3 Bugs can be reported on the help mailing list
4 sbcl-help@lists.sourceforge.net
5 or on the development mailing list
6 sbcl-devel@lists.sourceforge.net
8 Please include enough information in a bug report that someone reading
9 it can reproduce the problem, i.e. don't write
10 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
11 PRINT-OBJECT doesn't seem to work with *PRINT-LENGTH*. Is this a bug?
13 Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
14 In sbcl-1.2.3 running under OpenBSD 4.5 on my Alpha box, when
15 I compile and load the file
16 (DEFSTRUCT (FOO (:PRINT-OBJECT (LAMBDA (X Y)
17 (LET ((*PRINT-LENGTH* 4))
20 then at the command line type
22 the program loops endlessly instead of printing the object.
27 There is also some information on bugs in the manual page and
28 in the TODO file. Eventually more such information may move here.
30 The gaps in the number sequence belong to old bug descriptions which
31 have gone away (typically because they were fixed, but sometimes for
32 other reasons, e.g. because they were moved elsewhere).
35 KNOWN BUGS OF NO SPECIAL CLASS:
38 DEFSTRUCT almost certainly should overwrite the old LAYOUT information
39 instead of just punting when a contradictory structure definition
40 is loaded. As it is, if you redefine DEFSTRUCTs in a way which
41 changes their layout, you probably have to rebuild your entire
42 program, even if you know or guess enough about the internals of
43 SBCL to wager that this (undefined in ANSI) operation would be safe.
45 3: "type checking of structure slots"
47 ANSI specifies that a type mismatch in a structure slot
48 initialization value should not cause a warning.
50 This one might not be fixed for a while because while we're big
51 believers in ANSI compatibility and all, (1) there's no obvious
52 simple way to do it (short of disabling all warnings for type
53 mismatches everywhere), and (2) there's a good portable
54 workaround, and (3) by their own reasoning, it looks as though
55 ANSI may have gotten it wrong. ANSI justifies this specification
57 The restriction against issuing a warning for type mismatches
58 between a slot-initform and the corresponding slot's :TYPE
59 option is necessary because a slot-initform must be specified
60 in order to specify slot options; in some cases, no suitable
62 However, in SBCL (as in CMU CL or, for that matter, any compiler
63 which really understands Common Lisp types) a suitable default
64 does exist, in all cases, because the compiler understands the
65 concept of functions which never return (i.e. has return type NIL).
66 Thus, as a portable workaround, you can use a call to some
67 known-never-to-return function as the default. E.g.
69 (BAR (ERROR "missing :BAR argument")
70 :TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
72 (DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
73 (DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
74 (ERROR "missing required argument"))
76 (BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
77 (BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
78 (N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
79 Such code should compile without complaint and work correctly either
80 on SBCL or on any other completely compliant Common Lisp system.
82 b: &AUX argument in a boa-constructor without a default value means
83 "do not initilize this slot" and does not cause type error. But
84 an error may be signalled at read time and it would be good if
87 c: Reading of not initialized slot sometimes causes SEGV.
90 (declaim (optimize (safety 3) (speed 1) (space 1)))
93 (defstruct (stringwise-foo (:include foo
94 (x "x" :type simple-string)
95 (y "y" :type simple-string))))
96 (defparameter *stringwise-foo*
97 (make-stringwise-foo))
98 (setf (foo-x *stringwise-foo*) 0)
99 (defun frob-stringwise-foo (sf)
100 (aref (stringwise-foo-x sf) 0))
101 (frob-stringwise-foo *stringwise-foo*)
105 The "compiling top-level form:" output ought to be condensed.
106 Perhaps any number of such consecutive lines ought to turn into a
107 single "compiling top-level forms:" line.
110 The way that the compiler munges types with arguments together
111 with types with no arguments (in e.g. TYPE-EXPAND) leads to
112 weirdness visible to the user:
113 (DEFTYPE FOO () 'FIXNUM)
115 (TYPEP 11 '(FOO)) => T, which seems weird
116 (TYPEP 11 'FIXNUM) => T
117 (TYPEP 11 '(FIXNUM)) signals an error, as it should
118 The situation is complicated by the presence of Common Lisp types
119 like UNSIGNED-BYTE (which can either be used in list form or alone)
120 so I'm not 100% sure that the behavior above is actually illegal.
121 But I'm 90+% sure, and the following related behavior,
123 treating the bare symbol AND as equivalent to '(AND), is specifically
124 forbidden (by the ANSI specification of the AND type).
127 It would be nice if the
129 (during macroexpansion)
130 said what macroexpansion was at fault, e.g.
132 (during macroexpansion of IN-PACKAGE,
133 during macroexpansion of DEFFOO)
136 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
137 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
138 (Also, when this is fixed, we can enable the code in PROCLAIM which
139 checks for incompatible FTYPE redeclarations.)
142 (I *think* this is a bug. It certainly seems like strange behavior. But
143 the ANSI spec is scary, dark, and deep.. -- WHN)
144 (FORMAT NIL "~,1G" 1.4) => "1. "
145 (FORMAT NIL "~3,1G" 1.4) => "1. "
148 Sometimes (SB-EXT:QUIT) fails with
149 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
150 Process inferior-lisp exited abnormally with code 1
151 I haven't noticed a repeatable case of this yet.
154 The printer doesn't report closures very well. This is true in
158 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
159 It would be nice to make closures have a settable name slot,
160 and make things like DEFSTRUCT and FLET, which create closures,
161 set helpful values into this slot.
