaced by a - in the test)."
376 Raymond Toy comments that this is tricky on the X86 since its FPU
377 uses 80-bit precision internally.
380 Even in sbcl-0.pre7.x, which is supposed to be free of the old
381 non-ANSI behavior of treating the function return type inferred
382 from the current function definition as a declaration of the
383 return type from any function of that name, the return type of NIL
384 is attached to FOO in 120a above, and used to optimize code which
388 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
389 the entire lexical environment at the point of MACROLET available
390 in the bodies of the macroexpander functions. In particular, it
391 allows the function bodies (which run at compile time) to try to
392 access lexical variables (which are only defined at runtime).
393 It doesn't even issue a warning, which is bad.
395 The SBCL behavior arguably conforms to the ANSI spec (since the
396 spec says that the behavior is undefined, ergo anything conforms).
397 However, it would be better to issue a compile-time error.
398 Unfortunately I (WHN) don't see any simple way to detect this
399 condition in order to issue such an error, so for the meantime
400 SBCL just does this weird broken "conforming" thing.
402 The ANSI standard says, in the definition of the special operator
404 The macro-expansion functions defined by MACROLET are defined
405 in the lexical environment in which the MACROLET form appears.
406 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
407 the local macro definitions in a MACROLET, but the consequences
408 are undefined if the local macro definitions reference any
409 local variable or function bindings that are visible in that
411 Then it seems to contradict itself by giving the example
413 (macrolet ((fudge (z)
414 ;The parameters x and flag are not accessible
415 ; at this point; a reference to flag would be to
416 ; the global variable of that name.
417 ` (if flag (* ,z ,z) ,z)))
418 ;The parameters x and flag are accessible here.
422 The comment "a reference to flag would be to the global variable
423 of the same name" sounds like good behavior for the system to have.
424 but actual specification quoted above says that the actual behavior
427 (Since 0.7.8.23 macroexpanders are defined in a restricted version
428 of the lexical environment, containing no lexical variables and
429 functions, which seems to conform to ANSI and CLtL2, but signalling
430 a STYLE-WARNING for references to variables similar to locals might
434 (as reported by Gabe Garza on cmucl-help 2001-09-21)
436 (defun test-pred (x y)
440 (func (lambda () x)))
441 (print (eq func func))
442 (print (test-pred func func))
443 (delete func (list func))))
444 Now calling (TEST-CASE) gives output
447 (#<FUNCTION {500A9EF9}>)
448 Evidently Python thinks of the lambda as a code transformation so
449 much that it forgets that it's also an object.
452 Ideally, uninterning a symbol would allow it, and its associated
453 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
454 at least as of sbcl-0.7.0, this isn't the case. Information about
455 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
456 essentially in ordinary (non-weak) hash tables keyed by symbols.
457 Thus, once a system has an entry in this system, it tends to live
458 forever, even when it is uninterned and all other references to it
461 141: "pretty printing and backquote"
464 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
467 * (write '`(, .ala.) :readably t :pretty t)
470 (note the space between the comma and the point)
473 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
474 prominent SourceForge web/db bug tracking system, which is
475 unfortunately not a reliable way to get a timely response from
476 the SBCL maintainers)
477 In the course of trying to build a test case for an
478 application error, I encountered this behavior:
479 If you start up sbcl, and then lay on CTRL-C for a
480 minute or two, the lisp process will eventually say:
481 %PRIMITIVE HALT called; the party is over.
482 and throw you into the monitor. If I start up lisp,
483 attach to the process with strace, and then do the same
484 (abusive) thing, I get instead:
485 access failure in heap page not marked as write-protected
486 and the monitor again. I don't know enough to have the
487 faintest idea of what is going on here.
488 This is with sbcl 6.12, uname -a reports:
489 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
490 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
491 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
492 it took more than two minutes (but less than five) for me.
