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.
46 ANSI specifies that a type mismatch in a structure slot
47 initialization value should not cause a warning.
49 This one might not be fixed for a while because while we're big
50 believers in ANSI compatibility and all, (1) there's no obvious
51 simple way to do it (short of disabling all warnings for type
52 mismatches everywhere), and (2) there's a good portable
53 workaround, and (3) by their own reasoning, it looks as though
54 ANSI may have gotten it wrong. ANSI justifies this specification
56 The restriction against issuing a warning for type mismatches
57 between a slot-initform and the corresponding slot's :TYPE
58 option is necessary because a slot-initform must be specified
59 in order to specify slot options; in some cases, no suitable
61 However, in SBCL (as in CMU CL or, for that matter, any compiler
62 which really understands Common Lisp types) a suitable default
63 does exist, in all cases, because the compiler understands the
64 concept of functions which never return (i.e. has return type NIL).
65 Thus, as a portable workaround, you can use a call to some
66 known-never-to-return function as the default. E.g.
68 (BAR (ERROR "missing :BAR argument")
69 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
71 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
72 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
73 (ERROR "missing required argument"))
75 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
76 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
77 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
78 Such code should compile without complaint and work correctly either
79 on SBCL or on any other completely compliant Common Lisp system.
82 bogus warnings about undefined functions for magic functions like
83 SB!C::%%DEFUN and SB!C::%DEFCONSTANT when cross-compiling files
84 like src/code/float.lisp. Fixing this will probably require
85 straightening out enough bootstrap consistency issues that
86 the cross-compiler can run with *TYPE-SYSTEM-INITIALIZED*.
87 Instead, the cross-compiler runs in a slightly flaky state
88 which is sane enough to compile SBCL itself, but which is
89 also unstable in several ways, including its inability
90 to really grok function declarations.
92 As of sbcl-0.7.5, sbcl's cross-compiler does run with
93 *TYPE-SYSTEM-INITIALIZED*; however, this bug remains.
96 The "compiling top-level form:" output ought to be condensed.
97 Perhaps any number of such consecutive lines ought to turn into a
98 single "compiling top-level forms:" line.
101 The way that the compiler munges types with arguments together
102 with types with no arguments (in e.g. TYPE-EXPAND) leads to
103 weirdness visible to the user:
104 (DEFTYPE FOO () 'FIXNUM)
106 (TYPEP 11 '(FOO)) => T, which seems weird
107 (TYPEP 11 'FIXNUM) => T
108 (TYPEP 11 '(FIXNUM)) signals an error, as it should
109 The situation is complicated by the presence of Common Lisp types
110 like UNSIGNED-BYTE (which can either be used in list form or alone)
111 so I'm not 100% sure that the behavior above is actually illegal.
112 But I'm 90+% sure, and the following related behavior,
114 treating the bare symbol AND as equivalent to '(AND), is specifically
115 forbidden (by the ANSI specification of the AND type).
118 It would be nice if the
120 (during macroexpansion)
121 said what macroexpansion was at fault, e.g.
123 (during macroexpansion of IN-PACKAGE,
124 during macroexpansion of DEFFOO)
127 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
128 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
129 (Also, when this is fixed, we can enable the code in PROCLAIM which
130 checks for incompatible FTYPE redeclarations.)
133 (I *think* this is a bug. It certainly seems like strange behavior. But
134 the ANSI spec is scary, dark, and deep.. -- WHN)
135 (FORMAT NIL "~,1G" 1.4) => "1. "
136 (FORMAT NIL "~3,1G" 1.4) => "1. "
139 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
141 (setf (find-class 'ccc1) (find-class 'ccc))
142 (defmethod zut ((c ccc1)) 123)
143 In sbcl-0.7.1.13, this gives an error,
144 There is no class named CCC1.
145 DTC's recommended workaround from the mailing list 3 Mar 2000:
146 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
149 Sometimes (SB-EXT:QUIT) fails with
150 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
151 Process inferior-lisp exited abnormally with code 1
152 I haven't noticed a repeatable case of this yet.
155 The printer doesn't report closures very well. This is true in
159 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
160 It would be nice to make closures have a settable name slot,
161 and make things like DEFSTRUCT and FLET, which create closures,
162 set helpful values into this slot.
165 And as long as we're wishing, it would be awfully nice if INSPECT could
166 also report on closures, telling about the values of the bound variables.
169 The compiler assumes that any time a function of declared FTYPE
170 doesn't signal an error, its arguments were of the declared type.
171 E.g. compiling and loading
172 (DECLAIM (OPTIMIZE (SAFETY 3)))
173 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
175 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
177 (COND ((> (FACTORIAL X) 1.0E6)
178 (FORMAT T "too big~%"))
180 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
182 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
185 will cause the INTEGERP case to be selected, giving bogus output a la
187 (or (FOO 1000.5), "exactly 1001.5")
188 This violates the "declarations are assertions" principle.
189 According to the ANSI spec, in the section "System Class FUNCTION",
190 this is a case of "lying to the compiler", but the lying is done
191 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
192 not by the unexpectedly general definition of FACTORIAL. In any case,
193 "declarations are assertions" means that lying to the compiler should
194 cause an error to be signalled, and should not cause a bogus
195 result to be returned. Thus, the compiler should not assume
196 that arbitrary functions check their argument types. (It might
197 make sense to add another flag (CHECKED?) to DEFKNOWN to
198 identify functions which *do* check their argument types.)
199 (Also, verify that the compiler handles declared function
200 return types as assertions.)
203 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
204 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
206 (DECLARE (TYPE INDEXOID X))
207 (THE (VALUES INDEXOID)
209 where the implementation of the type check in function FOO
210 includes a full call to %TYPEP. There are also some fundamental problems
211 with the interpretation of VALUES types (inherited from CMU CL, and
212 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
213 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
216 The definitions of SIGCONTEXT-FLOAT-REGISTER and
217 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
218 supported on FreeBSD because the floating point state is not saved,
219 but at least as of FreeBSD 4.0, the floating point state *is* saved,
220 so they could be supported after all. Very likely
221 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
224 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
225 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
226 issue with 'union, 'and, 'or etc. So even though according to the
227 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
228 (and now SBCL) interpret them as legal types.