164 And as long as we're wishing, it would be awfully nice if INSPECT could
165 also report on closures, telling about the values of the bound variables.
168 The compiler assumes that any time a function of declared FTYPE
169 doesn't signal an error, its arguments were of the declared type.
170 E.g. compiling and loading
171 (DECLAIM (OPTIMIZE (SAFETY 3)))
172 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
174 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
176 (COND ((> (FACTORIAL X) 1.0E6)
177 (FORMAT T "too big~%"))
179 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
181 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
184 will cause the INTEGERP case to be selected, giving bogus output a la
186 This violates the "declarations are assertions" principle.
187 According to the ANSI spec, in the section "System Class FUNCTION",
188 this is a case of "lying to the compiler", but the lying is done
189 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
190 not by the unexpectedly general definition of FACTORIAL. In any case,
191 "declarations are assertions" means that lying to the compiler should
192 cause an error to be signalled, and should not cause a bogus
193 result to be returned. Thus, the compiler should not assume
194 that arbitrary functions check their argument types. (It might
195 make sense to add another flag (CHECKED?) to DEFKNOWN to
196 identify functions which *do* check their argument types.)
197 (Also, verify that the compiler handles declared function
198 return types as assertions.)
201 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
202 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
204 (DECLARE (TYPE INDEXOID X))
205 (THE (VALUES INDEXOID)
207 where the implementation of the type check in function FOO
208 includes a full call to %TYPEP. There are also some fundamental problems
209 with the interpretation of VALUES types (inherited from CMU CL, and
210 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
211 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
214 The definitions of SIGCONTEXT-FLOAT-REGISTER and
215 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
216 supported on FreeBSD because the floating point state is not saved,
217 but at least as of FreeBSD 4.0, the floating point state *is* saved,
218 so they could be supported after all. Very likely
219 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
222 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
223 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
224 issue with 'union, 'and, 'or etc. So even though according to the
225 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
226 (and now SBCL) interpret them as legal types.
229 a slew of floating-point-related errors reported by Peter Van Eynde
231 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT on the x86 is
232 bogus, and should probably be 1.4012985e-45. In SBCL,
233 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
234 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
235 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
236 and LEAST-NEGATIVE-LONG-FLOAT.
237 c: Many expressions generate floating infinity on x86/Linux:
242 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
243 on x86/Linux generates the infinities instead. That might or
244 might not be conforming behavior, but it's also inconsistent,
245 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
246 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
247 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
249 d: (in section12.erg) various forms a la
250 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
251 don't give the right behavior.
254 type safety errors reported by Peter Van Eynde July 25, 2000:
255 c: (COERCE 'AND 'FUNCTION) returns something related to
256 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
257 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
258 not a binary input stream, but instead cheerfully reads from
259 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
262 The debugger LIST-LOCATIONS command doesn't work properly.
265 Compiling and loading
266 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
268 then requesting a BACKTRACE at the debugger prompt gives no information
269 about where in the user program the problem occurred.
272 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
273 I am looking into this problem that showed up on the cmucl-help
274 list. It seems to me that the "implementation specific environment
275 hacking functions" found in pcl/walker.lisp are completely messed
276 up. The good thing is that they appear to be barely used within
277 PCL and the munged environment object is passed to cmucl only
278 in calls to macroexpand-1, which is probably why this case fails.
279 SBCL uses essentially the same code, so if the environment hacking
280 is screwed up, it affects us too.
283 Using the pretty-printer from the command prompt gives funny
284 results, apparently because the pretty-printer doesn't know
285 about user's command input, including the user's carriage return
286 that the user, and therefore the pretty-printer thinks that
287 the new output block should start indented 2 or more characters
288 rightward of the correct location.
291 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
292 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
293 crashes SBCL. In general tracing anything which is used in the
294 implementation of TRACE is likely to have the same problem.
297 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
298 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
299 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
300 WITH-OUTPUT-TO-STRING.
303 ANSI says in one place that type declarations can be abbreviated even
304 when the type name is not a symbol, e.g.
305 (DECLAIM ((VECTOR T) *FOOVECTOR*))
306 SBCL doesn't support this. But ANSI says in another place that this
307 isn't allowed. So it's not clear this is a bug after all. (See the
308 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
312 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
313 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
314 an easily guessable temporary filename in a way which might open
315 applications using LOAD-FOREIGN to hijacking by malicious users
316 on the same machine. Incantations for doing this safely are
317 floating around the net in various "how to write secure programs
318 despite Unix" documents, and it would be good to (1) fix this in
319 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
320 files and make it share the same new safe logic.
322 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
323 make the temporary filename less easily guessable)
326 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
327 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
328 the mean of the distribution can be systematically O(0.1%) wrong.
329 Just increasing R-I-E-B is probably not a good solution, since
330 it would decrease efficiency more than is probably necessary. Perhaps
331 using some sort of accept/reject method would be better.
