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 bogus warnings about undefined functions for magic functions like
106 SB!C::%%DEFUN and SB!C::%DEFCONSTANT when cross-compiling files
107 like src/code/float.lisp. Fixing this will probably require
108 straightening out enough bootstrap consistency issues that
109 the cross-compiler can run with *TYPE-SYSTEM-INITIALIZED*.
110 Instead, the cross-compiler runs in a slightly flaky state
111 which is sane enough to compile SBCL itself, but which is
112 also unstable in several ways, including its inability
113 to really grok function declarations.
115 As of sbcl-0.7.5, sbcl's cross-compiler does run with
116 *TYPE-SYSTEM-INITIALIZED*; however, this bug remains.
119 The "compiling top-level form:" output ought to be condensed.
120 Perhaps any number of such consecutive lines ought to turn into a
121 single "compiling top-level forms:" line.
124 The way that the compiler munges types with arguments together
125 with types with no arguments (in e.g. TYPE-EXPAND) leads to
126 weirdness visible to the user:
127 (DEFTYPE FOO () 'FIXNUM)
129 (TYPEP 11 '(FOO)) => T, which seems weird
130 (TYPEP 11 'FIXNUM) => T
131 (TYPEP 11 '(FIXNUM)) signals an error, as it should
132 The situation is complicated by the presence of Common Lisp types
133 like UNSIGNED-BYTE (which can either be used in list form or alone)
134 so I'm not 100% sure that the behavior above is actually illegal.
135 But I'm 90+% sure, and the following related behavior,
137 treating the bare symbol AND as equivalent to '(AND), is specifically
138 forbidden (by the ANSI specification of the AND type).
141 It would be nice if the
143 (during macroexpansion)
144 said what macroexpansion was at fault, e.g.
146 (during macroexpansion of IN-PACKAGE,
147 during macroexpansion of DEFFOO)
150 (SUBTYPEP '(FUNCTION (T BOOLEAN) NIL)
151 '(FUNCTION (FIXNUM FIXNUM) NIL)) => T, T
152 (Also, when this is fixed, we can enable the code in PROCLAIM which
153 checks for incompatible FTYPE redeclarations.)
156 (I *think* this is a bug. It certainly seems like strange behavior. But
157 the ANSI spec is scary, dark, and deep.. -- WHN)
158 (FORMAT NIL "~,1G" 1.4) => "1. "
159 (FORMAT NIL "~3,1G" 1.4) => "1. "
162 from Marco Antoniotti on cmucl-imp mailing list 1 Mar 2000:
164 (setf (find-class 'ccc1) (find-class 'ccc))
165 (defmethod zut ((c ccc1)) 123)
166 In sbcl-0.7.1.13, this gives an error,
167 There is no class named CCC1.
168 DTC's recommended workaround from the mailing list 3 Mar 2000:
169 (setf (pcl::find-class 'ccc1) (pcl::find-class 'ccc))
172 Sometimes (SB-EXT:QUIT) fails with
173 Argh! maximum interrupt nesting depth (4096) exceeded, exiting
174 Process inferior-lisp exited abnormally with code 1
175 I haven't noticed a repeatable case of this yet.
178 The printer doesn't report closures very well. This is true in
182 #<Closure Over Function "DEFUN STRUCTURE-SLOT-ACCESSOR" {134D1A1}>
183 It would be nice to make closures have a settable name slot,
184 and make things like DEFSTRUCT and FLET, which create closures,
185 set helpful values into this slot.
188 And as long as we're wishing, it would be awfully nice if INSPECT could
189 also report on closures, telling about the values of the bound variables.
192 The compiler assumes that any time a function of declared FTYPE
193 doesn't signal an error, its arguments were of the declared type.
194 E.g. compiling and loading
195 (DECLAIM (OPTIMIZE (SAFETY 3)))
196 (DEFUN FACTORIAL (X) (GAMMA (1+ X)))
198 (DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
200 (COND ((> (FACTORIAL X) 1.0E6)
201 (FORMAT T "too big~%"))
203 (FORMAT T "exactly ~S~%" (FACTORIAL X)))
205 (FORMAT T "approximately ~S~%" (FACTORIAL X)))))
208 will cause the INTEGERP case to be selected, giving bogus output a la
210 This violates the "declarations are assertions" principle.
211 According to the ANSI spec, in the section "System Class FUNCTION",
212 this is a case of "lying to the compiler", but the lying is done
213 by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
214 not by the unexpectedly general definition of FACTORIAL. In any case,
215 "declarations are assertions" means that lying to the compiler should
216 cause an error to be signalled, and should not cause a bogus
217 result to be returned. Thus, the compiler should not assume
218 that arbitrary functions check their argument types. (It might
219 make sense to add another flag (CHECKED?) to DEFKNOWN to
220 identify functions which *do* check their argument types.)