231 a slew of floating-point-related errors reported by Peter Van Eynde
233 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
234 should probably be 1.4012985e-45. In SBCL,
235 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
236 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
237 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
238 and LEAST-NEGATIVE-LONG-FLOAT.
239 c: Many expressions generate floating infinity on x86/Linux:
244 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
245 on x86/Linux generates the infinities instead. That might or
246 might not be conforming behavior, but it's also inconsistent,
247 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
248 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
249 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
251 d: (in section12.erg) various forms a la
252 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
253 don't give the right behavior.
256 type safety errors reported by Peter Van Eynde July 25, 2000:
257 a: (COERCE (QUOTE (A B C)) (QUOTE (VECTOR * 4)))
259 In general lengths of array type specifications aren't
260 checked by COERCE, so it fails when the spec is
261 (VECTOR 4), (STRING 2), (SIMPLE-BIT-VECTOR 3), or whatever.
262 b: CONCATENATE has the same problem of not checking the length
263 of specified output array types. MAKE-SEQUENCE and MAP and
264 MERGE also have the same problem.
265 c: (COERCE 'AND 'FUNCTION) returns something related to
266 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
267 h: (MAKE-CONCATENATED-STREAM (MAKE-STRING-OUTPUT-STREAM))
268 should signal TYPE-ERROR.
269 i: MAKE-TWO-WAY-STREAM doesn't check that its arguments can
270 be used for input and output as needed. It should fail with
271 TYPE-ERROR when handed e.g. the results of
272 MAKE-STRING-INPUT-STREAM or MAKE-STRING-OUTPUT-STREAM in
273 the inappropriate positions, but doesn't.
274 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
275 not a binary input stream, but instead cheerfully reads from
276 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
279 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
280 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
281 causes a COMPILER-ERROR.
284 SYMBOL-MACROLET bugs reported by Peter Van Eynde July 25, 2000:
285 c: SYMBOL-MACROLET should signal PROGRAM-ERROR if something
286 it binds is declared SPECIAL inside.
289 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
291 (DEFGENERIC FOO02 (X))
292 (DEFMETHOD FOO02 ((X NUMBER)) T)
293 (LET ((M (FIND-METHOD (FUNCTION FOO02)
295 (LIST (FIND-CLASS (QUOTE NUMBER))))))
296 (REMOVE-METHOD (FUNCTION FOO02) M)
297 (DEFGENERIC FOO03 (X))
298 (ADD-METHOD (FUNCTION FOO03) M)))
299 should give an error, but SBCL allows it.
300 b: READ should probably return READER-ERROR, not the bare
301 arithmetic error, when input a la "1/0" or "1e1000" causes
305 It has been reported (e.g. by Peter Van Eynde) that there are
306 several metaobject protocol "errors". (In order to fix them, we might
307 need to document exactly what metaobject protocol specification
308 we're following -- the current code is just inherited from PCL.)
311 The implementation of #'+ returns its single argument without
312 type checking, e.g. (+ "illegal") => "illegal".
315 The debugger LIST-LOCATIONS command doesn't work properly.
318 Compiling and loading
319 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
321 then requesting a BACKTRACE at the debugger prompt gives no information
322 about where in the user program the problem occurred.
325 The compiler is supposed to do type inference well enough that
328 ((SIMPLE-ARRAY SINGLE-FLOAT)
330 (DECLARE (TYPE (SIMPLE-ARRAY SINGLE-FLOAT) X))
333 is redundant. However, as reported by Juan Jose Garcia Ripoll for
334 CMU CL, it sometimes doesn't. Adding declarations is a pretty good
335 workaround for the problem for now, but can't be done by the TYPECASE
336 macros themselves, since it's too hard for the macro to detect
337 assignments to the variable within the clause.
338 Note: The compiler *is* smart enough to do the type inference in
339 many cases. This case, derived from a couple of MACROEXPAND-1
340 calls on Ripoll's original test case,
342 (DECLARE (OPTIMIZE SPEED (SAFETY 0)))
343 (COND ((TYPEP A '(SIMPLE-ARRAY SINGLE-FLOAT)) NIL
344 (LET ((LENGTH (ARRAY-TOTAL-SIZE A)))
345 (LET ((I 0) (G2554 LENGTH))
346 (DECLARE (TYPE REAL G2554) (TYPE REAL I))
349 (WHEN (>= I G2554) (GO SB-LOOP::END-LOOP))
350 (SETF (ROW-MAJOR-AREF A I) (- (ROW-MAJOR-AREF A I)))
351 (GO SB-LOOP::NEXT-LOOP)
352 SB-LOOP::END-LOOP))))))
353 demonstrates the problem; but the problem goes away if the TAGBODY
354 and GO forms are removed (leaving the SETF in ordinary, non-looping
355 code), or if the TAGBODY and GO forms are retained, but the
356 assigned value becomes 0.0 instead of (- (ROW-MAJOR-AREF A I)).
359 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
360 I am looking into this problem that showed up on the cmucl-help
361 list. It seems to me that the "implementation specific environment
362 hacking functions" found in pcl/walker.lisp are completely messed
363 up. The good thing is that they appear to be barely used within
364 PCL and the munged environment object is passed to cmucl only
365 in calls to macroexpand-1, which is probably why this case fails.
366 SBCL uses essentially the same code, so if the environment hacking
367 is screwed up, it affects us too.
370 Using the pretty-printer from the command prompt gives funny
371 results, apparently because the pretty-printer doesn't know
372 about user's command input, including the user's carriage return
373 that the user, and therefore the pretty-printer thinks that
374 the new output block should start indented 2 or more characters
375 rightward of the correct location.