334 Internally the compiler sometimes evaluates
335 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
336 (I stumbled across this when I added an
337 (assert (not (eq type1 *wild-type*)))
338 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
339 in a type context should probably be translated to T, and so it's
340 probably wrong to ask whether it's equal to the T type and then (using
341 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
342 (I haven't tried to investigate this bug enough to guess whether
343 there might be any user-level symptoms.)
345 In fact, the type system is likely to depend on this inequality not
346 holding... * is not equivalent to T in many cases, such as
347 (VECTOR *) /= (VECTOR T).
350 Inconsistencies between derived and declared VALUES return types for
351 DEFUN aren't checked very well. E.g. the logic which successfully
352 catches problems like
353 (declaim (ftype (function (fixnum) float) foo))
355 (declare (type integer x))
356 (values x)) ; wrong return type, detected, gives warning, good!
358 (declaim (ftype (function (t) (values t t)) bar))
360 (values x)) ; wrong number of return values, no warning, bad!
361 The cause of this is seems to be that (1) the internal function
362 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
363 arguments symmetrically, and (2) when the type checking code was
364 written back when when SBCL's code was still CMU CL, the intent
366 (declaim (ftype (function (t) t) bar))
368 (values x x)) ; wrong number of return values; should give warning?
369 not be warned for, because a two-valued return value is considered
370 to be compatible with callers who expects a single value to be
371 returned. That intent is probably not appropriate for modern ANSI
372 Common Lisp, but fixing this might be complicated because of other
373 divergences between auld-style and new-style handling of
374 multiple-VALUES types. (Some issues related to this were discussed
375 on cmucl-imp at some length sometime in 2000.)
378 The facility for dumping a running Lisp image to disk gets confused
379 when run without the PURIFY option, and creates an unnecessarily large
380 core file (apparently representing memory usage up to the previous
381 high-water mark). Moreover, when the file is loaded, it confuses the
382 GC, so that thereafter memory usage can never be reduced below that
386 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
387 CL), out-of-line structure slot setters are horribly inefficient
388 whenever the type of the slot is declared, because out-of-line
389 structure slot setters are implemented as closures to save space,
390 so the compiler doesn't compile the type test into code, but
391 instead just saves the type in a lexical closure and interprets it
393 A proper solution involves deciding whether it's really worth
394 saving space by implementing structure slot accessors as closures.
395 (If it's not worth it, the problem vanishes automatically. If it
396 is worth it, there are hacks we could use to force type tests to
397 be compiled anyway, and even shared. E.g. we could implement
398 an EQUAL hash table mapping from types to compiled type tests,
399 and save the appropriate compiled type test as part of each lexical
400 closure; or we could make the lexical closures be placeholders
401 which overwrite their old definition as a lexical closure with
402 a new compiled definition the first time that they're called.)
403 As a workaround for the problem, #'(SETF FOO) expressions can
404 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
405 (defmacro efficient-setf-function (place-function-name)
406 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
407 ;; a workaround for the problem, encouraging the
408 ;; inline expansion of the structure accessor, so
409 ;; that the compiler can optimize its type test
410 (let ((new-value (gensym "NEW-VALUE-"))
411 (structure-value (gensym "STRUCTURE-VALUE-")))
412 `(lambda (,new-value ,structure-value)
413 (setf (,place-function-name ,structure-value)
415 ;; no problem, can just use the ordinary expansion
416 `(function (setf ,place-function-name))))
419 There's apparently a bug in CEILING optimization which caused
420 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
421 applied the patches to SBCL and they didn't seem to cause problems
422 (as reported sbcl-devel 2001-05-04). However, since the patches
423 modify nontrivial code which was apparently written incorrectly
424 the first time around, until regression tests are written I'm not
425 comfortable merging the patches in the CVS version of SBCL.
428 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
429 a clean, just-started SBCL system. And it seems to be right:
430 (ROOM T) can bring a small computer to its knees for a *long*
431 time trying to GC afterwards. Surely there's some more economical
432 way to implement (ROOM T).
435 When the compiler inline expands functions, it may be that different
436 kinds of return values are generated from different code branches.
437 E.g. an inline expansion of POSITION generates integer results
438 from one branch, and NIL results from another. When that inline
439 expansion is used in a context where only one of those results
442 (aref *a1* (position x *a2*)))
443 and the compiler can't prove that the unacceptable branch is
444 never taken, then bogus type mismatch warnings can be generated.
445 If you need to suppress the type mismatch warnings, you can
446 suppress the inline expansion,
448 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
449 (aref *a1* (position x *a2*)))
450 or, sometimes, suppress them by declaring the result to be of an
453 (aref *a1* (the integer (position x *a2*))))
455 This is not a new compiler problem in 0.7.0, but the new compiler
456 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
457 more conspicuous. If you don't need performance from these functions,
458 and the bogus warnings are a nuisance for you, you can return to
459 your pre-0.7.0 state of grace with
460 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
463 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
464 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
465 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
466 when of course it should be NIL. (He says it only fails for X86,
467 not SPARC; dunno about Alpha.)
469 Also, "the same problem exists for LONG-FLOAT-EPSILON,
470 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
471 for the -negative- the + is replaced by a - in the test)."
473 Raymond Toy comments that this is tricky on the X86 since its FPU
474 uses 80-bit precision internally.