496 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
497 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
498 COMPLEX implementation didn't deal with this, and hasn't been
499 upgraded to do so. (This doesn't seem to be a high priority
500 conformance problem, since seems hard to construct useful code
505 (declare (type (double-float -0d0) x))
506 (declare (optimize speed))
507 (+ x (sqrt (log (random 1d0)))))
508 debugger invoked on condition of type SIMPLE-ERROR:
509 bad thing to be a type specifier: ((COMPLEX
511 #.SB-EXT:DOUBLE-FLOAT-POSITIVE-INFINITY))
512 #C(0.0d0 #.SB-EXT:DOUBLE-FLOAT-POSITIVE-INFINITY)
513 #C(0.0d0 #.SB-EXT:DOUBLE-FLOAT-POSITIVE-INFINITY))
516 Floating point errors are reported poorly. E.g. on x86 OpenBSD
519 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
520 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
521 No traps are enabled? How can this be?
522 It should be possible to be much more specific (overflow, division
523 by zero, etc.) and of course the "How can this be?" should be fixable.
525 See also bugs #45.c and #183
528 (reported by Robert E. Brown 2002-04-16)
529 When a function is called with too few arguments, causing the
530 debugger to be entered, the uninitialized slots in the bad call frame
531 seem to cause GCish problems, being interpreted as tagged data even
532 though they're not. In particular, executing ROOM in the
533 debugger at that point causes AVER failures:
536 * (lisp-implementation-version)
542 failed AVER: "(SAP= CURRENT END)"
543 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
544 isn't too surprising since there are many differences in stack
545 implementation and GC conservatism between the X86 and other ports.)
548 In sbcl-0.7.3.11, compiling the (illegal) code
549 (in-package :cl-user)
550 (defmethod prove ((uustk uustk))
553 gives the (not terribly clear) error message
555 ; (during macroexpansion of (DEFMETHOD PROVE ...))
556 ; can't get template for (FROB NIL NIL)
557 The problem seems to be that the code walker used by the DEFMETHOD
558 macro is unhappy with the illegal syntax in the method body, and
559 is giving an unclear error message.
562 The compiler sometimes tries to constant-fold expressions before
563 it checks to see whether they can be reached. This can lead to
564 bogus warnings about errors in the constant folding, e.g. in code
567 (WRITE-STRING (> X 0) "+" "0"))
568 compiled in a context where the compiler can prove that X is NIL,
569 and the compiler complains that (> X 0) causes a type error because
570 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
571 caused a full WARNING, which made the bug really annoying because then
572 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
573 code. Since then the warning has been downgraded to STYLE-WARNING,
574 so it's still a bug but at least it's a little less annoying.
576 183: "IEEE floating point issues"
577 Even where floating point handling is being dealt with relatively
578 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
579 accrued-exceptions and current-exceptions part of the fp control
580 word don't seem to bear much relation to reality. E.g. on
584 debugger invoked on condition of type DIVISION-BY-ZERO:
585 arithmetic error DIVISION-BY-ZERO signalled
586 0] (sb-vm::get-floating-point-modes)
588 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
589 :ROUNDING-MODE :NEAREST
590 :CURRENT-EXCEPTIONS NIL
591 :ACCRUED-EXCEPTIONS (:INEXACT)
594 * (sb-vm::get-floating-point-modes)
595 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
596 :ROUNDING-MODE :NEAREST
597 :CURRENT-EXCEPTIONS (:INEXACT)
598 :ACCRUED-EXCEPTIONS (:INEXACT)
601 188: "compiler performance fiasco involving type inference and UNION-TYPE"
605 (declare (optimize (safety 3)))
606 (declare (optimize (compilation-speed 2)))
607 (declare (optimize (speed 1) (debug 1) (space 1)))
609 (declare (type (integer 0) start))
610 (print (incf start 22))
611 (print (incf start 26))
612 (print (incf start 28)))
614 (declare (type (integer 0) start))
615 (print (incf start 22))
616 (print (incf start 26)))
618 (declare (type (integer 0) start))
619 (print (incf start 22))
620 (print (incf start 26))))))
622 This example could be solved with clever enough constraint
623 propagation or with SSA, but consider
628 The careful type of X is {2k} :-(. Is it really important to be
629 able to work with unions of many intervals?