378 ANSI specifies that the RESULT-TYPE argument of CONCATENATE must be
379 a subtype of SEQUENCE, but CONCATENATE doesn't check this properly:
380 (CONCATENATE 'SIMPLE-ARRAY #(1 2) '(3)) => #(1 2 3)
381 This also leads to funny behavior when derived type specifiers
382 are used, as originally reported by Milan Zamazal for CMU CL (on the
383 Debian bugs mailing list (?) 2000-02-27), then reported by Martin
384 Atzmueller for SBCL (2000-10-01 on sbcl-devel@lists.sourceforge.net):
385 (DEFTYPE FOO () 'SIMPLE-ARRAY)
386 (CONCATENATE 'FOO #(1 2) '(3))
387 => #<ARRAY-TYPE SIMPLE-ARRAY> is a bad type specifier for
389 The derived type specifier FOO should act the same way as the
390 built-in type SIMPLE-ARRAY here, but it doesn't. That problem
391 doesn't seem to exist for sequence types:
392 (DEFTYPE BAR () 'SIMPLE-VECTOR)
393 (CONCATENATE 'BAR #(1 2) '(3)) => #(1 2 3)
394 See also bug #46a./b., and discussion and patch sbcl-devel and
398 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
399 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
400 crashes SBCL. In general tracing anything which is used in the
401 implementation of TRACE is likely to have the same problem.
404 (DECLAIM (OPTIMIZE ..)) doesn't work properly inside LOCALLY forms.
407 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
408 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
409 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
410 WITH-OUTPUT-TO-STRING.
413 ANSI says in one place that type declarations can be abbreviated even
414 when the type name is not a symbol, e.g.
415 (DECLAIM ((VECTOR T) *FOOVECTOR*))
416 SBCL doesn't support this. But ANSI says in another place that this
417 isn't allowed. So it's not clear this is a bug after all. (See the
418 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
422 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
423 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
424 an easily guessable temporary filename in a way which might open
425 applications using LOAD-FOREIGN to hijacking by malicious users
426 on the same machine. Incantations for doing this safely are
427 floating around the net in various "how to write secure programs
428 despite Unix" documents, and it would be good to (1) fix this in
429 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
430 files and make it share the same new safe logic.
433 Functions are assigned names based on the context in which they're
434 defined. This is less than ideal for the functions which are
435 used to implement CLOS methods. E.g. the output of
436 (DESCRIBE 'PRINT-OBJECT) lists functions like
437 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
439 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
440 It would be better if these functions' names always identified
441 them as methods, and identified their generic functions and
445 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
446 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
447 the mean of the distribution can be systematically O(0.1%) wrong.
448 Just increasing R-I-E-B is probably not a good solution, since
449 it would decrease efficiency more than is probably necessary. Perhaps
450 using some sort of accept/reject method would be better.
453 Internally the compiler sometimes evaluates
454 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
455 (I stumbled across this when I added an
456 (assert (not (eq type1 *wild-type*)))
457 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
458 in a type context should probably be translated to T, and so it's
459 probably wrong to ask whether it's equal to the T type and then (using
460 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
461 (I haven't tried to investigate this bug enough to guess whether
462 there might be any user-level symptoms.)
465 Inconsistencies between derived and declared VALUES return types for
466 DEFUN aren't checked very well. E.g. the logic which successfully
467 catches problems like
468 (declaim (ftype (function (fixnum) float) foo))
470 (declare (type integer x))
471 (values x)) ; wrong return type, detected, gives warning, good!
473 (declaim (ftype (function (t) (values t t)) bar))
475 (values x)) ; wrong number of return values, no warning, bad!
476 The cause of this is seems to be that (1) the internal function
477 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
478 arguments symmetrically, and (2) when the type checking code was
479 written back when when SBCL's code was still CMU CL, the intent
481 (declaim (ftype (function (t) t) bar))
483 (values x x)) ; wrong number of return values; should give warning?
484 not be warned for, because a two-valued return value is considered
485 to be compatible with callers who expects a single value to be
486 returned. That intent is probably not appropriate for modern ANSI
487 Common Lisp, but fixing this might be complicated because of other
488 divergences between auld-style and new-style handling of
489 multiple-VALUES types. (Some issues related to this were discussed
490 on cmucl-imp at some length sometime in 2000.)
493 The facility for dumping a running Lisp image to disk gets confused
494 when run without the PURIFY option, and creates an unnecessarily large
495 core file (apparently representing memory usage up to the previous
496 high-water mark). Moreover, when the file is loaded, it confuses the
497 GC, so that thereafter memory usage can never be reduced below that
501 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
502 CL), out-of-line structure slot setters are horribly inefficient
503 whenever the type of the slot is declared, because out-of-line
504 structure slot setters are implemented as closures to save space,
505 so the compiler doesn't compile the type test into code, but
506 instead just saves the type in a lexical closure and interprets it
508 A proper solution involves deciding whether it's really worth
509 saving space by implementing structure slot accessors as closures.
510 (If it's not worth it, the problem vanishes automatically. If it
511 is worth it, there are hacks we could use to force type tests to
512 be compiled anyway, and even shared. E.g. we could implement
513 an EQUAL hash table mapping from types to compiled type tests,
514 and save the appropriate compiled type test as part of each lexical
515 closure; or we could make the lexical closures be placeholders
516 which overwrite their old definition as a lexical closure with
517 a new compiled definition the first time that they're called.)