477 Even in sbcl-0.pre7.x, which is supposed to be free of the old
478 non-ANSI behavior of treating the function return type inferred
479 from the current function definition as a declaration of the
480 return type from any function of that name, the return type of NIL
481 is attached to FOO in 120a above, and used to optimize code which
485 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
486 the entire lexical environment at the point of MACROLET available
487 in the bodies of the macroexpander functions. In particular, it
488 allows the function bodies (which run at compile time) to try to
489 access lexical variables (which are only defined at runtime).
490 It doesn't even issue a warning, which is bad.
492 The SBCL behavior arguably conforms to the ANSI spec (since the
493 spec says that the behavior is undefined, ergo anything conforms).
494 However, it would be better to issue a compile-time error.
495 Unfortunately I (WHN) don't see any simple way to detect this
496 condition in order to issue such an error, so for the meantime
497 SBCL just does this weird broken "conforming" thing.
499 The ANSI standard says, in the definition of the special operator
501 The macro-expansion functions defined by MACROLET are defined
502 in the lexical environment in which the MACROLET form appears.
503 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
504 the local macro definitions in a MACROLET, but the consequences
505 are undefined if the local macro definitions reference any
506 local variable or function bindings that are visible in that
508 Then it seems to contradict itself by giving the example
510 (macrolet ((fudge (z)
511 ;The parameters x and flag are not accessible
512 ; at this point; a reference to flag would be to
513 ; the global variable of that name.
514 ` (if flag (* ,z ,z) ,z)))
515 ;The parameters x and flag are accessible here.
519 The comment "a reference to flag would be to the global variable
520 of the same name" sounds like good behavior for the system to have.
521 but actual specification quoted above says that the actual behavior
524 (Since 0.7.8.23 macroexpanders are defined in a restricted version
525 of the lexical environment, containing no lexical variables and
526 functions, which seems to conform to ANSI and CLtL2, but signalling
527 a STYLE-WARNING for references to variables similar to locals might
531 (as reported by Gabe Garza on cmucl-help 2001-09-21)
533 (defun test-pred (x y)
537 (func (lambda () x)))
538 (print (eq func func))
539 (print (test-pred func func))
540 (delete func (list func))))
541 Now calling (TEST-CASE) gives output
544 (#<FUNCTION {500A9EF9}>)
545 Evidently Python thinks of the lambda as a code transformation so
546 much that it forgets that it's also an object.
549 Ideally, uninterning a symbol would allow it, and its associated
550 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
551 at least as of sbcl-0.7.0, this isn't the case. Information about
552 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
553 essentially in ordinary (non-weak) hash tables keyed by symbols.
554 Thus, once a system has an entry in this system, it tends to live
555 forever, even when it is uninterned and all other references to it
558 141: "pretty printing and backquote"
561 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
564 * (write '`(, .ala.) :readably t :pretty t)
567 (note the space between the comma and the point)
570 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
571 prominent SourceForge web/db bug tracking system, which is
572 unfortunately not a reliable way to get a timely response from
573 the SBCL maintainers)
574 In the course of trying to build a test case for an
575 application error, I encountered this behavior:
576 If you start up sbcl, and then lay on CTRL-C for a
577 minute or two, the lisp process will eventually say:
578 %PRIMITIVE HALT called; the party is over.
579 and throw you into the monitor. If I start up lisp,
580 attach to the process with strace, and then do the same
581 (abusive) thing, I get instead:
582 access failure in heap page not marked as write-protected
583 and the monitor again. I don't know enough to have the
584 faintest idea of what is going on here.
585 This is with sbcl 6.12, uname -a reports:
586 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
587 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
588 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
589 it took more than two minutes (but less than five) for me.
592 (This was once known as IR1-4, but it lived on even after the
593 IR1 interpreter went to the big bit bucket in the sky.)
594 The system accepts DECLAIM in most places where DECLARE would be
595 accepted, without even issuing a warning. ANSI allows this, but since
596 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
597 meaning is rather different, and it's unlikely that the user
598 has a good reason for doing DECLAIM not at top level, it would be
599 good to issue a STYLE-WARNING when this happens. A possible
600 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
601 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
602 [This is considered an IR1-interpreter-related bug because until
603 EVAL-WHEN is rewritten, which won't happen until after the IR1
604 interpreter is gone, the system's notion of what's a top-level form
605 and what's not will remain too confused to fix this problem.]
608 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
609 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
610 COMPLEX implementation didn't deal with this, and hasn't been
611 upgraded to do so. (This doesn't seem to be a high priority
612 conformance problem, since seems hard to construct useful code
616 Floating point errors are reported poorly. E.g. on x86 OpenBSD
619 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
620 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
621 No traps are enabled? How can this be?
622 It should be possible to be much more specific (overflow, division
623 by zero, etc.) and of course the "How can this be?" should be fixable.
625 See also bugs #45.c and #183
628 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
629 (in-package :cl-user)
632 (defstruct foo bar bletch)
634 (labels ((kidify1 (kid)
642 (declare (inline kid-frob))
645 (the simple-vector (foo-bar perd)))))
647 debugger invoked on condition of type TYPE-ERROR:
648 The value NIL is not of type SB-C::NODE.
649 The location of this failure has moved around as various related
650 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
651 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
653 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
654 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
655 of KIDIFY1, which already does not exist. So it gives up on
656 expansion, leaving garbage consisting of infinished blocks of the
657 partially converted function.)