221 (Also, verify that the compiler handles declared function
222 return types as assertions.)
225 TYPEP of VALUES types is sometimes implemented very inefficiently, e.g. in
226 (DEFTYPE INDEXOID () '(INTEGER 0 1000))
228 (DECLARE (TYPE INDEXOID X))
229 (THE (VALUES INDEXOID)
231 where the implementation of the type check in function FOO
232 includes a full call to %TYPEP. There are also some fundamental problems
233 with the interpretation of VALUES types (inherited from CMU CL, and
234 from the ANSI CL standard) as discussed on the cmucl-imp@cons.org
235 mailing list, e.g. in Robert Maclachlan's post of 21 Jun 2000.
238 The definitions of SIGCONTEXT-FLOAT-REGISTER and
239 %SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
240 supported on FreeBSD because the floating point state is not saved,
241 but at least as of FreeBSD 4.0, the floating point state *is* saved,
242 so they could be supported after all. Very likely
243 SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
246 (as discussed by Douglas Crosher on the cmucl-imp mailing list ca.
247 Aug. 10, 2000): CMUCL currently interprets 'member as '(member); same
248 issue with 'union, 'and, 'or etc. So even though according to the
249 ANSI spec, bare 'MEMBER, 'AND, and 'OR are not legal types, CMUCL
250 (and now SBCL) interpret them as legal types.
253 a slew of floating-point-related errors reported by Peter Van Eynde
255 b: SBCL's value for LEAST-POSITIVE-SHORT-FLOAT is bogus, and
256 should probably be 1.4012985e-45. In SBCL,
257 (/ LEAST-POSITIVE-SHORT-FLOAT 2) returns a number smaller
258 than LEAST-POSITIVE-SHORT-FLOAT. Similar problems
259 exist for LEAST-NEGATIVE-SHORT-FLOAT, LEAST-POSITIVE-LONG-FLOAT,
260 and LEAST-NEGATIVE-LONG-FLOAT.
261 c: Many expressions generate floating infinity on x86/Linux:
266 PVE's regression tests want them to raise errors. sbcl-0.7.0.5
267 on x86/Linux generates the infinities instead. That might or
268 might not be conforming behavior, but it's also inconsistent,
269 which is almost certainly wrong. (Inconsistency: (/ 1 0.0)
270 should give the same result as (/ 1.0 0.0), but instead (/ 1 0.0)
271 generates SINGLE-FLOAT-POSITIVE-INFINITY and (/ 1.0 0.0)
273 d: (in section12.erg) various forms a la
274 (FLOAT 1 DOUBLE-FLOAT-EPSILON)
275 don't give the right behavior.
278 type safety errors reported by Peter Van Eynde July 25, 2000:
279 c: (COERCE 'AND 'FUNCTION) returns something related to
280 (MACRO-FUNCTION 'AND), but ANSI says it should raise an error.
281 k: READ-BYTE is supposed to signal TYPE-ERROR when its argument is
282 not a binary input stream, but instead cheerfully reads from
283 character streams, e.g. (MAKE-STRING-INPUT-STREAM "abc").
286 DEFCLASS bugs reported by Peter Van Eynde July 25, 2000:
287 d: (DEFGENERIC IF (X)) should signal a PROGRAM-ERROR, but instead
288 causes a COMPILER-ERROR.
291 miscellaneous errors reported by Peter Van Eynde July 25, 2000:
293 (DEFGENERIC FOO02 (X))
294 (DEFMETHOD FOO02 ((X NUMBER)) T)
295 (LET ((M (FIND-METHOD (FUNCTION FOO02)
297 (LIST (FIND-CLASS (QUOTE NUMBER))))))
298 (REMOVE-METHOD (FUNCTION FOO02) M)
299 (DEFGENERIC FOO03 (X))
300 (ADD-METHOD (FUNCTION FOO03) M)))
301 should give an error, but SBCL allows it.
304 It has been reported (e.g. by Peter Van Eynde) that there are
305 several metaobject protocol "errors". (In order to fix them, we might
306 need to document exactly what metaobject protocol specification
307 we're following -- the current code is just inherited from PCL.)
310 The debugger LIST-LOCATIONS command doesn't work properly.
313 Compiling and loading
314 (DEFUN FAIL (X) (THROW 'FAIL-TAG X))
316 then requesting a BACKTRACE at the debugger prompt gives no information
317 about where in the user program the problem occurred.