518 As a workaround for the problem, #'(SETF FOO) expressions can
519 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
520 (defmacro efficient-setf-function (place-function-name)
521 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
522 ;; a workaround for the problem, encouraging the
523 ;; inline expansion of the structure accessor, so
524 ;; that the compiler can optimize its type test
525 (let ((new-value (gensym "NEW-VALUE-"))
526 (structure-value (gensym "STRUCTURE-VALUE-")))
527 `(lambda (,new-value ,structure-value)
528 (setf (,place-function-name ,structure-value)
530 ;; no problem, can just use the ordinary expansion
531 `(function (setf ,place-function-name))))
534 There's apparently a bug in CEILING optimization which caused
535 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
536 applied the patches to SBCL and they didn't seem to cause problems
537 (as reported sbcl-devel 2001-05-04). However, since the patches
538 modify nontrivial code which was apparently written incorrectly
539 the first time around, until regression tests are written I'm not
540 comfortable merging the patches in the CVS version of SBCL.
543 (DESCRIBE 'SB-ALIEN:DEF-ALIEN-TYPE) reports the macro argument list
547 in #<PACKAGE "SB-ALIEN">.
548 Macro-function: #<FUNCTION "DEF!MACRO DEF-ALIEN-TYPE" {19F4A39}>
549 Macro arguments: (#:whole-470 #:environment-471)
550 On Sat, May 26, 2001 09:45:57 AM CDT it was compiled from:
551 /usr/stuff/sbcl/src/code/host-alieneval.lisp
552 Created: Monday, March 12, 2001 07:47:43 AM CST
555 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
556 a clean, just-started SBCL system. And it seems to be right:
557 (ROOM T) can bring a small computer to its knees for a *long*
558 time trying to GC afterwards. Surely there's some more economical
559 way to implement (ROOM T).
562 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
564 ;;; The compiler is flushing the argument type test, and the default
565 ;;; case in the cond, so that calling with say a fixnum 0 causes a
567 (declaim (optimize (safety 2) (speed 3)))
569 (declare (type (or string stream) x))
570 (cond ((typep x 'string) 'string)
571 ((typep x 'stream) 'stream)
574 The symptom in sbcl-0.6.12.42 on OpenBSD is actually (TST 0)=>STREAM
575 (not the SIGBUS reported in the comment) but that's broken too;
576 type declarations are supposed to be treated as assertions unless
577 SAFETY 0, so we should be getting a TYPE-ERROR.
580 reported by Martin Atzmueller 2001-06-25; originally from CMU CL bugs
582 (in-package :cl-user)
583 ;;; The following invokes a compiler error.
584 (declaim (optimize (speed 2) (debug 3)))
587 (unwind-protect nil)))
591 The error message in sbcl-0.6.12.42 is
592 internal error, failed AVER:
593 "(COMMON-LISP:EQ (SB!C::TN-ENVIRONMENT SB!C:TN) SB!C::TN-ENV)"
596 When the compiler inline expands functions, it may be that different
597 kinds of return values are generated from different code branches.
598 E.g. an inline expansion of POSITION generates integer results
599 from one branch, and NIL results from another. When that inline
600 expansion is used in a context where only one of those results
603 (aref *a1* (position x *a2*)))
604 and the compiler can't prove that the unacceptable branch is
605 never taken, then bogus type mismatch warnings can be generated.
606 If you need to suppress the type mismatch warnings, you can
607 suppress the inline expansion,
609 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
610 (aref *a1* (position x *a2*)))
611 or, sometimes, suppress them by declaring the result to be of an
614 (aref *a1* (the integer (position x *a2*))))
616 This is not a new compiler problem in 0.7.0, but the new compiler
617 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
618 more conspicuous. If you don't need performance from these functions,
619 and the bogus warnings are a nuisance for you, you can return to
620 your pre-0.7.0 state of grace with
621 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
624 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
625 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
626 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
627 when of course it should be NIL. (He says it only fails for X86,
628 not SPARC; dunno about Alpha.)
630 Also, "the same problem exists for LONG-FLOAT-EPSILON,
631 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
632 for the -negative- the + is replaced by a - in the test)."
634 Raymond Toy comments that this is tricky on the X86 since its FPU
635 uses 80-bit precision internally.
638 Even in sbcl-0.pre7.x, which is supposed to be free of the old
639 non-ANSI behavior of treating the function return type inferred
640 from the current function definition as a declaration of the
641 return type from any function of that name, the return type of NIL
642 is attached to FOO in 120a above, and used to optimize code which
646 There was some sort of screwup in handling of
647 (IF (NOT (IGNORE-ERRORS ..))). E.g.
649 (if (not (ignore-errors
650 (make-pathname :host "foo"
654 (error "notunlessnot")))
655 The (NOT (IGNORE-ERRORS ..)) form evaluates to T, so this should be
656 printing "ok", but instead it's going to the ERROR. This problem
657 seems to've been introduced by MNA's HANDLER-CASE patch (sbcl-devel
658 2001-07-17) and as a workaround (put in sbcl-0.pre7.14.flaky4.12)
659 I reverted back to the old weird HANDLER-CASE code. However, I
660 think the problem looks like a compiler bug in handling RETURN-FROM,
661 so I left the MNA-patched code in HANDLER-CASE (suppressed with
662 #+NIL) and I'd like to go back to see whether this really is
663 a compiler bug before I delete this BUGS entry.
666 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
667 the entire lexical environment at the point of MACROLET available
668 in the bodies of the macroexpander functions. In particular, it
669 allows the function bodies (which run at compile time) to try to
670 access lexical variables (which are only defined at runtime).
671 It doesn't even issue a warning, which is bad.
673 The SBCL behavior arguably conforms to the ANSI spec (since the
674 spec says that the behavior is undefined, ergo anything conforms).
675 However, it would be better to issue a compile-time error.
676 Unfortunately I (WHN) don't see any simple way to detect this
677 condition in order to issue such an error, so for the meantime
678 SBCL just does this weird broken "conforming" thing.
680 The ANSI standard says, in the definition of the special operator
682 The macro-expansion functions defined by MACROLET are defined
683 in the lexical environment in which the MACROLET form appears.