631 190: "PPC/Linux pipe? buffer? bug"
632 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
633 on the PPC/Linux platform, waiting for a zombie env process. This
634 is a classic symptom of buffer filling and deadlock, but it seems
635 only sporadically reproducible.
637 191: "Miscellaneous PCL deficiencies"
638 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
639 a. DEFCLASS does not inform the compiler about generated
640 functions. Compiling a file with
644 (WITH-SLOTS (A-CLASS-X) A
646 results in a STYLE-WARNING:
648 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
650 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
651 Mai points out that the declamation of functions is in fact
652 incorrect in some cases (most notably for structure
653 classes). This means that at present erroneous attempts to use
654 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
655 won't get the corresponding STYLE-WARNING.
656 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
657 lists and &KEY arguments) do not signal errors when they should.
659 201: "Incautious type inference from compound types"
660 a. (reported by APD sbcl-devel 2002-09-17)
662 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
664 (FORMAT NIL "~S IS ~S, Y = ~S"
671 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
675 (declare (type (array * (4 4)) x))
677 (setq x (make-array '(4 4)))
678 (adjust-array y '(3 5))
679 (= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
681 * (foo (make-array '(4 4) :adjustable t))
684 205: "environment issues in cross compiler"
685 (These bugs have no impact on user code, but should be fixed or
687 a. Macroexpanders introduced with MACROLET are defined in the null
689 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
690 the null lexical environment.
691 c. The cross-compiler cannot inline functions defined in a non-null
694 206: ":SB-FLUID feature broken"
695 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
696 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
699 207: "poorly distributed SXHASH results for compound data"
700 SBCL's SXHASH could probably try a little harder. ANSI: "the
701 intent is that an implementation should make a good-faith
702 effort to produce hash-codes that are well distributed
703 within the range of non-negative fixnums". But
704 (let ((hits (make-hash-table)))
707 (let* ((ij (cons i j))
708 (newlist (push ij (gethash (sxhash ij) hits))))
710 (format t "~&collision: ~S~%" newlist))))))
711 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
712 would be an obvious way of fix. Maybe it would be acceptably efficient
713 to redo MIX using a lookup into a 256-entry s-box containing
714 29-bit pseudorandom numbers?
716 211: "keywords processing"
717 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
718 number of keyword arguments.
721 (flet ((foo (&key y) (list y)))
722 (list (foo :y 1 :y 2)))
724 issues confusing message
729 ; caught STYLE-WARNING:
730 ; The variable #:G15 is defined but never used.
732 212: "Sequence functions and circular arguments"
733 COERCE, MERGE and CONCATENATE go into an infinite loop when given
734 circular arguments; it would be good for the user if they could be
735 given an error instead (ANSI 17.1.1 allows this behaviour on the part
736 of the implementation, as conforming code cannot give non-proper
737 sequences to these functions. MAP also has this problem (and
738 solution), though arguably the convenience of being able to do
739 (MAP 'LIST '+ FOO '#1=(1 . #1#))
740 might be classed as more important (though signalling an error when
741 all of the arguments are circular is probably desireable).
743 213: "Sequence functions and type checking"
744 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
745 various complicated, though recognizeable, CONS types [e.g.
746 (CONS * (CONS * NULL))
747 which according to ANSI should be recognized] (and, in SAFETY 3
748 code, should return a list of LENGTH 2 or signal an error)
749 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
750 check that it will return a sequence of the given type. Fixing
751 it along the same lines as the others (cf. work done around
752 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
753 entirely straightforward.
754 c. All of these functions will silently accept a type of the form
756 whether or not the return value is of this type. This is
757 probably permitted by ANSI (see "Exceptional Situations" under
758 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
759 know about this escape clause, so code of the form
760 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
761 can erroneously return T.
764 SBCL 0.6.12.43 fails to compile
767 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
768 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
771 or a more simple example:
774 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
775 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
777 215: ":TEST-NOT handling by functions"
778 a. FIND and POSITION currently signal errors when given non-NIL for
779 both their :TEST and (deprecated) :TEST-NOT arguments, but by
780 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
781 means that the effect is "unpredictable but harmless". It's not
782 clear what that actually means; it may preclude conforming
783 implementations from signalling errors.