320 Paul Werkowski wrote on cmucl-imp@cons.org 2000-11-15
321 I am looking into this problem that showed up on the cmucl-help
322 list. It seems to me that the "implementation specific environment
323 hacking functions" found in pcl/walker.lisp are completely messed
324 up. The good thing is that they appear to be barely used within
325 PCL and the munged environment object is passed to cmucl only
326 in calls to macroexpand-1, which is probably why this case fails.
327 SBCL uses essentially the same code, so if the environment hacking
328 is screwed up, it affects us too.
331 Using the pretty-printer from the command prompt gives funny
332 results, apparently because the pretty-printer doesn't know
333 about user's command input, including the user's carriage return
334 that the user, and therefore the pretty-printer thinks that
335 the new output block should start indented 2 or more characters
336 rightward of the correct location.
339 As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
340 and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
341 crashes SBCL. In general tracing anything which is used in the
342 implementation of TRACE is likely to have the same problem.
345 As reported by Martin Atzmueller on sbcl-devel 26 Dec 2000,
346 ANSI says that WITH-OUTPUT-TO-STRING should have a keyword
347 :ELEMENT-TYPE, but in sbcl-0.6.9 this is not defined for
348 WITH-OUTPUT-TO-STRING.
351 ANSI says in one place that type declarations can be abbreviated even
352 when the type name is not a symbol, e.g.
353 (DECLAIM ((VECTOR T) *FOOVECTOR*))
354 SBCL doesn't support this. But ANSI says in another place that this
355 isn't allowed. So it's not clear this is a bug after all. (See the
356 e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
360 as pointed out by Dan Barlow on sbcl-devel 2000-07-02:
361 The PICK-TEMPORARY-FILE-NAME utility used by LOAD-FOREIGN uses
362 an easily guessable temporary filename in a way which might open
363 applications using LOAD-FOREIGN to hijacking by malicious users
364 on the same machine. Incantations for doing this safely are
365 floating around the net in various "how to write secure programs
366 despite Unix" documents, and it would be good to (1) fix this in
367 LOAD-FOREIGN, and (2) hunt for any other code which uses temporary
368 files and make it share the same new safe logic.
370 (partially alleviated in sbcl-0.7.9.32 by a fix by Matthew Danish to
371 make the temporary filename less easily guessable)
374 Functions are assigned names based on the context in which they're
375 defined. This is less than ideal for the functions which are
376 used to implement CLOS methods. E.g. the output of
377 (DESCRIBE 'PRINT-OBJECT) lists functions like
378 #<FUNCTION "DEF!STRUCT (TRACE-INFO (:MAKE-LOAD-FORM-FUN SB-KERNEL:JUST-DUMP-IT-NORMALLY) (:PRINT-OBJECT #))" {1020E49}>
380 #<FUNCTION "MACROLET ((FORCE-DELAYED-DEF!METHODS NIL #))" {1242871}>
381 It would be better if these functions' names always identified
382 them as methods, and identified their generic functions and
386 RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
387 RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
388 the mean of the distribution can be systematically O(0.1%) wrong.
389 Just increasing R-I-E-B is probably not a good solution, since
390 it would decrease efficiency more than is probably necessary. Perhaps
391 using some sort of accept/reject method would be better.
394 Internally the compiler sometimes evaluates
395 (sb-kernel:type/= (specifier-type '*) (specifier-type t))
396 (I stumbled across this when I added an
397 (assert (not (eq type1 *wild-type*)))
398 in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
399 in a type context should probably be translated to T, and so it's
400 probably wrong to ask whether it's equal to the T type and then (using
401 the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
402 (I haven't tried to investigate this bug enough to guess whether
403 there might be any user-level symptoms.)
405 In fact, the type system is likely to depend on this inequality not
406 holding... * is not equivalent to T in many cases, such as
407 (VECTOR *) /= (VECTOR T).
410 Inconsistencies between derived and declared VALUES return types for
411 DEFUN aren't checked very well. E.g. the logic which successfully
412 catches problems like
413 (declaim (ftype (function (fixnum) float) foo))
415 (declare (type integer x))
416 (values x)) ; wrong return type, detected, gives warning, good!
418 (declaim (ftype (function (t) (values t t)) bar))
420 (values x)) ; wrong number of return values, no warning, bad!
421 The cause of this is seems to be that (1) the internal function
422 VALUES-TYPES-EQUAL-OR-INTERSECT used to make the check handles its
423 arguments symmetrically, and (2) when the type checking code was
424 written back when when SBCL's code was still CMU CL, the intent
426 (declaim (ftype (function (t) t) bar))
428 (values x x)) ; wrong number of return values; should give warning?
