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 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
150 (setf (find-class 'ccc1) (find-class 'ccc))
151 (defmethod zut ((c ccc1)) 123)
152 In sbcl-0.7.1.13, this gives an error,
153 There is no class named CCC1.
154 In sbcl-0.pre8.20, this works, but prints style warnings about
158 Sometimes (SB-EXT:QUIT) fails with
159 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
160 Process inferior-lisp exited abnormally with code 1
161 I haven't noticed a repeatable case of this yet.
164 The printer doesn't report closures very well. This is true in
168 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
169 It would be nice to make closures have a settable name slot,
170 and make things like DEFSTRUCT and FLET, which create closures,
171 set helpful values into this slot.
174 And as long as we're wishing, it would be awfully nice if INSPECT could
175 also report on closures, telling about the values of the bound variables.
178 The compiler assumes that any time a function of declared FTYPE
179 doesn't signal an error, its arguments were of the declared type.
180 E.g. compiling and loading
181 (DECLAIM (OPTIMIZE (SAFETY 3)))
182 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
184 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
186 (COND ((> (FACTORIAL X) 1.0E6)
187 (FORMAT T "too big~%"))
189 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
191 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
194 will cause the INTEGERP case to be selected, giving bogus output a la
196 This violates the "declarations are assertions" principle.
197 According to the ANSI spec, in the section "System Class FUNCTION",
198 this is a case of "lying to the compiler", but the lying is done
199 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
200 not by the unexpectedly general definition of FACTORIAL. In any case,
201 "declarations are assertions" means that lying to the compiler should
202 cause an error to be signalled, and should not cause a bogus
203 result to be returned. Thus, the compiler should not assume
204 that arbitrary functions check their argument types. (It might
205 make sense to add another flag (CHECKED?) to DEFKNOWN to
206 identify functions which *do* check their argument types.)
207 (Also, verify that the compiler handles declared function
208 return types as assertions.)
211 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
212 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
214 (DECLARE (TYPE INDEXOID X))
215 (THE (VALUES INDEXOID)
217 where the implementation of the type check in function FOO
218 includes a full call to %TYPEP. There are also some fundamental problems
219 with the interpretation of VALUES types (inherited from CMU CL, and
220 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
221 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
224 The definitions of SIGCONTEXT-FLOAT-REGISTER and
225 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
226 supported on FreeBSD because the floating point state is not saved,
227 but at least as of FreeBSD 4.0, the floating point state *is* saved,
228 so they could be supported after all. Very likely
229 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
232 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
233 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
234 issue with 'union, 'and, 'or etc. So even though according to the
235 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
236 (and now SBCL) interpret them as legal types.
239 a slew of floating-point-related errors reported by Peter Van Eynde
241 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT on the x86 is
242 bogus, and should probably be 1.4012985e-45. In SBCL,
243 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
244 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
245 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
246 and LEAST-NEGATIVE-LONG-FLOAT.
247 c: Many expressions generate floating infinity on x86/Linux:
252 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
253 on x86/Linux generates the infinities instead. That might or
254 might not be conforming behavior, but it's also inconsistent,
255 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
256 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
257 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
259 d: (in section12.erg) various forms a la
260 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
261 don't give the right behavior.
264 type safety errors reported by Peter Van Eynde July 25, 2000:
265 c: (COERCE 'AND 'FUNCTION) returns something related to
266 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
267 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
268 not a binary input stream, but instead cheerfully reads from
269 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
272 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
273 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
274 causes a COMPILER-ERROR.
277 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
279 (DEFGENERIC FOO02 (X))
280 (DEFMETHOD FOO02 ((X NUMBER)) T)
281 (LET ((M (FIND-METHOD (FUNCTION FOO02)
283 (LIST (FIND-CLASS (QUOTE NUMBER))))))
284 (REMOVE-METHOD (FUNCTION FOO02) M)
285 (DEFGENERIC FOO03 (X))
286 (ADD-METHOD (FUNCTION FOO03) M)))
287 should give an error, but SBCL allows it.
290 The debugger LIST-LOCATIONS command doesn't work properly.
293 Compiling and loading
294 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
296 then requesting a BACKTRACE at the debugger prompt gives no information
297 about where in the user program the problem occurred.