659 (due to reordering of the compiler this example is compiled
660 successfully by 0.7.14, but the bug probably remains)
663 (reported by Robert E. Brown 2002-04-16)
664 When a function is called with too few arguments, causing the
665 debugger to be entered, the uninitialized slots in the bad call frame
666 seem to cause GCish problems, being interpreted as tagged data even
667 though they're not. In particular, executing ROOM in the
668 debugger at that point causes AVER failures:
671 * (lisp-implementation-version)
677 failed AVER: "(SAP= CURRENT END)"
678 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
679 isn't too surprising since there are many differences in stack
680 implementation and GC conservatism between the X86 and other ports.)
683 In sbcl-0.7.3.11, compiling the (illegal) code
684 (in-package :cl-user)
685 (defmethod prove ((uustk uustk))
688 gives the (not terribly clear) error message
690 ; (during macroexpansion of (DEFMETHOD PROVE ...))
691 ; can't get template for (FROB NIL NIL)
692 The problem seems to be that the code walker used by the DEFMETHOD
693 macro is unhappy with the illegal syntax in the method body, and
694 is giving an unclear error message.
697 The compiler sometimes tries to constant-fold expressions before
698 it checks to see whether they can be reached. This can lead to
699 bogus warnings about errors in the constant folding, e.g. in code
702 (WRITE-STRING (> X 0) "+" "0"))
703 compiled in a context where the compiler can prove that X is NIL,
704 and the compiler complains that (> X 0) causes a type error because
705 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
706 caused a full WARNING, which made the bug really annoying because then
707 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
708 code. Since then the warning has been downgraded to STYLE-WARNING,
709 so it's still a bug but at least it's a little less annoying.
711 183: "IEEE floating point issues"
712 Even where floating point handling is being dealt with relatively
713 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
714 accrued-exceptions and current-exceptions part of the fp control
715 word don't seem to bear much relation to reality. E.g. on
719 debugger invoked on condition of type DIVISION-BY-ZERO:
720 arithmetic error DIVISION-BY-ZERO signalled
721 0] (sb-vm::get-floating-point-modes)
723 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
724 :ROUNDING-MODE :NEAREST
725 :CURRENT-EXCEPTIONS NIL
726 :ACCRUED-EXCEPTIONS (:INEXACT)
729 * (sb-vm::get-floating-point-modes)
730 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
731 :ROUNDING-MODE :NEAREST
732 :CURRENT-EXCEPTIONS (:INEXACT)
733 :ACCRUED-EXCEPTIONS (:INEXACT)
736 187: "type inference confusion around DEFTRANSFORM time"
737 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
738 bug in DEFTRANSFORM")
739 After the file below is compiled and loaded in sbcl-0.7.5, executing
740 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
741 at the REPL returns an adjustable vector, which is wrong. Presumably
742 somehow the DERIVE-TYPE information for the output values of %WAD is
743 being mispropagated as a type constraint on the input values of %WAD,
744 and so causing the type test to be optimized away. It's unclear how
745 hand-expanding the DEFTRANSFORM would change this, but it suggests
746 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
747 invoked at a particular phase) is involved.
748 (cl:in-package :sb-c)
749 (eval-when (:compile-toplevel)
750 ;;; standin for %DATA-VECTOR-AND-INDEX
751 (defknown %dvai (array index)
753 (foldable flushable))
754 (deftransform %dvai ((array index)
758 (let* ((atype (continuation-type array))
759 (eltype (array-type-specialized-element-type atype)))
760 (when (eq eltype *wild-type*)
761 (give-up-ir1-transform
762 "specialized array element type not known at compile-time"))
763 (when (not (array-type-complexp atype))
764 (give-up-ir1-transform "SIMPLE array!"))
765 `(if (array-header-p array)
766 (%wad array index nil)
767 (values array index))))
768 ;;; standin for %WITH-ARRAY-DATA
769 (defknown %wad (array index (or index null))
770 (values (simple-array * (*)) index index index)
771 (foldable flushable))
772 ;;; (Commenting out this optimizer causes the bug to go away.)
773 (defoptimizer (%wad derive-type) ((array start end))
774 (let ((atype (continuation-type array)))
775 (when (array-type-p atype)
776 (values-specifier-type
777 `(values (simple-array ,(type-specifier
778 (array-type-specialized-element-type atype))
780 index index index)))))
782 (defun %wad (array start end)
783 (format t "~&in %WAD~%")
784 (%with-array-data array start end))
785 (cl:in-package :cl-user)
787 (declare (type (vector t) v))
788 (declare (notinline sb-kernel::%with-array-data))
789 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
793 188: "compiler performance fiasco involving type inference and UNION-TYPE"
794 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
795 bug could be fixed properly, so you won't see the bug unless you restore
796 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
800 (declare (optimize (safety 3)))
801 (declare (optimize (compilation-speed 2)))
802 (declare (optimize (speed 1) (debug 1) (space 1)))
803 (let ((fn "if-this-file-exists-the-universe-is-strange"))
804 (load fn :if-does-not-exist nil)
805 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
806 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
807 (load (concatenate 'string fn ".misc-garbage")
808 :if-does-not-exist nil)))))
810 134.552 seconds of real time
811 133.35156 seconds of user run time
812 0.03125 seconds of system run time
813 [Run times include 2.787 seconds GC run time.]