429 not be warned for, because a two-valued return value is considered
430 to be compatible with callers who expects a single value to be
431 returned. That intent is probably not appropriate for modern ANSI
432 Common Lisp, but fixing this might be complicated because of other
433 divergences between auld-style and new-style handling of
434 multiple-VALUES types. (Some issues related to this were discussed
435 on cmucl-imp at some length sometime in 2000.)
438 The facility for dumping a running Lisp image to disk gets confused
439 when run without the PURIFY option, and creates an unnecessarily large
440 core file (apparently representing memory usage up to the previous
441 high-water mark). Moreover, when the file is loaded, it confuses the
442 GC, so that thereafter memory usage can never be reduced below that
446 In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
447 CL), out-of-line structure slot setters are horribly inefficient
448 whenever the type of the slot is declared, because out-of-line
449 structure slot setters are implemented as closures to save space,
450 so the compiler doesn't compile the type test into code, but
451 instead just saves the type in a lexical closure and interprets it
453 A proper solution involves deciding whether it's really worth
454 saving space by implementing structure slot accessors as closures.
455 (If it's not worth it, the problem vanishes automatically. If it
456 is worth it, there are hacks we could use to force type tests to
457 be compiled anyway, and even shared. E.g. we could implement
458 an EQUAL hash table mapping from types to compiled type tests,
459 and save the appropriate compiled type test as part of each lexical
460 closure; or we could make the lexical closures be placeholders
461 which overwrite their old definition as a lexical closure with
462 a new compiled definition the first time that they're called.)
463 As a workaround for the problem, #'(SETF FOO) expressions can
464 be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
465 (defmacro efficient-setf-function (place-function-name)
466 (or #+sbcl (and (sb-impl::info :function :accessor-for place-function-name)
467 ;; a workaround for the problem, encouraging the
468 ;; inline expansion of the structure accessor, so
469 ;; that the compiler can optimize its type test
470 (let ((new-value (gensym "NEW-VALUE-"))
471 (structure-value (gensym "STRUCTURE-VALUE-")))
472 `(lambda (,new-value ,structure-value)
473 (setf (,place-function-name ,structure-value)
475 ;; no problem, can just use the ordinary expansion
476 `(function (setf ,place-function-name))))
479 There's apparently a bug in CEILING optimization which caused
480 Douglas Crosher to patch the CMU CL version. Martin Atzmueller
481 applied the patches to SBCL and they didn't seem to cause problems
482 (as reported sbcl-devel 2001-05-04). However, since the patches
483 modify nontrivial code which was apparently written incorrectly
484 the first time around, until regression tests are written I'm not
485 comfortable merging the patches in the CVS version of SBCL.
488 (TIME (ROOM T)) reports more than 200 Mbytes consed even for
489 a clean, just-started SBCL system. And it seems to be right:
490 (ROOM T) can bring a small computer to its knees for a *long*
491 time trying to GC afterwards. Surely there's some more economical
492 way to implement (ROOM T).
495 When the compiler inline expands functions, it may be that different
496 kinds of return values are generated from different code branches.
497 E.g. an inline expansion of POSITION generates integer results
498 from one branch, and NIL results from another. When that inline
499 expansion is used in a context where only one of those results
502 (aref *a1* (position x *a2*)))
503 and the compiler can't prove that the unacceptable branch is
504 never taken, then bogus type mismatch warnings can be generated.
505 If you need to suppress the type mismatch warnings, you can
506 suppress the inline expansion,
508 #+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
509 (aref *a1* (position x *a2*)))
510 or, sometimes, suppress them by declaring the result to be of an
513 (aref *a1* (the integer (position x *a2*))))
515 This is not a new compiler problem in 0.7.0, but the new compiler
516 transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
517 more conspicuous. If you don't need performance from these functions,
518 and the bogus warnings are a nuisance for you, you can return to
519 your pre-0.7.0 state of grace with
520 #+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
523 as reported by Eric Marsden on cmucl-imp@cons.org 2001-08-14:
524 (= (FLOAT 1 DOUBLE-FLOAT-EPSILON)
525 (+ (FLOAT 1 DOUBLE-FLOAT-EPSILON) DOUBLE-FLOAT-EPSILON)) => T
526 when of course it should be NIL. (He says it only fails for X86,
527 not SPARC; dunno about Alpha.)
529 Also, "the same problem exists for LONG-FLOAT-EPSILON,
530 DOUBLE-FLOAT-NEGATIVE-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON (though
531 for the -negative- the + is replaced by a - in the test)."