684 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
685 the local macro definitions in a MACROLET, but the consequences
686 are undefined if the local macro definitions reference any
687 local variable or function bindings that are visible in that
689 Then it seems to contradict itself by giving the example
691 (macrolet ((fudge (z)
692 ;The parameters x and flag are not accessible
693 ; at this point; a reference to flag would be to
694 ; the global variable of that name.
695 ` (if flag (* ,z ,z) ,z)))
696 ;The parameters x and flag are accessible here.
700 The comment "a reference to flag would be to the global variable
701 of the same name" sounds like good behavior for the system to have.
702 but actual specification quoted above says that the actual behavior
706 (as reported by Gabe Garza on cmucl-help 2001-09-21)
708 (defun test-pred (x y)
712 (func (lambda () x)))
713 (print (eq func func))
714 (print (test-pred func func))
715 (delete func (list func))))
716 Now calling (TEST-CASE) gives output
719 (#<FUNCTION {500A9EF9}>)
720 Evidently Python thinks of the lambda as a code transformation so
721 much that it forgets that it's also an object.
724 The DEFSTRUCT section of the ANSI spec, in the :CONC-NAME section,
725 specifies a precedence rule for name collisions between slot accessors of
726 structure classes related by inheritance. As of 0.7.0, SBCL still
730 insufficient syntax checking in MACROLET:
732 (macrolet ((defmacro bar (z) `(+ z z)))
734 shouldn't compile without error (because of the extra DEFMACRO symbol).
737 As of sbcl-0.pre7.86.flaky7.3, the cross-compiler, and probably
738 the CL:COMPILE function (which is based on the same %COMPILE
739 mechanism) get confused by
741 (labels ((sxhash-number (x)
743 (fixnum (sxhash x)) ; through DEFTRANSFORM
744 (integer (sb!bignum:sxhash-bignum x))
745 (single-float (sxhash x)) ; through DEFTRANSFORM
746 (double-float (sxhash x)) ; through DEFTRANSFORM
747 #!+long-float (long-float (error "stub: no LONG-FLOAT"))
748 (ratio (let ((result 127810327))
749 (declare (type fixnum result))
750 (mixf result (sxhash-number (numerator x)))
751 (mixf result (sxhash-number (denominator x)))
753 (complex (let ((result 535698211))
754 (declare (type fixnum result))
755 (mixf result (sxhash-number (realpart x)))
756 (mixf result (sxhash-number (imagpart x)))
758 (sxhash-recurse (x &optional (depthoid +max-hash-depthoid+))
759 (declare (type index depthoid))
763 (mix (sxhash-recurse (car x) (1- depthoid))
764 (sxhash-recurse (cdr x) (1- depthoid)))
767 (if (typep x 'structure-object)
769 (sxhash ; through DEFTRANSFORM
770 (class-name (layout-class (%instance-layout x)))))
772 (symbol (sxhash x)) ; through DEFTRANSFORM
773 (number (sxhash-number x))
776 (simple-string (sxhash x)) ; through DEFTRANSFORM
777 (string (%sxhash-substring x))
778 (bit-vector (let ((result 410823708))
779 (declare (type fixnum result))
780 (dotimes (i (min depthoid (length x)))
781 (mixf result (aref x i)))
783 (t (logxor 191020317 (sxhash (array-rank x))))))
786 (sxhash (char-code x)))) ; through DEFTRANSFORM
789 complaining "function called with two arguments, but wants exactly
790 one" about SXHASH-RECURSE. (This might not be strictly a new bug,
791 since IIRC post-fork CMU CL has also had problems with &OPTIONAL
792 arguments in FLET/LABELS: it might be an old Python bug which is
793 only exercised by the new arrangement of the SBCL compiler.)
796 Ideally, uninterning a symbol would allow it, and its associated
797 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
798 at least as of sbcl-0.7.0, this isn't the case. Information about
799 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
800 essentially in ordinary (non-weak) hash tables keyed by symbols.
801 Thus, once a system has an entry in this system, it tends to live
802 forever, even when it is uninterned and all other references to it
806 (reported by Arnaud Rouanet on cmucl-imp 2001-12-18)
807 (defmethod foo ((x integer))
809 (defmethod foo :around ((x integer))
812 Now (FOO 3) should return 3, but instead it returns 4.
815 (reported by Alexey Dejneka sbcl-devel 2002-01-03)
817 SUBTYPEP does not work well with redefined classes:
826 * (defclass b (a) ())
839 This bug was fixed in sbcl-0.7.4.1 by invalidating the PCL wrapper
840 class upon redefinition. Unfortunately, doing so causes bug #176 to
841 appear. Pending further investication, one or other of these bugs
842 might be present at any given time.
845 Pretty-printing nested backquotes doesn't work right, as
846 reported by Alexey Dejneka sbcl-devel 2002-01-13:
848 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
849 * (lisp-implementation-version)
853 (as reported by Lynn Quam on cmucl-imp ca. 2002-01-16)
854 %NATURALIZE-C-STRING conses a lot, like 16 bytes per byte
855 of the naturalized string. We could probably port the patches
856 from the cmucl-imp mailing list.
859 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
860 prominent SourceForge web/db bug tracking system, which is
861 unfortunately not a reliable way to get a timely response from
862 the SBCL maintainers)
863 In the course of trying to build a test case for an
864 application error, I encountered this behavior:
865 If you start up sbcl, and then lay on CTRL-C for a
866 minute or two, the lisp process will eventually say:
867 %PRIMITIVE HALT called; the party is over.
868 and throw you into the monitor. If I start up lisp,
869 attach to the process with strace, and then do the same
870 (abusive) thing, I get instead:
871 access failure in heap page not marked as write-protected
872 and the monitor again. I don't know enough to have the
873 faintest idea of what is going on here.
874 This is with sbcl 6.12, uname -a reports:
875 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
876 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
877 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
878 it took more than two minutes (but less than five) for me.