300 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
301 I am looking into this problem that showed up on the cmucl-help
302 list. It seems to me that the "implementation specific environment
303 hacking functions" found in pcl/walker.lisp are completely messed
304 up. The good thing is that they appear to be barely used within
305 PCL and the munged environment object is passed to cmucl only
306 in calls to macroexpand-1, which is probably why this case fails.
307 SBCL uses essentially the same code, so if the environment hacking
308 is screwed up, it affects us too.
311 Using the pretty-printer from the command prompt gives funny
312 results, apparently because the pretty-printer doesn't know
313 about user's command input, including the user's carriage return
314 that the user, and therefore the pretty-printer thinks that
315 the new output block should start indented 2 or more characters
316 rightward of the correct location.
319 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
320 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
321 crashes SBCL. In general tracing anything which is used in the
322 implementation of TRACE is likely to have the same problem.
325 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
326 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
327 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
328 WITH-OUTPUT-TO-STRING.
331 ANSI says in one place that type declarations can be abbreviated even
332 when the type name is not a symbol, e.g.
333 (DECLAIM ((VECTOR T) *FOOVECTOR*))
334 SBCL doesn't support this. But ANSI says in another place that this
335 isn't allowed. So it's not clear this is a bug after all. (See the
336 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
340 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
341 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
342 an easily guessable temporary filename in a way which might open
343 applications using LOAD-FOREIGN to hijacking by malicious users
344 on the same machine. Incantations for doing this safely are
345 floating around the net in various "how to write secure programs
346 despite Unix" documents, and it would be good to (1) fix this in
347 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
348 files and make it share the same new safe logic.
350 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
351 make the temporary filename less easily guessable)
354 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
355 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
356 the mean of the distribution can be systematically O(0.1%) wrong.
357 Just increasing R-I-E-B is probably not a good solution, since
358 it would decrease efficiency more than is probably necessary. Perhaps
359 using some sort of accept/reject method would be better.
362 Internally the compiler sometimes evaluates
363 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
364 (I stumbled across this when I added an
365 (assert (not (eq type1 *wild-type*)))
366 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
367 in a type context should probably be translated to T, and so it's
368 probably wrong to ask whether it's equal to the T type and then (using
369 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
370 (I haven't tried to investigate this bug enough to guess whether
371 there might be any user-level symptoms.)
373 In fact, the type system is likely to depend on this inequality not
374 holding... * is not equivalent to T in many cases, such as
375 (VECTOR *) /= (VECTOR T).
378 Inconsistencies between derived and declared VALUES return types for
379 DEFUN aren't checked very well. E.g. the logic which successfully
380 catches problems like
381 (declaim (ftype (function (fixnum) float) foo))
383 (declare (type integer x))
384 (values x)) ; wrong return type, detected, gives warning, good!
386 (declaim (ftype (function (t) (values t t)) bar))
388 (values x)) ; wrong number of return values, no warning, bad!
389 The cause of this is seems to be that (1) the internal function
390 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
391 arguments symmetrically, and (2) when the type checking code was
392 written back when when SBCL's code was still CMU CL, the intent
394 (declaim (ftype (function (t) t) bar))
396 (values x x)) ; wrong number of return values; should give warning?
397 not be warned for, because a two-valued return value is considered
398 to be compatible with callers who expects a single value to be
399 returned. That intent is probably not appropriate for modern ANSI
400 Common Lisp, but fixing this might be complicated because of other
401 divergences between auld-style and new-style handling of
402 multiple-VALUES types. (Some issues related to this were discussed
403 on cmucl-imp at some length sometime in 2000.)
406 The facility for dumping a running Lisp image to disk gets confused
407 when run without the PURIFY option, and creates an unnecessarily large
408 core file (apparently representing memory usage up to the previous
409 high-water mark). Moreover, when the file is loaded, it confuses the
410 GC, so that thereafter memory usage can never be reduced below that
414 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
415 CL), out-of-line structure slot setters are horribly inefficient
416 whenever the type of the slot is declared, because out-of-line
417 structure slot setters are implemented as closures to save space,
418 so the compiler doesn't compile the type test into code, but
419 instead just saves the type in a lexical closure and interprets it
421 A proper solution involves deciding whether it's really worth
422 saving space by implementing structure slot accessors as closures.
423 (If it's not worth it, the problem vanishes automatically. If it
424 is worth it, there are hacks we could use to force type tests to
425 be compiled anyway, and even shared. E.g. we could implement
426 an EQUAL hash table mapping from types to compiled type tests,
427 and save the appropriate compiled type test as part of each lexical
428 closure; or we could make the lexical closures be placeholders
429 which overwrite their old definition as a lexical closure with
430 a new compiled definition the first time that they're called.)