533 Raymond Toy comments that this is tricky on the X86 since its FPU
534 uses 80-bit precision internally.
537 Even in sbcl-0.pre7.x, which is supposed to be free of the old
538 non-ANSI behavior of treating the function return type inferred
539 from the current function definition as a declaration of the
540 return type from any function of that name, the return type of NIL
541 is attached to FOO in 120a above, and used to optimize code which
545 As of version 0.pre7.14, SBCL's implementation of MACROLET makes
546 the entire lexical environment at the point of MACROLET available
547 in the bodies of the macroexpander functions. In particular, it
548 allows the function bodies (which run at compile time) to try to
549 access lexical variables (which are only defined at runtime).
550 It doesn't even issue a warning, which is bad.
552 The SBCL behavior arguably conforms to the ANSI spec (since the
553 spec says that the behavior is undefined, ergo anything conforms).
554 However, it would be better to issue a compile-time error.
555 Unfortunately I (WHN) don't see any simple way to detect this
556 condition in order to issue such an error, so for the meantime
557 SBCL just does this weird broken "conforming" thing.
559 The ANSI standard says, in the definition of the special operator
561 The macro-expansion functions defined by MACROLET are defined
562 in the lexical environment in which the MACROLET form appears.
563 Declarations and MACROLET and SYMBOL-MACROLET definitions affect
564 the local macro definitions in a MACROLET, but the consequences
565 are undefined if the local macro definitions reference any
566 local variable or function bindings that are visible in that
568 Then it seems to contradict itself by giving the example
570 (macrolet ((fudge (z)
571 ;The parameters x and flag are not accessible
572 ; at this point; a reference to flag would be to
573 ; the global variable of that name.
574 ` (if flag (* ,z ,z) ,z)))
575 ;The parameters x and flag are accessible here.
579 The comment "a reference to flag would be to the global variable
580 of the same name" sounds like good behavior for the system to have.
581 but actual specification quoted above says that the actual behavior
584 (Since 0.7.8.23 macroexpanders are defined in a restricted version
585 of the lexical environment, containing no lexical variables and
586 functions, which seems to conform to ANSI and CLtL2, but signalling
587 a STYLE-WARNING for references to variables similar to locals might
591 (as reported by Gabe Garza on cmucl-help 2001-09-21)
593 (defun test-pred (x y)
597 (func (lambda () x)))
598 (print (eq func func))
599 (print (test-pred func func))
600 (delete func (list func))))
601 Now calling (TEST-CASE) gives output
604 (#<FUNCTION {500A9EF9}>)
605 Evidently Python thinks of the lambda as a code transformation so
606 much that it forgets that it's also an object.
609 Ideally, uninterning a symbol would allow it, and its associated
610 FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
611 at least as of sbcl-0.7.0, this isn't the case. Information about
612 FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
613 essentially in ordinary (non-weak) hash tables keyed by symbols.
614 Thus, once a system has an entry in this system, it tends to live
615 forever, even when it is uninterned and all other references to it
618 141: "pretty printing and backquote"
621 ``(FOO SB-IMPL::BACKQ-COMMA-AT S)
624 * (write '`(, .ala.) :readably t :pretty t)
627 (note the space between the comma and the point)
631 (reported by Jesse Bouwman 2001-10-24 through the unfortunately
632 prominent SourceForge web/db bug tracking system, which is
633 unfortunately not a reliable way to get a timely response from
634 the SBCL maintainers)
635 In the course of trying to build a test case for an
636 application error, I encountered this behavior:
637 If you start up sbcl, and then lay on CTRL-C for a
638 minute or two, the lisp process will eventually say:
639 %PRIMITIVE HALT called; the party is over.
640 and throw you into the monitor. If I start up lisp,
641 attach to the process with strace, and then do the same
642 (abusive) thing, I get instead:
643 access failure in heap page not marked as write-protected
644 and the monitor again. I don't know enough to have the
645 faintest idea of what is going on here.
646 This is with sbcl 6.12, uname -a reports:
647 Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
648 I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
649 under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
650 it took more than two minutes (but less than five) for me.
653 (This was once known as IR1-4, but it lived on even after the
654 IR1 interpreter went to the big bit bucket in the sky.)
655 The system accepts DECLAIM in most places where DECLARE would be
656 accepted, without even issuing a warning. ANSI allows this, but since
657 it's fairly easy to mistype DECLAIM instead of DECLARE, and the
658 meaning is rather different, and it's unlikely that the user
659 has a good reason for doing DECLAIM not at top level, it would be
660 good to issue a STYLE-WARNING when this happens. A possible
661 fix would be to issue STYLE-WARNINGs for DECLAIMs not at top level,
662 or perhaps to issue STYLE-WARNINGs for any EVAL-WHEN not at top level.