881 (This was once known as IR1-4, but it lived on even after the
882 IR1 interpreter went to the big bit bucket in the sky.)
883 The system accepts DECLAIM in most places where DECLARE would be
884 accepted, without even issuing a warning. ANSI allows this, but since
885 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
886 meaning is rather different, and it's unlikely that the user
887 has a good reason for doing DECLAIM not at top level, it would be
888 good to issue a STYLE-WARNING when this happens. A possible
889 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
890 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
891 [This is considered an IR1-interpreter-related bug because until
892 EVAL-WHEN is rewritten, which won't happen until after the IR1
893 interpreter is gone, the system's notion of what's a top-level form
894 and what's not will remain too confused to fix this problem.]
897 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
898 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
899 COMPLEX implementation didn't deal with this, and hasn't been
900 upgraded to do so. (This doesn't seem to be a high priority
901 conformance problem, since seems hard to construct useful code
905 Floating point errors are reported poorly. E.g. on x86 OpenBSD
908 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
909 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
910 No traps are enabled? How can this be?
911 It should be possible to be much more specific (overflow, division
912 by zero, etc.) and of course the "How can this be?" should be fixable.
914 See also bugs #45.c and #183
917 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
918 (in-package :cl-user)
921 (defstruct foo bar bletch)
923 (labels ((kidify1 (kid)
931 (declare (inline kid-frob))
934 (the simple-vector (foo-bar perd)))))
936 debugger invoked on condition of type TYPE-ERROR:
937 The value NIL is not of type SB-C::NODE.
938 The location of this failure has moved around as various related
939 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
940 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
943 (essentially the same problem as a CMU CL bug reported by Martin
944 Cracauer on cmucl-imp 2002-02-19)
945 There is a hole in structure slot type checking. Compiling and LOADing
946 (declaim (optimize safety))
948 (bla 0 :type fixnum))
950 (let ((foo (make-foo)))
951 (setf (foo-bla foo) '(1 . 1))
952 (format t "Is ~a of type ~a a cons? => ~a~%"
954 (type-of (foo-bla foo))
955 (consp (foo-bla foo)))))
957 should signal an error, but in sbcl-0.7.1.21 instead gives the output
958 Is (1 . 1) of type CONS a cons? => NIL
959 without signalling an error.
962 Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
963 UPGRADED-COMPLEX-PART-TYPE should have an optional environment argument.
964 (reported by Alexey Dejneka sbcl-devel 2002-04-12)
967 (reported by Robert E. Brown 2002-04-16)
968 When a function is called with too few arguments, causing the
969 debugger to be entered, the uninitialized slots in the bad call frame
970 seem to cause GCish problems, being interpreted as tagged data even
971 though they're not. In particular, executing ROOM in the
972 debugger at that point causes AVER failures:
975 * (lisp-implementation-version)
981 failed AVER: "(SAP= CURRENT END)"
982 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
983 isn't too surprising since there are many differences in stack
984 implementation and GC conservatism between the X86 and other ports.)
988 (in-package :cl-user)
990 (defmethod permanentize ((uustk uustk))
991 (flet ((frob (hash-table test-for-deletion)
993 (obj-entry.stale? (oe)
994 (destructuring-bind (key . datum) oe
995 (declare (type simple-vector key))
996 (deny0 (void? datum))
997 (some #'stale? key))))
998 (declare (inline frob obj-entry.stale?))
999 (frob (uustk.args-hash->obj-alist uustk)
1001 (frob (uustk.hash->memoized-objs-list uustk)
1004 in sbcl-0.7.3.11 causes an assertion failure,
1007 (AND (NULL (BLOCK-SUCC B))
1008 (NOT (BLOCK-DELETE-P B))
1009 (NOT (EQ B (COMPONENT-HEAD #)))))"
1012 In sbcl-0.7.3.11, compiling the (illegal) code
1013 (in-package :cl-user)
1014 (defmethod prove ((uustk uustk))
1015 (zap ((frob () nil))
1017 gives the (not terribly clear) error message
1019 ; (during macroexpansion of (DEFMETHOD PROVE ...))
1020 ; can't get template for (FROB NIL NIL)
1021 The problem seems to be that the code walker used by the DEFMETHOD
1022 macro is unhappy with the illegal syntax in the method body, and
1023 is giving an unclear error message.
1026 (reported by Dan Barlow on sbcl-devel 2002-05-10)
1027 In sbcl-0.7.3.12, doing
1028 (defstruct foo bar baz)
1029 (compile nil (lambda (x) (or x (foo-baz x))))
1031 debugger invoked on condition of type SB-INT:BUG:
1032 full call to SB-KERNEL:%INSTANCE-REF
1033 This is probably a bug in SBCL itself. [...]
1034 Since this is a reasonable user error, it shouldn't be reported as
1038 sbcl's treatment of at least macro lambda lists is too permissive;
1039 e.g., in sbcl-0.7.3.7:
1040 (defmacro foo (&rest rest bar) `(,bar ,rest))
1041 (macroexpand '(foo quux zot)) -> (QUUX (QUUX ZOT))
1042 whereas section 3.4.4 of the CLHS doesn't allow required parameters
1043 to come after the rest argument.
1046 The compiler sometimes tries to constant-fold expressions before
1047 it checks to see whether they can be reached. This can lead to
1048 bogus warnings about errors in the constant folding, e.g. in code
1051 (WRITE-STRING (> X 0) "+" "0"))
1052 compiled in a context where the compiler can prove that X is NIL,
1053 and the compiler complains that (> X 0) causes a type error because
1054 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
1055 caused a full WARNING, which made the bug really annoying because then
1056 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
1057 code. Since then the warning has been downgraded to STYLE-WARNING,
1058 so it's still a bug but at least it's a little less annoying.
1061 reported by Alexey Dejneka 08 Jun 2002 in sbcl-devel:
1062 Playing with McCLIM, I've received an error "Unbound variable WRAPPER
1063 in SB-PCL::CHECK-WRAPPER-VALIDITY".