431 As a workaround for the problem, #'(SETF FOO) expressions can
432 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
433 (defmacro efficient-setf-function (place-function-name)
434 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
435 ;; a workaround for the problem, encouraging the
436 ;; inline expansion of the structure accessor, so
437 ;; that the compiler can optimize its type test
438 (let ((new-value (gensym "NEW-VALUE-"))
439 (structure-value (gensym "STRUCTURE-VALUE-")))
440 `(lambda (,new-value ,structure-value)
441 (setf (,place-function-name ,structure-value)
443 ;; no problem, can just use the ordinary expansion
444 `(function (setf ,place-function-name))))
447 There's apparently a bug in CEILING optimization which caused
448 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
449 applied the patches to SBCL and they didn't seem to cause problems
450 (as reported sbcl-devel 2001-05-04). However, since the patches
451 modify nontrivial code which was apparently written incorrectly
452 the first time around, until regression tests are written I'm not
453 comfortable merging the patches in the CVS version of SBCL.
456 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
457 a clean, just-started SBCL system. And it seems to be right:
458 (ROOM T) can bring a small computer to its knees for a *long*
459 time trying to GC afterwards. Surely there's some more economical
460 way to implement (ROOM T).
463 When the compiler inline expands functions, it may be that different
464 kinds of return values are generated from different code branches.
465 E.g. an inline expansion of POSITION generates integer results
466 from one branch, and NIL results from another. When that inline
467 expansion is used in a context where only one of those results
470 (aref *a1* (position x *a2*)))
471 and the compiler can't prove that the unacceptable branch is
472 never taken, then bogus type mismatch warnings can be generated.
473 If you need to suppress the type mismatch warnings, you can
474 suppress the inline expansion,
476 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
477 (aref *a1* (position x *a2*)))
478 or, sometimes, suppress them by declaring the result to be of an
481 (aref *a1* (the integer (position x *a2*))))
483 This is not a new compiler problem in 0.7.0, but the new compiler
484 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
485 more conspicuous. If you don't need performance from these functions,
486 and the bogus warnings are a nuisance for you, you can return to
487 your pre-0.7.0 state of grace with
488 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
491 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
492 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
493 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
494 when of course it should be NIL. (He says it only fails for X86,
495 not SPARC; dunno about Alpha.)
497 Also, "the same problem exists for LONG-FLOAT-EPSILON,
498 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
499 for the -negative- the + is replaced by a - in the test)."
501 Raymond Toy comments that this is tricky on the X86 since its FPU
502 uses 80-bit precision internally.
505 Even in sbcl-0.pre7.x, which is supposed to be free of the old
506 non-ANSI behavior of treating the function return type inferred
507 from the current function definition as a declaration of the
508 return type from any function of that name, the return type of NIL
509 is attached to FOO in 120a above, and used to optimize code which
513 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
514 the entire lexical environment at the point of MACROLET available
515 in the bodies of the macroexpander functions. In particular, it
516 allows the function bodies (which run at compile time) to try to
517 access lexical variables (which are only defined at runtime).
518 It doesn't even issue a warning, which is bad.
520 The SBCL behavior arguably conforms to the ANSI spec (since the
521 spec says that the behavior is undefined, ergo anything conforms).
522 However, it would be better to issue a compile-time error.
523 Unfortunately I (WHN) don't see any simple way to detect this
524 condition in order to issue such an error, so for the meantime
525 SBCL just does this weird broken "conforming" thing.
527 The ANSI standard says, in the definition of the special operator
529 The macro-expansion functions defined by MACROLET are defined
530 in the lexical environment in which the MACROLET form appears.
531 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
532 the local macro definitions in a MACROLET, but the consequences
533 are undefined if the local macro definitions reference any
534 local variable or function bindings that are visible in that
536 Then it seems to contradict itself by giving the example
538 (macrolet ((fudge (z)
539 ;The parameters x and flag are not accessible
540 ; at this point; a reference to flag would be to
541 ; the global variable of that name.
542 ` (if flag (* ,z ,z) ,z)))
543 ;The parameters x and flag are accessible here.
547 The comment "a reference to flag would be to the global variable
548 of the same name" sounds like good behavior for the system to have.