784 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
785 when conflict occurs. As a quality of implementation issue, it
786 might be preferable to treat :TEST and :TEST-NOT as being in some
787 sense the same &KEY, and effectively take the first test function in
789 c. Again, a quality of implementation issue: it would be good to issue a
790 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
791 WARNING for calls with both :TEST and :TEST-NOT; possibly this
792 latter should be WARNed about at execute-time too.
794 216: "debugger confused by frames with invalid number of arguments"
795 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
796 leaves the system confused, enough so that (QUIT) no longer works.
797 It's as though the process of working with the uninitialized slot in
798 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
799 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
801 217: "Bad type operations with FUNCTION types"
804 * (values-type-union (specifier-type '(function (base-char)))
805 (specifier-type '(function (integer))))
807 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
809 It causes insertion of wrong type assertions into generated
813 (let ((f (etypecase x
814 (character #'write-char)
815 (integer #'write-byte))))
818 (character (write-char x s))
819 (integer (write-byte x s)))))
821 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
823 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
824 produce invalid code, but type checking is not accurate.)
826 233: bugs in constraint propagation
829 (declare (optimize (speed 2) (safety 3)))
835 (quux y (+ y 2d0) (* y 3d0)))))
836 (foo 4) => segmentation violation
838 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
842 (declaim (optimize (speed 2) (safety 3)))
844 (if (typep (prog1 x (setq x y)) 'double-float)
847 (foo 1d0 5) => segmentation violation
849 235: "type system and inline expansion"
851 (declaim (ftype (function (cons) number) acc))
852 (declaim (inline acc))
854 (the number (car c)))
857 (values (locally (declare (optimize (safety 0)))
859 (locally (declare (optimize (safety 3)))
862 (foo '(nil) '(t)) => NIL, T.
864 237: "Environment arguments to type functions"
865 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
866 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
867 argument, but they ignore it completely. This is almost
868 certainly not correct.
869 b. Also, the compiler's optimizers for TYPEP have not been informed
870 about the new argument; consequently, they will not transform
871 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
872 just as optimizeable as (TYPEP 1 'INTEGER).
874 238: "REPL compiler overenthusiasm for CLOS code"
876 * (defclass foo () ())
877 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
878 causes approximately 100 lines of code deletion notes. Some
879 discussion on this issue happened under the title 'Three "interesting"
880 bugs in PCL', resulting in a fix for this oververbosity from the
881 compiler proper; however, the problem persists in the interactor
882 because the notion of original source is not preserved: for the
883 compiler, the original source of the above expression is (DEFMETHOD
884 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
885 compiler gets its hands on the code needing compilation from the REPL,
886 it has been macroexpanded several times.
888 A symptom of the same underlying problem, reported by Tony Martinez:
890 (with-input-from-string (*query-io* " no")
892 (simple-type-error () 'error))
894 ; (SB-KERNEL:FLOAT-WAIT)
896 ; note: deleting unreachable code
897 ; compilation unit finished
900 241: "DEFCLASS mysteriously remembers uninterned accessor names."
901 (from tonyms on #lisp IRC 2003-02-25)
902 In sbcl-0.7.12.55, typing
903 (defclass foo () ((bar :accessor foo-bar)))
906 (defclass foo () ((bar :accessor foo-bar)))
907 gives the error message
908 "#:FOO-BAR already names an ordinary function or a macro."
909 So it's somehow checking the uninterned old accessor name instead
910 of the new requested accessor name, which seems broken to me (WHN).
912 242: "WRITE-SEQUENCE suboptimality"
913 (observed from clx performance)
914 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
915 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
916 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
917 rather than writing the sequence in one go, leading to severe
918 performance degradation.
920 243: "STYLE-WARNING overenthusiasm for unused variables"
921 (observed from clx compilation)
922 In sbcl-0.7.14, in the presence of the macros
923 (DEFMACRO FOO (X) `(BAR ,X))
924 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
925 somewhat surprising style warnings are emitted for
926 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
928 ; (LAMBDA (Y) (FOO Y))
930 ; caught STYLE-WARNING:
931 ; The variable Y is defined but never used.