1064 (defun check-wrapper-validity (instance)
1065 (let* ((owrapper (wrapper-of instance)))
1066 (if (not (invalid-wrapper-p owrapper))
1068 (let* ((state (wrapper-state wrapper)) ; !!!
1070 I've tried to replace it with OWRAPPER, but now OBSOLETE-INSTANCE-TRAP
1071 breaks with "NIL is not of type SB-KERNEL:LAYOUT".
1073 partial fix: The undefined variable WRAPPER resulted from an error
1074 in recent refactoring, as can be seen by comparing to the code in e.g.
1075 sbcl-0.7.2. Replacing WRAPPER with OWRAPPER (done by WHN in sbcl-0.7.4.22)
1076 should bring the code back to its behavior as of sbcl-0.7.2, but
1077 that still leaves the OBSOLETE-INSTANCE-TRAP bug. An example of
1078 input which triggers that bug is
1080 (let ((lastname (intern (format nil "C~D" (1- i))))
1081 (name (intern (format nil "C~D" i))))
1082 (eval `(defclass ,name
1083 (,@(if (= i 0) nil (list lastname)))
1085 (eval `(defmethod initialize-instance :after ((x ,name) &rest any)
1086 (declare (ignore any))))))
1088 (defclass c0 (b) ())
1089 (make-instance 'c19)
1093 178: "AVER failure compiling confused THEs in FUNCALL"
1094 In sbcl-0.7.4.24, compiling
1096 (funcall (the function (the standard-object x))))
1099 "(AND (EQ (IR2-CONTINUATION-PRIMITIVE-TYPE 2CONT) FUNCTION-PTYPE) (EQ CHECK T))"
1100 This variant compiles OK, though:
1101 (defun bug178alternative (x)
1102 (funcall (the nil x)))
1104 181: "bad type specifier drops compiler into debugger"
1106 (in-package :cl-user)
1108 (declare (type 0 x))
1111 bad thing to be a type specifier: 0
1112 which seems fine, but also enters the debugger (instead of having
1113 the compiler handle the error, convert it into a COMPILER-ERROR, and
1114 continue compiling) which seems wrong.
1116 183: "IEEE floating point issues"
1117 Even where floating point handling is being dealt with relatively
1118 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
1119 accrued-exceptions and current-exceptions part of the fp control
1120 word don't seem to bear much relation to reality. E.g. on
1124 debugger invoked on condition of type DIVISION-BY-ZERO:
1125 arithmetic error DIVISION-BY-ZERO signalled
1126 0] (sb-vm::get-floating-point-modes)
1128 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1129 :ROUNDING-MODE :NEAREST
1130 :CURRENT-EXCEPTIONS NIL
1131 :ACCRUED-EXCEPTIONS (:INEXACT)
1134 * (sb-vm::get-floating-point-modes)
1135 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
1136 :ROUNDING-MODE :NEAREST
1137 :CURRENT-EXCEPTIONS (:INEXACT)
1138 :ACCRUED-EXCEPTIONS (:INEXACT)
1141 185: "top-level forms at the REPL"
1142 * (locally (defstruct foo (a 0 :type fixnum)))
1145 ; (in macroexpansion of (SB-KERNEL::%DELAYED-GET-COMPILER-LAYOUT BAR))
1146 however, compiling and loading the same expression in a file works
1149 187: "type inference confusion around DEFTRANSFORM time"
1150 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
1151 bug in DEFTRANSFORM")
1152 After the file below is compiled and loaded in sbcl-0.7.5, executing
1153 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
1154 at the REPL returns an adjustable vector, which is wrong. Presumably
1155 somehow the DERIVE-TYPE information for the output values of %WAD is
1156 being mispropagated as a type constraint on the input values of %WAD,
1157 and so causing the type test to be optimized away. It's unclear how
1158 hand-expanding the DEFTRANSFORM would change this, but it suggests
1159 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
1160 invoked at a particular phase) is involved.
1161 (cl:in-package :sb-c)
1162 (eval-when (:compile-toplevel)
1163 ;;; standin for %DATA-VECTOR-AND-INDEX
1164 (defknown %dvai (array index)
1166 (foldable flushable))
1167 (deftransform %dvai ((array index)
1171 (let* ((atype (continuation-type array))
1172 (eltype (array-type-specialized-element-type atype)))
1173 (when (eq eltype *wild-type*)
1174 (give-up-ir1-transform
1175 "specialized array element type not known at compile-time"))
1176 (when (not (array-type-complexp atype))
1177 (give-up-ir1-transform "SIMPLE array!"))
1178 `(if (array-header-p array)
1179 (%wad array index nil)
1180 (values array index))))
1181 ;;; standin for %WITH-ARRAY-DATA
1182 (defknown %wad (array index (or index null))
1183 (values (simple-array * (*)) index index index)
1184 (foldable flushable))
1185 ;;; (Commenting out this optimizer causes the bug to go away.)
1186 (defoptimizer (%wad derive-type) ((array start end))
1187 (let ((atype (continuation-type array)))
1188 (when (array-type-p atype)
1189 (values-specifier-type
1190 `(values (simple-array ,(type-specifier
1191 (array-type-specialized-element-type atype))
1193 index index index)))))
1195 (defun %wad (array start end)
1196 (format t "~&in %WAD~%")
1197 (%with-array-data array start end))
1198 (cl:in-package :cl-user)
1200 (declare (type (vector t) v))
1201 (declare (notinline sb-kernel::%with-array-data))
1202 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
1206 188: "compiler performance fiasco involving type inference and UNION-TYPE"
1207 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
1208 bug could be fixed properly, so you won't see the bug unless you restore
1209 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
1213 (declare (optimize (safety 3)))
1214 (declare (optimize (compilation-speed 2)))
1215 (declare (optimize (speed 1) (debug 1) (space 1)))
1216 (let ((fn "if-this-file-exists-the-universe-is-strange"))
1217 (load fn :if-does-not-exist nil)
1218 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
1219 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
1220 (load (concatenate 'string fn ".misc-garbage")
1221 :if-does-not-exist nil)))))
1223 134.552 seconds of real time
1224 133.35156 seconds of user run time
1225 0.03125 seconds of system run time
1226 [Run times include 2.787 seconds GC run time.]