549 but actual specification quoted above says that the actual behavior
552 (Since 0.7.8.23 macroexpanders are defined in a restricted version
553 of the lexical environment, containing no lexical variables and
554 functions, which seems to conform to ANSI and CLtL2, but signalling
555 a STYLE-WARNING for references to variables similar to locals might
559 (as reported by Gabe Garza on cmucl-help 2001-09-21)
561 (defun test-pred (x y)
565 (func (lambda () x)))
566 (print (eq func func))
567 (print (test-pred func func))
568 (delete func (list func))))
569 Now calling (TEST-CASE) gives output
572 (#<FUNCTION {500A9EF9}>)
573 Evidently Python thinks of the lambda as a code transformation so
574 much that it forgets that it's also an object.
577 Ideally, uninterning a symbol would allow it, and its associated
578 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
579 at least as of sbcl-0.7.0, this isn't the case. Information about
580 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
581 essentially in ordinary (non-weak) hash tables keyed by symbols.
582 Thus, once a system has an entry in this system, it tends to live
583 forever, even when it is uninterned and all other references to it
586 141: "pretty printing and backquote"
589 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
592 * (write '`(, .ala.) :readably t :pretty t)
595 (note the space between the comma and the point)
598 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
599 prominent SourceForge web/db bug tracking system, which is
600 unfortunately not a reliable way to get a timely response from
601 the SBCL maintainers)
602 In the course of trying to build a test case for an
603 application error, I encountered this behavior:
604 If you start up sbcl, and then lay on CTRL-C for a
605 minute or two, the lisp process will eventually say:
606 %PRIMITIVE HALT called; the party is over.
607 and throw you into the monitor. If I start up lisp,
608 attach to the process with strace, and then do the same
609 (abusive) thing, I get instead:
610 access failure in heap page not marked as write-protected
611 and the monitor again. I don't know enough to have the
612 faintest idea of what is going on here.
613 This is with sbcl 6.12, uname -a reports:
614 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
615 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
616 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
617 it took more than two minutes (but less than five) for me.
620 (This was once known as IR1-4, but it lived on even after the
621 IR1 interpreter went to the big bit bucket in the sky.)
622 The system accepts DECLAIM in most places where DECLARE would be
623 accepted, without even issuing a warning. ANSI allows this, but since
624 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
625 meaning is rather different, and it's unlikely that the user
626 has a good reason for doing DECLAIM not at top level, it would be
627 good to issue a STYLE-WARNING when this happens. A possible
628 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
629 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
630 [This is considered an IR1-interpreter-related bug because until
631 EVAL-WHEN is rewritten, which won't happen until after the IR1
632 interpreter is gone, the system's notion of what's a top-level form
633 and what's not will remain too confused to fix this problem.]
636 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
637 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
638 COMPLEX implementation didn't deal with this, and hasn't been
639 upgraded to do so. (This doesn't seem to be a high priority
640 conformance problem, since seems hard to construct useful code
644 Floating point errors are reported poorly. E.g. on x86 OpenBSD
647 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
648 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
649 No traps are enabled? How can this be?
650 It should be possible to be much more specific (overflow, division
651 by zero, etc.) and of course the "How can this be?" should be fixable.
653 See also bugs #45.c and #183
656 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
657 (in-package :cl-user)
660 (defstruct foo bar bletch)
662 (labels ((kidify1 (kid)
670 (declare (inline kid-frob))
673 (the simple-vector (foo-bar perd)))))
675 debugger invoked on condition of type TYPE-ERROR:
676 The value NIL is not of type SB-C::NODE.
677 The location of this failure has moved around as various related
678 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
679 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
681 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
682 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
683 of KIDIFY1, which already does not exist. So it gives up on
684 expansion, leaving garbage consisting of infinished blocks of the
685 partially converted function.)
687 (due to reordering of the compiler this example is compiled
688 successfully by 0.7.14, but the bug probably remains)
691 (reported by Robert E. Brown 2002-04-16)
692 When a function is called with too few arguments, causing the
693 debugger to be entered, the uninitialized slots in the bad call frame
694 seem to cause GCish problems, being interpreted as tagged data even
695 though they're not. In particular, executing ROOM in the
696 debugger at that point causes AVER failures:
699 * (lisp-implementation-version)
705 failed AVER: "(SAP= CURRENT END)"
706 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
707 isn't too surprising since there are many differences in stack
708 implementation and GC conservatism between the X86 and other ports.)