663 [This is considered an IR1-interpreter-related bug because until
664 EVAL-WHEN is rewritten, which won't happen until after the IR1
665 interpreter is gone, the system's notion of what's a top-level form
666 and what's not will remain too confused to fix this problem.]
669 ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
670 FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
671 COMPLEX implementation didn't deal with this, and hasn't been
672 upgraded to do so. (This doesn't seem to be a high priority
673 conformance problem, since seems hard to construct useful code
677 Floating point errors are reported poorly. E.g. on x86 OpenBSD
680 debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
681 An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
682 No traps are enabled? How can this be?
683 It should be possible to be much more specific (overflow, division
684 by zero, etc.) and of course the "How can this be?" should be fixable.
686 See also bugs #45.c and #183
689 In sbcl-0.7.1.3 on x86, COMPILE-FILE on the file
690 (in-package :cl-user)
693 (defstruct foo bar bletch)
695 (labels ((kidify1 (kid)
703 (declare (inline kid-frob))
706 (the simple-vector (foo-bar perd)))))
708 debugger invoked on condition of type TYPE-ERROR:
709 The value NIL is not of type SB-C::NODE.
710 The location of this failure has moved around as various related
711 issues were cleaned up. As of sbcl-0.7.1.9, it occurs in
712 NODE-BLOCK called by LAMBDA-COMPONENT called by IR2-CONVERT-CLOSURE.
714 (Python LET-converts KIDIFY1 into KID-FROB, then tries to inline
715 expand KID-FROB into %ZEEP. Having partially done it, it sees a call
716 of KIDIFY1, which already does not exist. So it gives up on
717 expansion, leaving garbage consisting of infinished blocks of the
718 partially converted function.)
721 Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
722 UPGRADED-COMPLEX-PART-TYPE should have an optional environment argument.
723 (reported by Alexey Dejneka sbcl-devel 2002-04-12)
726 (reported by Robert E. Brown 2002-04-16)
727 When a function is called with too few arguments, causing the
728 debugger to be entered, the uninitialized slots in the bad call frame
729 seem to cause GCish problems, being interpreted as tagged data even
730 though they're not. In particular, executing ROOM in the
731 debugger at that point causes AVER failures:
734 * (lisp-implementation-version)
740 failed AVER: "(SAP= CURRENT END)"
741 (Christophe Rhodes reports that this doesn't occur on the SPARC, which
742 isn't too surprising since there are many differences in stack
743 implementation and GC conservatism between the X86 and other ports.)
746 In sbcl-0.7.3.11, compiling the (illegal) code
747 (in-package :cl-user)
748 (defmethod prove ((uustk uustk))
751 gives the (not terribly clear) error message
753 ; (during macroexpansion of (DEFMETHOD PROVE ...))
754 ; can't get template for (FROB NIL NIL)
755 The problem seems to be that the code walker used by the DEFMETHOD
756 macro is unhappy with the illegal syntax in the method body, and
757 is giving an unclear error message.
760 The compiler sometimes tries to constant-fold expressions before
761 it checks to see whether they can be reached. This can lead to
762 bogus warnings about errors in the constant folding, e.g. in code
765 (WRITE-STRING (> X 0) "+" "0"))
766 compiled in a context where the compiler can prove that X is NIL,
767 and the compiler complains that (> X 0) causes a type error because
768 NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
769 caused a full WARNING, which made the bug really annoying because then
770 COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
771 code. Since then the warning has been downgraded to STYLE-WARNING,
772 so it's still a bug but at least it's a little less annoying.
774 183: "IEEE floating point issues"
775 Even where floating point handling is being dealt with relatively
776 well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
777 accrued-exceptions and current-exceptions part of the fp control
778 word don't seem to bear much relation to reality. E.g. on
782 debugger invoked on condition of type DIVISION-BY-ZERO:
783 arithmetic error DIVISION-BY-ZERO signalled
784 0] (sb-vm::get-floating-point-modes)
786 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
787 :ROUNDING-MODE :NEAREST
788 :CURRENT-EXCEPTIONS NIL
789 :ACCRUED-EXCEPTIONS (:INEXACT)
792 * (sb-vm::get-floating-point-modes)
793 (:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
794 :ROUNDING-MODE :NEAREST
795 :CURRENT-EXCEPTIONS (:INEXACT)
796 :ACCRUED-EXCEPTIONS (:INEXACT)
799 187: "type inference confusion around DEFTRANSFORM time"
800 (reported even more verbosely on sbcl-devel 2002-06-28 as "strange
801 bug in DEFTRANSFORM")
802 After the file below is compiled and loaded in sbcl-0.7.5, executing
803 (TCX (MAKE-ARRAY 4 :FILL-POINTER 2) 0)
804 at the REPL returns an adjustable vector, which is wrong. Presumably
805 somehow the DERIVE-TYPE information for the output values of %WAD is
806 being mispropagated as a type constraint on the input values of %WAD,
807 and so causing the type test to be optimized away. It's unclear how
808 hand-expanding the DEFTRANSFORM would change this, but it suggests
809 the DEFTRANSFORM machinery (or at least the way DEFTRANSFORMs are
810 invoked at a particular phase) is involved.