815 246883368 bytes consed.
816 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
817 thinking about type relationships involving types like
819 (OR (INTEGER 576 576)
830 In recent SBCL the following example also illustrates this bug:
835 (declare (optimize (safety 3)))
836 (declare (optimize (compilation-speed 2)))
837 (declare (optimize (speed 1) (debug 1) (space 1)))
839 (declare (type (integer 0) start))
840 (print (incf start 22))
841 (print (incf start 26))
842 (print (incf start 28)))
844 (declare (type (integer 0) start))
845 (print (incf start 22))
846 (print (incf start 26)))
848 (declare (type (integer 0) start))
849 (print (incf start 22))
850 (print (incf start 26))))))
852 190: "PPC/Linux pipe? buffer? bug"
853 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
854 on the PPC/Linux platform, waiting for a zombie env process. This
855 is a classic symptom of buffer filling and deadlock, but it seems
856 only sporadically reproducible.
858 191: "Miscellaneous PCL deficiencies"
859 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
860 a. DEFCLASS does not inform the compiler about generated
861 functions. Compiling a file with
865 (WITH-SLOTS (A-CLASS-X) A
867 results in a STYLE-WARNING:
869 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
871 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
872 Mai points out that the declamation of functions is in fact
873 incorrect in some cases (most notably for structure
874 classes). This means that at present erroneous attempts to use
875 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
876 won't get the corresponding STYLE-WARNING.
877 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
878 lists and &KEY arguments) do not signal errors when they should.
880 192: "Python treats free type declarations as promises."
881 b. What seemed like the same fundamental problem as bug 192a, but
882 was not fixed by the same (APD "more strict type checking
883 sbcl-devel 2002-08-97) patch:
884 (DOTIMES (I ...) (DOTIMES (J ...) (DECLARE ...) ...)):
885 (declaim (optimize (speed 1) (safety 3)))
886 (defun trust-assertion (i)
888 (declare (type (mod 4) i)) ; when commented out, behavior changes!
891 (trust-assertion 6) ; prints nothing unless DECLARE is commented out
896 (locally (declare (type fixnum x y))
900 194: "no error from (THE REAL '(1 2 3)) in some cases"
903 (multiple-value-prog1 (progn (the real '(1 2 3))))
904 returns (1 2 3) instead of signalling an error. This was fixed by
905 APD's "more strict type checking patch", but although the fixed
906 code (in sbcl-0.7.7.19) works (signals TYPE-ERROR) interactively,
907 it's difficult to write a regression test for it, because
908 (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
909 still returns (1 2 3).
911 b. (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
912 returns (1 2 3). (As above, this shows up when writing regression
913 tests for fixed-ness of part a.)
914 c. Also in sbcl-0.7.7.9, (IGNORE-ERRORS (THE REAL '(1 2 3))) => (1 2 3).
918 (arg2 (identity (the real #(1 2 3)))))
919 (if (< arg1 arg2) arg1 arg2))))
921 but putting the same expression inside (DEFUN FOO () ...),
924 * Actually this entry is probably multiple bugs, as
925 Alexey Dejneka commented on sbcl-devel 2002-09-03:)
926 I don't think that placing these two bugs in one entry is
927 a good idea: they have different explanations. The second
928 (min 1 nil) is caused by flushing of unused code--IDENTITY
929 can do nothing with it. So it is really bug 122. The first
930 (min nil) is due to M-V-PROG1: substituting a continuation
931 for the result, it forgets about type assertion. The purpose
932 of IDENTITY is to save the restricted continuation from
933 inaccurate transformations.
934 * Alexey Dejneka pointed out that
935 (IGNORE-ERRORS (IDENTITY (THE REAL '(1 2 3))))
937 (IGNORE-ERRORS (VALUES (THE REAL '(1 2 3))))
940 201: "Incautious type inference from compound CONS types"
941 (reported by APD sbcl-devel 2002-09-17)
943 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
945 (FORMAT NIL "~S IS ~S, Y = ~S"
952 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
955 Compiler does not check THEs on unused values, e.g. in
957 (progn (the real (list 1)) t)
959 This situation may appear during optimizing away degenerate cases of
960 certain functions: see bug 192b.
962 205: "environment issues in cross compiler"
963 (These bugs have no impact on user code, but should be fixed or
965 a. Macroexpanders introduced with MACROLET are defined in the null
967 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
968 the null lexical environment.
969 c. The cross-compiler cannot inline functions defined in a non-null
972 206: ":SB-FLUID feature broken"
973 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
974 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
977 207: "poorly distributed SXHASH results for compound data"
978 SBCL's SXHASH could probably try a little harder. ANSI: "the
979 intent is that an implementation should make a good-faith
980 effort to produce hash-codes that are well distributed
981 within the range of non-negative fixnums". But
982 (let ((hits (make-hash-table)))
985 (let* ((ij (cons i j))
986 (newlist (push ij (gethash (sxhash ij) hits))))
988 (format t "~&collision: ~S~%" newlist))))))
989 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
990 would be an obvious way of fix. Maybe it would be acceptably efficient
991 to redo MIX using a lookup into a 256-entry s-box containing
992 29-bit pseudorandom numbers?