1228 246883368 bytes consed.
1229 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
1230 thinking about type relationships involving types like
1232 (OR (INTEGER 576 576)
1243 190: "PPC/Linux pipe? buffer? bug"
1244 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
1245 on the PPC/Linux platform, waiting for a zombie env process. This
1246 is a classic symptom of buffer filling and deadlock, but it seems
1247 only sporadically reproducible.
1249 191: "Miscellaneous PCL deficiencies"
1250 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
1251 a. DEFCLASS does not inform the compiler about generated
1252 functions. Compiling a file with
1253 (DEFCLASS A-CLASS ()
1255 (DEFUN A-CLASS-X (A)
1256 (WITH-SLOTS (A-CLASS-X) A
1258 results in a STYLE-WARNING:
1260 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
1262 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
1263 Mai points out that the declamation of functions is in fact
1264 incorrect in some cases (most notably for structure
1265 classes). This means that at present erroneous attempts to use
1266 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
1267 won't get the corresponding STYLE-WARNING.
1268 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
1269 lists and &KEY arguments) do not signal errors when they should.
1271 192: "Python treats free type declarations as promises."
1272 b. What seemed like the same fundamental problem as bug 192a, but
1273 was not fixed by the same (APD "more strict type checking
1274 sbcl-devel 2002-08-97) patch:
1275 (DOTIMES (I ...) (DOTIMES (J ...) (DECLARE ...) ...)):
1276 (declaim (optimize (speed 1) (safety 3)))
1277 (defun trust-assertion (i)
1279 (declare (type (mod 4) i)) ; when commented out, behavior changes!
1282 (trust-assertion 6) ; prints nothing unless DECLARE is commented out
1284 193: "unhelpful CLOS error reporting when the primary method is missing"
1286 (defmethod foo :before ((x t)) (print x))
1287 is the only method defined on FOO, the error reporting when e.g.
1289 is relatively unhelpful:
1290 There is no primary method for the generic function
1291 #<STANDARD-GENERIC-FUNCTION FOO (1)>.
1292 with the offending argument nowhere visible in the backtrace. This
1293 continues even if there *are* primary methods, just not for the
1294 specified arg type, e.g.
1295 (defmethod foo ((x character)) (print x))
1296 (defmethod foo ((x string)) (print x))
1297 (defmethod foo ((x pathname)) ...)
1298 In that case it could be very helpful to know what argument value is
1299 falling through the cracks of the defined primary methods, but the
1300 error message stays the same (even BACKTRACE doesn't tell you what the
1301 bad argument value is).
1303 194: "no error from (THE REAL '(1 2 3)) in some cases"
1306 (multiple-value-prog1 (progn (the real '(1 2 3))))
1307 returns (1 2 3) instead of signalling an error. This was fixed by
1308 APD's "more strict type checking patch", but although the fixed
1309 code (in sbcl-0.7.7.19) works (signals TYPE-ERROR) interactively,
1310 it's difficult to write a regression test for it, because
1311 (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
1312 still returns (1 2 3).
1314 b. (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
1315 returns (1 2 3). (As above, this shows up when writing regression
1316 tests for fixed-ness of part a.)
1317 c. Also in sbcl-0.7.7.9, (IGNORE-ERRORS (THE REAL '(1 2 3))) => (1 2 3).
1319 (null (ignore-errors
1321 (arg2 (identity (the real #(1 2 3)))))
1322 (if (< arg1 arg2) arg1 arg2))))
1324 but putting the same expression inside (DEFUN FOO () ...),
1327 * Actually this entry is probably multiple bugs, as
1328 Alexey Dejneka commented on sbcl-devel 2002-09-03:)
1329 I don't think that placing these two bugs in one entry is
1330 a good idea: they have different explanations. The second
1331 (min 1 nil) is caused by flushing of unused code--IDENTITY
1332 can do nothing with it. So it is really bug 122. The first
1333 (min nil) is due to M-V-PROG1: substituting a continuation
1334 for the result, it forgets about type assertion. The purpose
1335 of IDENTITY is to save the restricted continuation from
1336 inaccurate transformations.
1337 * Alexey Dejneka pointed out that
1338 (IGNORE-ERRORS (IDENTITY (THE REAL '(1 2 3))))
1339 works as it should. Also
1340 (IGNORE-ERRORS (VALUES (THE REAL '(1 2 3))))
1341 works as it should. Perhaps this is another case of VALUES type
1342 intersections behaving in non-useful ways?
1344 195: "confusing reporting of not-a-REAL TYPE-ERRORs from THE REAL"
1345 In sbcl-0.7.7.10, (THE REAL #(1 2 3)) signals a type error which
1346 prints as "This is not a (OR SINGLE-FLOAT DOUBLE-FLOAT RATIONAL)".
1347 The (OR SINGLE-FLOAT DOUBLE-FLOAT RATIONAL) representation of
1348 REAL is unnecessarily confusing, especially since it relies on
1349 internal implementation knowledge that even with SHORT-FLOAT
1350 and LONG-FLOAT left out of the union, this type is equal to REAL.
1351 So it'd be better just to say "This is not a REAL".
1353 196: "confusing error message for unREAL second arg to ATAN"
1354 (fixed in sbcl-0.7.7.18)
1356 DEFUNCT CATEGORIES OF BUGS
1358 These labels were used for bugs related to the old IR1 interpreter.
1359 The # values reached 6 before the category was closed down.