711 In sbcl-0.7.3.11, compiling the (illegal) code
712 (in-package :cl-user)
713 (defmethod prove ((uustk uustk))
716 gives the (not terribly clear) error message
718 ; (during macroexpansion of (DEFMETHOD PROVE ...))
719 ; can't get template for (FROB NIL NIL)
720 The problem seems to be that the code walker used by the DEFMETHOD
721 macro is unhappy with the illegal syntax in the method body, and
722 is giving an unclear error message.
725 The compiler sometimes tries to constant-fold expressions before
726 it checks to see whether they can be reached. This can lead to
727 bogus warnings about errors in the constant folding, e.g. in code
730 (WRITE-STRING (> X 0) "+" "0"))
731 compiled in a context where the compiler can prove that X is NIL,
732 and the compiler complains that (> X 0) causes a type error because
733 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
734 caused a full WARNING, which made the bug really annoying because then
735 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
736 code. Since then the warning has been downgraded to STYLE-WARNING,
737 so it's still a bug but at least it's a little less annoying.
739 183: "IEEE floating point issues"
740 Even where floating point handling is being dealt with relatively
741 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
742 accrued-exceptions and current-exceptions part of the fp control
743 word don't seem to bear much relation to reality. E.g. on
747 debugger invoked on condition of type DIVISION-BY-ZERO:
748 arithmetic error DIVISION-BY-ZERO signalled
749 0] (sb-vm::get-floating-point-modes)
751 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
752 :ROUNDING-MODE :NEAREST
753 :CURRENT-EXCEPTIONS NIL
754 :ACCRUED-EXCEPTIONS (:INEXACT)
757 * (sb-vm::get-floating-point-modes)
758 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
759 :ROUNDING-MODE :NEAREST
760 :CURRENT-EXCEPTIONS (:INEXACT)
761 :ACCRUED-EXCEPTIONS (:INEXACT)
764 187: "type inference confusion around DEFTRANSFORM time"
765 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
766 bug in DEFTRANSFORM")
767 After the file below is compiled and loaded in sbcl-0.7.5, executing
768 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
769 at the REPL returns an adjustable vector, which is wrong. Presumably
770 somehow the DERIVE-TYPE information for the output values of %WAD is
771 being mispropagated as a type constraint on the input values of %WAD,
772 and so causing the type test to be optimized away. It's unclear how
773 hand-expanding the DEFTRANSFORM would change this, but it suggests
774 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
775 invoked at a particular phase) is involved.
776 (cl:in-package :sb-c)
777 (eval-when (:compile-toplevel)
778 ;;; standin for %DATA-VECTOR-AND-INDEX
779 (defknown %dvai (array index)
781 (foldable flushable))
782 (deftransform %dvai ((array index)
786 (let* ((atype (continuation-type array))
787 (eltype (array-type-specialized-element-type atype)))
788 (when (eq eltype *wild-type*)
789 (give-up-ir1-transform
790 "specialized array element type not known at compile-time"))
791 (when (not (array-type-complexp atype))
792 (give-up-ir1-transform "SIMPLE array!"))
793 `(if (array-header-p array)
794 (%wad array index nil)
795 (values array index))))
796 ;;; standin for %WITH-ARRAY-DATA
797 (defknown %wad (array index (or index null))
798 (values (simple-array * (*)) index index index)
799 (foldable flushable))
800 ;;; (Commenting out this optimizer causes the bug to go away.)
801 (defoptimizer (%wad derive-type) ((array start end))
802 (let ((atype (continuation-type array)))
803 (when (array-type-p atype)
804 (values-specifier-type
805 `(values (simple-array ,(type-specifier
806 (array-type-specialized-element-type atype))
808 index index index)))))
810 (defun %wad (array start end)
811 (format t "~&in %WAD~%")
812 (%with-array-data array start end))
813 (cl:in-package :cl-user)
815 (declare (type (vector t) v))
816 (declare (notinline sb-kernel::%with-array-data))
817 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
821 188: "compiler performance fiasco involving type inference and UNION-TYPE"
822 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
823 bug could be fixed properly, so you won't see the bug unless you restore
824 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
828 (declare (optimize (safety 3)))
829 (declare (optimize (compilation-speed 2)))
830 (declare (optimize (speed 1) (debug 1) (space 1)))
831 (let ((fn "if-this-file-exists-the-universe-is-strange"))
832 (load fn :if-does-not-exist nil)
833 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
834 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
835 (load (concatenate 'string fn ".misc-garbage")
836 :if-does-not-exist nil)))))
838 134.552 seconds of real time
839 133.35156 seconds of user run time
840 0.03125 seconds of system run time
841 [Run times include 2.787 seconds GC run time.]