811 (cl:in-package :sb-c)
812 (eval-when (:compile-toplevel)
813 ;;; standin for %DATA-VECTOR-AND-INDEX
814 (defknown %dvai (array index)
816 (foldable flushable))
817 (deftransform %dvai ((array index)
821 (let* ((atype (continuation-type array))
822 (eltype (array-type-specialized-element-type atype)))
823 (when (eq eltype *wild-type*)
824 (give-up-ir1-transform
825 "specialized array element type not known at compile-time"))
826 (when (not (array-type-complexp atype))
827 (give-up-ir1-transform "SIMPLE array!"))
828 `(if (array-header-p array)
829 (%wad array index nil)
830 (values array index))))
831 ;;; standin for %WITH-ARRAY-DATA
832 (defknown %wad (array index (or index null))
833 (values (simple-array * (*)) index index index)
834 (foldable flushable))
835 ;;; (Commenting out this optimizer causes the bug to go away.)
836 (defoptimizer (%wad derive-type) ((array start end))
837 (let ((atype (continuation-type array)))
838 (when (array-type-p atype)
839 (values-specifier-type
840 `(values (simple-array ,(type-specifier
841 (array-type-specialized-element-type atype))
843 index index index)))))
845 (defun %wad (array start end)
846 (format t "~&in %WAD~%")
847 (%with-array-data array start end))
848 (cl:in-package :cl-user)
850 (declare (type (vector t) v))
851 (declare (notinline sb-kernel::%with-array-data))
852 ;; (Hand-expending DEFTRANSFORM %DVAI here also causes the bug to
856 188: "compiler performance fiasco involving type inference and UNION-TYPE"
857 (In sbcl-0.7.6.10, DEFTRANSFORM CONCATENATE was commented out until this
858 bug could be fixed properly, so you won't see the bug unless you restore
859 the DEFTRANSFORM by hand.) In sbcl-0.7.5.11 on a 700 MHz Pentium III,
863 (declare (optimize (safety 3)))
864 (declare (optimize (compilation-speed 2)))
865 (declare (optimize (speed 1) (debug 1) (space 1)))
866 (let ((fn "if-this-file-exists-the-universe-is-strange"))
867 (load fn :if-does-not-exist nil)
868 (load (concatenate 'string fn ".lisp") :if-does-not-exist nil)
869 (load (concatenate 'string fn ".fasl") :if-does-not-exist nil)
870 (load (concatenate 'string fn ".misc-garbage")
871 :if-does-not-exist nil)))))
873 134.552 seconds of real time
874 133.35156 seconds of user run time
875 0.03125 seconds of system run time
876 [Run times include 2.787 seconds GC run time.]
878 246883368 bytes consed.
879 BACKTRACE from Ctrl-C in the compilation shows that the compiler is
880 thinking about type relationships involving types like
882 (OR (INTEGER 576 576)
893 190: "PPC/Linux pipe? buffer? bug"
894 In sbcl-0.7.6, the run-program.test.sh test script sometimes hangs
895 on the PPC/Linux platform, waiting for a zombie env process. This
896 is a classic symptom of buffer filling and deadlock, but it seems
897 only sporadically reproducible.
899 191: "Miscellaneous PCL deficiencies"
900 (reported by Alexey Dejneka sbcl-devel 2002-08-04)
901 a. DEFCLASS does not inform the compiler about generated
902 functions. Compiling a file with
906 (WITH-SLOTS (A-CLASS-X) A
908 results in a STYLE-WARNING:
910 SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
912 APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
913 Mai points out that the declamation of functions is in fact
914 incorrect in some cases (most notably for structure
915 classes). This means that at present erroneous attempts to use
916 WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
917 won't get the corresponding STYLE-WARNING.
918 c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
919 lists and &KEY arguments) do not signal errors when they should.