933 245: bugs in disassembler
934 a. On X86 an immediate operand for IMUL is printed incorrectly.
935 b. On X86 operand size prefix is not recognized.
937 248: "reporting errors in type specifier syntax"
938 (TYPEP 1 '(SYMBOL NIL)) says something about "unknown type
942 (defun foo (&key (a :x))
946 does not cause a warning. (BTW: old SBCL issued a warning, but for a
947 function, which was never called!)
950 Compiler does not emit warnings for
952 a. (lambda () (svref (make-array 8 :adjustable t) 1))
955 (list (let ((y (the real x)))
956 (unless (floatp y) (error ""))
961 (declare (optimize (debug 0)))
962 (declare (type vector x))
963 (list (fill-pointer x)
967 Complex array type does not have corresponding type specifier.
969 This is a problem because the compiler emits optimization notes when
970 you use a non-simple array, and without a type specifier for hairy
971 array types, there's no good way to tell it you're doing it
972 intentionally so that it should shut up and just compile the code.
974 Another problem is confusing error message "asserted type ARRAY
975 conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
976 compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
978 The last problem is that when type assertions are converted to type
979 checks, types are represented with type specifiers, so we could lose
980 complex attribute. (Now this is probably not important, because
981 currently checks for complex arrays seem to be performed by
985 (compile nil '(lambda () (aref (make-array 0) 0))) compiles without
986 warning. Analogous cases with the index and length being equal and
987 greater than 0 are warned for; the problem here seems to be that the
988 type required for an array reference of this type is (INTEGER 0 (0))
989 which is canonicalized to NIL.
994 (t1 (specifier-type s)))
995 (eval `(defstruct ,s))
996 (type= t1 (specifier-type s)))
1001 b. The same for CSUBTYPEP.
1004 * (let () (list (the (values &optional fixnum) (eval '(values)))))
1005 debugger invoked on condition of type TYPE-ERROR:
1006 The value NIL is not of type FIXNUM.
1008 262: "yet another bug in inline expansion of local functions"
1012 (declare (integer x y))
1015 (declare (integer u))
1016 (if (> (1+ (the unsigned-byte u)) 0)
1018 (return (+ 38 (cos (/ u 78)))))))
1019 (declare (inline xyz))
1021 (* (funcall (eval #'xyz) x)
1023 (funcall (if (> x 5) #'xyz #'identity)
1028 Urgh... It's time to write IR1-copier.
1031 SB-EXT:RUN-PROGRAM is currently non-functional on Linux/PPC;
1032 attempting to use it leads to segmentation violations. This is
1033 probably because of a bogus implementation of
1034 os_restore_fp_control().
1037 David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
1038 behaviour of streams with element-type (SIGNED-BYTE 8). The patch
1039 looks reasonable, if not obviously correct; however, it caused the
1040 PPC/Linux port to segfault during warm-init while loading
1041 src/pcl/std-class.fasl. A workaround patch was made, but it would
1042 be nice to understand why the first patch caused problems, and to
1043 fix the cause if possible.
1045 268: "wrong free declaration scope"
1046 The following code must signal type error:
1048 (locally (declare (optimize (safety 3)))
1049 (flet ((foo (x &optional (y (car x)))
1050 (declare (optimize (safety 0)))
1052 (funcall (eval #'foo) 1)))
1055 SCALE-FLOAT should accept any integer for its second argument.
1058 In the following function constraint propagator optimizes nothing:
1061 (declare (integer x))
1062 (declare (optimize speed))
1070 All forms of GC hooks (including notifiers and finalisers) are currently
1071 (since 0.8.0) broken for gencgc (i.e. x86) users
1074 Compilation of the following two forms causes "X is unbound" error:
1076 (symbol-macrolet ((x pi))
1077 (macrolet ((foo (y) (+ x y)))
1078 (declaim (inline bar))
1084 (See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
1087 CLHS says that type declaration of a symbol macro should not affect
1088 its expansion, but in SBCL it does.