843 246883368 bytes consed.
844 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
845 thinking about type relationships involving types like
847 (OR (INTEGER 576 576)
858 190: "PPC/Linux pipe? buffer? bug"
859 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
860 on the PPC/Linux platform, waiting for a zombie env process. This
861 is a classic symptom of buffer filling and deadlock, but it seems
862 only sporadically reproducible.
864 191: "Miscellaneous PCL deficiencies"
865 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
866 a. DEFCLASS does not inform the compiler about generated
867 functions. Compiling a file with
871 (WITH-SLOTS (A-CLASS-X) A
873 results in a STYLE-WARNING:
875 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
877 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
878 Mai points out that the declamation of functions is in fact
879 incorrect in some cases (most notably for structure
880 classes). This means that at present erroneous attempts to use
881 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
882 won't get the corresponding STYLE-WARNING.
883 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
884 lists and &KEY arguments) do not signal errors when they should.
886 192: "Python treats free type declarations as promises."
887 b. What seemed like the same fundamental problem as bug 192a, but
888 was not fixed by the same (APD "more strict type checking
889 sbcl-devel 2002-08-97) patch:
890 (DOTIMES (I ...) (DOTIMES (J ...) (DECLARE ...) ...)):
891 (declaim (optimize (speed 1) (safety 3)))
892 (defun trust-assertion (i)
894 (declare (type (mod 4) i)) ; when commented out, behavior changes!
897 (trust-assertion 6) ; prints nothing unless DECLARE is commented out
902 (locally (declare (type fixnum x y))
906 194: "no error from (THE REAL '(1 2 3)) in some cases"
909 (multiple-value-prog1 (progn (the real '(1 2 3))))
910 returns (1 2 3) instead of signalling an error. This was fixed by
911 APD's "more strict type checking patch", but although the fixed
912 code (in sbcl-0.7.7.19) works (signals TYPE-ERROR) interactively,
913 it's difficult to write a regression test for it, because
914 (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
915 still returns (1 2 3).
917 b. (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
918 returns (1 2 3). (As above, this shows up when writing regression
919 tests for fixed-ness of part a.)
920 c. Also in sbcl-0.7.7.9, (IGNORE-ERRORS (THE REAL '(1 2 3))) => (1 2 3).
924 (arg2 (identity (the real #(1 2 3)))))
925 (if (< arg1 arg2) arg1 arg2))))
927 but putting the same expression inside (DEFUN FOO () ...),
930 * Actually this entry is probably multiple bugs, as
931 Alexey Dejneka commented on sbcl-devel 2002-09-03:)
932 I don't think that placing these two bugs in one entry is
933 a good idea: they have different explanations. The second
934 (min 1 nil) is caused by flushing of unused code--IDENTITY
935 can do nothing with it. So it is really bug 122. The first
936 (min nil) is due to M-V-PROG1: substituting a continuation
937 for the result, it forgets about type assertion. The purpose
938 of IDENTITY is to save the restricted continuation from
939 inaccurate transformations.
940 * Alexey Dejneka pointed out that
941 (IGNORE-ERRORS (IDENTITY (THE REAL '(1 2 3))))
943 (IGNORE-ERRORS (VALUES (THE REAL '(1 2 3))))
946 201: "Incautious type inference from compound CONS types"
947 (reported by APD sbcl-devel 2002-09-17)
949 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
951 (FORMAT NIL "~S IS ~S, Y = ~S"
958 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
961 Compiler does not check THEs on unused values, e.g. in
963 (progn (the real (list 1)) t)
965 This situation may appear during optimizing away degenerate cases of
966 certain functions: see bug 192b.
968 205: "environment issues in cross compiler"
969 (These bugs have no impact on user code, but should be fixed or
971 a. Macroexpanders introduced with MACROLET are defined in the null
973 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
974 the null lexical environment.