994 208: "package confusion in PCL handling of structure slot handlers"
995 In sbcl-0.7.8 compiling and loading
997 (defstruct foo (slot (error "missing")) :type list :read-only t)
998 (defmethod print-object ((foo foo) stream) (print nil stream))
999 causes CERROR "attempting to modify a symbol in the COMMON-LISP
1000 package: FOO-SLOT". (This is fairly bad code, but still it's hard
1001 to see that it should cause symbols to be interned in the CL package.)
1003 211: "keywords processing"
1004 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
1005 number of keyword arguments.
1008 (flet ((foo (&key y) (list y)))
1009 (list (foo :y 1 :y 2)))
1011 issues confusing message
1016 ; caught STYLE-WARNING:
1017 ; The variable #:G15 is defined but never used.
1019 212: "Sequence functions and circular arguments"
1020 COERCE, MERGE and CONCATENATE go into an infinite loop when given
1021 circular arguments; it would be good for the user if they could be
1022 given an error instead (ANSI 17.1.1 allows this behaviour on the part
1023 of the implementation, as conforming code cannot give non-proper
1024 sequences to these functions. MAP also has this problem (and
1025 solution), though arguably the convenience of being able to do
1026 (MAP 'LIST '+ FOO '#1=(1 . #1#))
1027 might be classed as more important (though signalling an error when
1028 all of the arguments are circular is probably desireable).
1030 213: "Sequence functions and type checking"
1031 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
1032 various complicated, though recognizeable, CONS types [e.g.
1033 (CONS * (CONS * NULL))
1034 which according to ANSI should be recognized] (and, in SAFETY 3
1035 code, should return a list of LENGTH 2 or signal an error)
1036 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
1037 check that it will return a sequence of the given type. Fixing
1038 it along the same lines as the others (cf. work done around
1039 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
1040 entirely straightforward.
1041 c. All of these functions will silently accept a type of the form
1043 whether or not the return value is of this type. This is
1044 probably permitted by ANSI (see "Exceptional Situations" under
1045 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
1046 know about this escape clause, so code of the form
1047 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
1048 can erroneously return T.
1051 SBCL 0.6.12.43 fails to compile
1054 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1055 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
1058 or a more simple example:
1061 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1062 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
1064 215: ":TEST-NOT handling by functions"
1065 a. FIND and POSITION currently signal errors when given non-NIL for
1066 both their :TEST and (deprecated) :TEST-NOT arguments, but by
1067 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
1068 means that the effect is "unpredictable but harmless". It's not
1069 clear what that actually means; it may preclude conforming
1070 implementations from signalling errors.
1071 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
1072 when conflict occurs. As a quality of implementation issue, it
1073 might be preferable to treat :TEST and :TEST-NOT as being in some
1074 sense the same &KEY, and effectively take the first test function in
1076 c. Again, a quality of implementation issue: it would be good to issue a
1077 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
1078 WARNING for calls with both :TEST and :TEST-NOT; possibly this
1079 latter should be WARNed about at execute-time too.
1081 216: "debugger confused by frames with invalid number of arguments"
1082 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
1083 leaves the system confused, enough so that (QUIT) no longer works.
1084 It's as though the process of working with the uninitialized slot in
1085 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
1086 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
1088 217: "Bad type operations with FUNCTION types"
1091 * (values-type-union (specifier-type '(function (base-char)))
1092 (specifier-type '(function (integer))))
1094 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
1096 It causes insertion of wrong type assertions into generated
1100 (let ((f (etypecase x
1101 (character #'write-char)
1102 (integer #'write-byte))))
1105 (character (write-char x s))
1106 (integer (write-byte x s)))))
1108 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
1110 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
1111 produce invalid code, but type checking is not accurate. Similar
1112 problems exist with VALUES-TYPE-INTERSECTION.)
1114 218: "VALUES type specifier semantics"
1115 (THE (VALUES ...) ...) in safe code discards extra values.
1117 (defun test (x y) (the (values integer) (truncate x y)))
1121 Sbcl 0.7.9 fails to compile
1123 (multiple-value-call #'list
1124 (the integer (helper))
1127 Type check for INTEGER, the result of which serves as the first
1128 argument of M-V-C, is inserted after evaluation of NIL. So arguments
1129 of M-V-C are pushed in the wrong order. As a temporary workaround
1130 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
1131 would be to put the check between evaluation of arguments, but it
1132 could be tricky to check result types of PROG1, IF etc.
1135 (subtypep 'function '(function)) => nil, t.
1137 233: bugs in constraint propagation
1140 (declare (optimize (speed 2) (safety 3)))
1143 (the double-float x)
1146 (quux y (+ y 2d0) (* y 3d0)))))
1147 (foo 4) => segmentation violation
1149 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
1153 (declaim (optimize (speed 2) (safety 3)))
1155 (if (typep (prog1 x (setq x y)) 'double-float)
1158 (foo 1d0 5) => segmentation violation
1160 235: "type system and inline expansion"
1162 (declaim (ftype (function (cons) number) acc))
1163 (declaim (inline acc))
1165 (the number (car c)))
1168 (values (locally (declare (optimize (safety 0)))
1170 (locally (declare (optimize (safety 3)))
1173 (foo '(nil) '(t)) => NIL, T.