1091 The following code (taken from CLOCC) takes a lot of time to compile:
1094 (declare (type (integer 0 #.large-constant) n))
1097 (fixed in 0.8.2.51, but a test case would be good)
1100 (defmethod fee ((x fixnum))
1103 (fee 1) => type error
1108 IGNORE/IGNORABLE declarations should be acceptable for symbol
1114 (declare (optimize speed))
1115 (loop for i of-type (integer 0) from 0 by 2 below 10
1118 uses generic arithmetic.
1120 b. For the example above, the compiler does not issue a note.
1121 (fixed in 0.8.3.6, but a test case would be good)
1123 279: type propagation error -- correctly inferred type goes astray?
1124 In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
1125 The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
1126 &OPTIONAL), not a (INTEGER 1 536870911)
1127 is emitted when compiling this file:
1128 (declaim (ftype (function ((integer 0 #.most-positive-fixnum))
1129 (integer #.most-negative-fixnum 0))
1134 (let* (;; Uncomment this for a type mismatch warning indicating
1135 ;; that the type of (FOO X) is correctly understood.
1136 #+nil (fs-foo (float-sign (foo x)))
1137 ;; Uncomment this for a type mismatch warning
1138 ;; indicating that the type of (ABS (FOO X)) is
1139 ;; correctly understood.
1140 #+nil (fs-abs-foo (float-sign (abs (foo x))))
1141 ;; something wrong with this one though
1142 (abs-foo (abs (foo x))))
1143 (declare (type (integer 1 100) abs-foo))
1148 280: bogus WARNING about duplicate function definition
1149 In sbcl-0.8.3 and sbcl-0.8.1.47, if BS.MIN is defined inline,
1151 (declaim (inline bs.min))
1152 (defun bs.min (bases) nil)
1153 before compiling the file below, the compiler warns
1154 Duplicate definition for BS.MIN found in one static
1155 unit (usually a file).
1157 (declaim (special *minus* *plus* *stagnant*))
1158 (defun b.*.min (&optional (x () xp) (y () yp) &rest rest)
1160 (define-compiler-macro b.*.min (&rest rest)
1162 (defun afish-d-rbd (pd)
1164 (b.*.min (foo-d-rbd *stagnant*))
1165 (multiple-value-bind (reduce-fn initial-value)
1167 (list (values #'bs.min 0))
1168 (vector (values #'bs.min *plus*)))
1169 (let ((cv-ks (cv (kpd.ks pd))))
1170 (funcall reduce-fn d-rbds)))))
1171 (defun bfish-d-rbd (pd)
1173 (b.*.min (foo-d-rbd *stagnant*))
1174 (multiple-value-bind (reduce-fn initial-value)
1176 (list (values #'bs.min *minus*))
1177 (vector (values #'bs.min 0)))
1178 (let ((cv-ks (cv (kpd.ks pd))))
1179 (funcall reduce-fn d-rbds)))))
1181 281: COMPUTE-EFFECTIVE-METHOD error signalling.
1182 (slightly obscured by a non-0 default value for
1183 SB-PCL::*MAX-EMF-PRECOMPUTE-METHODS*)
1184 It would be natural for COMPUTE-EFFECTIVE-METHOD to signal errors
1185 when it finds a method with invalid qualifiers. However, it
1186 shouldn't signal errors when any such methods are not applicable to
1187 the particular call being evaluated, and certainly it shouldn't when
1188 simply precomputing effective methods that may never be called.
1189 (setf sb-pcl::*max-emf-precompute-methods* 0)
1191 (:method-combination +)
1192 (:method ((x symbol)) 1)
1193 (:method + ((x number)) x))
1194 (foo 1) -> ERROR, but should simply return 1
1196 The issue seems to be that construction of a discriminating function
1197 calls COMPUTE-EFFECTIVE-METHOD with methods that are not all applicable.
1199 DEFUNCT CATEGORIES OF BUGS
1201 These labels were used for bugs related to the old IR1 interpreter.
1202 The # values reached 6 before the category was closed down.