975 c. The cross-compiler cannot inline functions defined in a non-null
978 206: ":SB-FLUID feature broken"
979 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
980 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
983 207: "poorly distributed SXHASH results for compound data"
984 SBCL's SXHASH could probably try a little harder. ANSI: "the
985 intent is that an implementation should make a good-faith
986 effort to produce hash-codes that are well distributed
987 within the range of non-negative fixnums". But
988 (let ((hits (make-hash-table)))
991 (let* ((ij (cons i j))
992 (newlist (push ij (gethash (sxhash ij) hits))))
994 (format t "~&collision: ~S~%" newlist))))))
995 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
996 would be an obvious way of fix. Maybe it would be acceptably efficient
997 to redo MIX using a lookup into a 256-entry s-box containing
998 29-bit pseudorandom numbers?
1000 208: "package confusion in PCL handling of structure slot handlers"
1001 In sbcl-0.7.8 compiling and loading
1003 (defstruct foo (slot (error "missing")) :type list :read-only t)
1004 (defmethod print-object ((foo foo) stream) (print nil stream))
1005 causes CERROR "attempting to modify a symbol in the COMMON-LISP
1006 package: FOO-SLOT". (This is fairly bad code, but still it's hard
1007 to see that it should cause symbols to be interned in the CL package.)
1009 211: "keywords processing"
1010 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
1011 number of keyword arguments.
1014 (flet ((foo (&key y) (list y)))
1015 (list (foo :y 1 :y 2)))
1017 issues confusing message
1022 ; caught STYLE-WARNING:
1023 ; The variable #:G15 is defined but never used.
1025 212: "Sequence functions and circular arguments"
1026 COERCE, MERGE and CONCATENATE go into an infinite loop when given
1027 circular arguments; it would be good for the user if they could be
1028 given an error instead (ANSI 17.1.1 allows this behaviour on the part
1029 of the implementation, as conforming code cannot give non-proper
1030 sequences to these functions. MAP also has this problem (and
1031 solution), though arguably the convenience of being able to do
1032 (MAP 'LIST '+ FOO '#1=(1 . #1#))
1033 might be classed as more important (though signalling an error when
1034 all of the arguments are circular is probably desireable).
1036 213: "Sequence functions and type checking"
1037 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
1038 various complicated, though recognizeable, CONS types [e.g.
1039 (CONS * (CONS * NULL))
1040 which according to ANSI should be recognized] (and, in SAFETY 3
1041 code, should return a list of LENGTH 2 or signal an error)
1042 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
1043 check that it will return a sequence of the given type. Fixing
1044 it along the same lines as the others (cf. work done around
1045 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
1046 entirely straightforward.
1047 c. All of these functions will silently accept a type of the form
1049 whether or not the return value is of this type. This is
1050 probably permitted by ANSI (see "Exceptional Situations" under
1051 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
1052 know about this escape clause, so code of the form
1053 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
1054 can erroneously return T.
1057 SBCL 0.6.12.43 fails to compile
1060 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1061 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
1064 or a more simple example:
1067 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1068 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
1070 215: ":TEST-NOT handling by functions"
1071 a. FIND and POSITION currently signal errors when given non-NIL for
1072 both their :TEST and (deprecated) :TEST-NOT arguments, but by
1073 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
1074 means that the effect is "unpredictable but harmless". It's not
1075 clear what that actually means; it may preclude conforming
1076 implementations from signalling errors.
1077 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
1078 when conflict occurs. As a quality of implementation issue, it
1079 might be preferable to treat :TEST and :TEST-NOT as being in some
1080 sense the same &KEY, and effectively take the first test function in
1082 c. Again, a quality of implementation issue: it would be good to issue a
1083 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
1084 WARNING for calls with both :TEST and :TEST-NOT; possibly this
1085 latter should be WARNed about at execute-time too.
1087 216: "debugger confused by frames with invalid number of arguments"
1088 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
1089 leaves the system confused, enough so that (QUIT) no longer works.
1090 It's as though the process of working with the uninitialized slot in
1091 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
1092 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
1094 217: "Bad type operations with FUNCTION types"
1097 * (values-type-union (specifier-type '(function (base-char)))
1098 (specifier-type '(function (integer))))
1100 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
1102 It causes insertion of wrong type assertions into generated
1106 (let ((f (etypecase x
1107 (character #'write-char)
1108 (integer #'write-byte))))
1111 (character (write-char x s))
1112 (integer (write-byte x s)))))
1114 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
1116 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
1117 produce invalid code, but type checking is not accurate. Similar
1118 problems exist with VALUES-TYPE-INTERSECTION.)