1175 b. (reported by brown on #lisp 2003-01-21)
1178 (declare (optimize (speed 3) (safety 0)))
1179 (declare (notinline mapcar))
1180 (let ((z (mapcar #'car x)))
1183 Without (DECLARE (NOTINLINE MAPCAR)), Python cannot derive that Z is
1186 236: "THE semantics is broken"
1189 (declare (optimize (speed 2) (safety 0)))
1192 (multiple-value-prog1
1194 (unless f (return-from foo 0))))))
1196 (foo #(4) nil) => SEGV
1198 VOP selection thinks that in unsafe code result type assertions
1199 should be valid immediately. (See also bug 233a.)
1201 The similar problem exists for TRULY-THE.
1203 237: "Environment arguments to type functions"
1204 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
1205 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
1206 argument, but they ignore it completely. This is almost
1207 certainly not correct.
1208 b. Also, the compiler's optimizers for TYPEP have not been informed
1209 about the new argument; consequently, they will not transform
1210 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
1211 just as optimizeable as (TYPEP 1 'INTEGER).
1213 238: "REPL compiler overenthusiasm for CLOS code"
1215 * (defclass foo () ())
1216 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
1217 causes approximately 100 lines of code deletion notes. Some
1218 discussion on this issue happened under the title 'Three "interesting"
1219 bugs in PCL', resulting in a fix for this oververbosity from the
1220 compiler proper; however, the problem persists in the interactor
1221 because the notion of original source is not preserved: for the
1222 compiler, the original source of the above expression is (DEFMETHOD
1223 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
1224 compiler gets its hands on the code needing compilation from the REPL,
1225 it has been macroexpanded several times.
1227 A symptom of the same underlying problem, reported by Tony Martinez:
1229 (with-input-from-string (*query-io* " no")
1231 (simple-type-error () 'error))
1233 ; (SB-KERNEL:FLOAT-WAIT)
1235 ; note: deleting unreachable code
1236 ; compilation unit finished
1239 241: "DEFCLASS mysteriously remembers uninterned accessor names."
1240 (from tonyms on #lisp IRC 2003-02-25)
1241 In sbcl-0.7.12.55, typing
1242 (defclass foo () ((bar :accessor foo-bar)))
1245 (defclass foo () ((bar :accessor foo-bar)))
1246 gives the error message
1247 "#:FOO-BAR already names an ordinary function or a macro."
1248 So it's somehow checking the uninterned old accessor name instead
1249 of the new requested accessor name, which seems broken to me (WHN).
1251 242: "WRITE-SEQUENCE suboptimality"
1252 (observed from clx performance)
1253 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
1254 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
1255 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
1256 rather than writing the sequence in one go, leading to severe
1257 performance degradation.
1259 243: "STYLE-WARNING overenthusiasm for unused variables"
1260 (observed from clx compilation)
1261 In sbcl-0.7.14, in the presence of the macros
1262 (DEFMACRO FOO (X) `(BAR ,X))
1263 (DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
1264 somewhat surprising style warnings are emitted for
1265 (COMPILE NIL '(LAMBDA (Y) (FOO Y))):
1267 ; (LAMBDA (Y) (FOO Y))
1269 ; caught STYLE-WARNING:
1270 ; The variable Y is defined but never used.
1272 244: "optimizing away tests for &KEY args of type declared in DEFKNOWN"
1273 (caught by clocc-ansi-test :EXCEPSIT-LEGACY-1050)
1274 In sbcl-0.pre8.44, (OPEN "foo" :DIRECTION :INPUT :EXTERNAL-FORMAT 'FOO)
1275 succeeds with no error (ignoring the bogus :EXTERNAL-FORMAT argument)
1276 apparently because the test is optimized away. The problem doesn't
1277 exist in sbcl-0.pre8.19. Deleting the (MEMBER :DEFAULT) declaration
1278 for :EXTERNAL-FORMAT in DEFKNOWN OPEN (and LOAD) is a workaround for
1279 the problem (and should be removed when the problem is fixed).
1281 245: bugs in disassembler
1282 a. On X86 an immediate operand for IMUL is printed incorrectly.
1283 b. On X86 operand size prefix is not recognized.
1285 246: "NTH-VALUE scaling problem"
1286 NTH-VALUE's current implementation for constant integers scales in
1287 compile-time as O(n^4), as indeed must the optional dispatch
1288 mechanism on which it is implemented. While it is unlikely to
1289 matter in real user code, it's still unpleasant to observe that
1290 (NTH-VALUE 1000 (VALUES-LIST (MAKE-LIST 1001))) takes several hours
1293 248: "reporting errors in type specifier syntax"
1294 (TYPEP 1 '(SYMBOL NIL)) says something about "unknown type
1297 DEFUNCT CATEGORIES OF BUGS
1299 These labels were used for bugs related to the old IR1 interpreter.
1300 The # values reached 6 before the category was closed down.