1120 218: "VALUES type specifier semantics"
1121 (THE (VALUES ...) ...) in safe code discards extra values.
1123 (defun test (x y) (the (values integer) (truncate x y)))
1127 Sbcl 0.7.9 fails to compile
1129 (multiple-value-call #'list
1130 (the integer (helper))
1133 Type check for INTEGER, the result of which serves as the first
1134 argument of M-V-C, is inserted after evaluation of NIL. So arguments
1135 of M-V-C are pushed in the wrong order. As a temporary workaround
1136 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
1137 would be to put the check between evaluation of arguments, but it
1138 could be tricky to check result types of PROG1, IF etc.
1141 (subtypep 'function '(function)) => nil, t.
1143 233: bugs in constraint propagation
1146 (declare (optimize (speed 2) (safety 3)))
1149 (the double-float x)
1152 (quux y (+ y 2d0) (* y 3d0)))))
1153 (foo 4) => segmentation violation
1155 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
1159 (declaim (optimize (speed 2) (safety 3)))
1161 (if (typep (prog1 x (setq x y)) 'double-float)
1164 (foo 1d0 5) => segmentation violation
1166 235: "type system and inline expansion"
1168 (declaim (ftype (function (cons) number) acc))
1169 (declaim (inline acc))
1171 (the number (car c)))
1174 (values (locally (declare (optimize (safety 0)))
1176 (locally (declare (optimize (safety 3)))
1179 (foo '(nil) '(t)) => NIL, T.
1181 b. (reported by brown on #lisp 2003-01-21)
1184 (declare (optimize (speed 3) (safety 0)))
1185 (declare (notinline mapcar))
1186 (let ((z (mapcar #'car x)))
1189 Without (DECLARE (NOTINLINE MAPCAR)), Python cannot derive that Z is
1192 236: "THE semantics is broken"
1195 (declare (optimize (speed 2) (safety 0)))
1198 (multiple-value-prog1
1200 (unless f (return-from foo 0))))))
1202 (foo #(4) nil) => SEGV
1204 VOP selection thinks that in unsafe code result type assertions
1205 should be valid immediately. (See also bug 233a.)
1207 The similar problem exists for TRULY-THE.
1209 237: "Environment arguments to type functions"
1210 a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
1211 UPGRADED-COMPLEX-PART-TYPE now have an optional environment
1212 argument, but they ignore it completely. This is almost
1213 certainly not correct.
1214 b. Also, the compiler's optimizers for TYPEP have not been informed
1215 about the new argument; consequently, they will not transform
1216 calls of the form (TYPEP 1 'INTEGER NIL), even though this is
1217 just as optimizeable as (TYPEP 1 'INTEGER).
1219 238: "REPL compiler overenthusiasm for CLOS code"
1221 * (defclass foo () ())
1222 * (defmethod bar ((x foo) (foo foo)) (call-next-method))
1223 causes approximately 100 lines of code deletion notes. Some
1224 discussion on this issue happened under the title 'Three "interesting"
1225 bugs in PCL', resulting in a fix for this oververbosity from the
1226 compiler proper; however, the problem persists in the interactor
1227 because the notion of original source is not preserved: for the
1228 compiler, the original source of the above expression is (DEFMETHOD
1229 BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
1230 compiler gets its hands on the code needing compilation from the REPL,
1231 it has been macroexpanded several times.
1233 241: "DEFCLASS mysteriously remembers uninterned accessor names."
1234 (from tonyms on #lisp IRC 2003-02-25)
1235 In sbcl-0.7.12.55, typing
1236 (defclass foo () ((bar :accessor foo-bar)))
1239 (defclass foo () ((bar :accessor foo-bar)))
1240 gives the error message
1241 "#:FOO-BAR already names an ordinary function or a macro."
1242 So it's somehow checking the uninterned old accessor name instead
1243 of the new requested accessor name, which seems broken to me (WHN).
1245 242: "WRITE-SEQUENCE suboptimality"
1246 (observed from clx performance)
1247 In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
1248 (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
1249 (UNSIGNED-BYTE 8) will write to the stream one byte at a time,
1250 rather than writing the sequence in one go, leading to severe
1251 performance degradation.
1253 DEFUNCT CATEGORIES OF BUGS
1255 These labels were used for bugs related to the old IR1 interpreter.
1256 The # values reached 6 before the category was closed down.