921 192: "Python treats free type declarations as promises."
922 b. What seemed like the same fundamental problem as bug 192a, but
923 was not fixed by the same (APD "more strict type checking
924 sbcl-devel 2002-08-97) patch:
925 (DOTIMES (I ...) (DOTIMES (J ...) (DECLARE ...) ...)):
926 (declaim (optimize (speed 1) (safety 3)))
927 (defun trust-assertion (i)
929 (declare (type (mod 4) i)) ; when commented out, behavior changes!
932 (trust-assertion 6) ; prints nothing unless DECLARE is commented out
937 (locally (declare (type fixnum x y))
941 194: "no error from (THE REAL '(1 2 3)) in some cases"
944 (multiple-value-prog1 (progn (the real '(1 2 3))))
945 returns (1 2 3) instead of signalling an error. This was fixed by
946 APD's "more strict type checking patch", but although the fixed
947 code (in sbcl-0.7.7.19) works (signals TYPE-ERROR) interactively,
948 it's difficult to write a regression test for it, because
949 (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
950 still returns (1 2 3).
952 b. (IGNORE-ERRORS (MULTIPLE-VALUE-PROG1 (PROGN (THE REAL '(1 2 3)))))
953 returns (1 2 3). (As above, this shows up when writing regression
954 tests for fixed-ness of part a.)
955 c. Also in sbcl-0.7.7.9, (IGNORE-ERRORS (THE REAL '(1 2 3))) => (1 2 3).
959 (arg2 (identity (the real #(1 2 3)))))
960 (if (< arg1 arg2) arg1 arg2))))
962 but putting the same expression inside (DEFUN FOO () ...),
965 * Actually this entry is probably multiple bugs, as
966 Alexey Dejneka commented on sbcl-devel 2002-09-03:)
967 I don't think that placing these two bugs in one entry is
968 a good idea: they have different explanations. The second
969 (min 1 nil) is caused by flushing of unused code--IDENTITY
970 can do nothing with it. So it is really bug 122. The first
971 (min nil) is due to M-V-PROG1: substituting a continuation
972 for the result, it forgets about type assertion. The purpose
973 of IDENTITY is to save the restricted continuation from
974 inaccurate transformations.
975 * Alexey Dejneka pointed out that
976 (IGNORE-ERRORS (IDENTITY (THE REAL '(1 2 3))))
978 (IGNORE-ERRORS (VALUES (THE REAL '(1 2 3))))
981 201: "Incautious type inference from compound CONS types"
982 (reported by APD sbcl-devel 2002-09-17)
984 (LET ((Y (CAR (THE (CONS INTEGER *) X))))
986 (FORMAT NIL "~S IS ~S, Y = ~S"
993 (FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
996 Compiler does not check THEs on unused values, e.g. in
998 (progn (the real (list 1)) t)
1000 This situation may appear during optimizing away degenerate cases of
1001 certain functions: see bug 192b.
1003 205: "environment issues in cross compiler"
1004 (These bugs have no impact on user code, but should be fixed or
1006 a. Macroexpanders introduced with MACROLET are defined in the null
1007 lexical environment.
1008 b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
1009 the null lexical environment.
1010 c. The cross-compiler cannot inline functions defined in a non-null
1011 lexical environment.
1013 206: ":SB-FLUID feature broken"
1014 (reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
1015 Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
1018 207: "poorly distributed SXHASH results for compound data"
1019 SBCL's SXHASH could probably try a little harder. ANSI: "the
1020 intent is that an implementation should make a good-faith
1021 effort to produce hash-codes that are well distributed
1022 within the range of non-negative fixnums". But
1023 (let ((hits (make-hash-table)))
1026 (let* ((ij (cons i j))
1027 (newlist (push ij (gethash (sxhash ij) hits))))
1029 (format t "~&collision: ~S~%" newlist))))))
1030 reports lots of collisions in sbcl-0.7.8. A stronger MIX function
1031 would be an obvious way of fix. Maybe it would be acceptably efficient
1032 to redo MIX using a lookup into a 256-entry s-box containing
1033 29-bit pseudorandom numbers?
1035 208: "package confusion in PCL handling of structure slot handlers"
1036 In sbcl-0.7.8 compiling and loading
1038 (defstruct foo (slot (error "missing")) :type list :read-only t)
1039 (defmethod print-object ((foo foo) stream) (print nil stream))
1040 causes CERROR "attempting to modify a symbol in the COMMON-LISP
1041 package: FOO-SLOT". (This is fairly bad code, but still it's hard
1042 to see that it should cause symbols to be interned in the CL package.)
1044 211: "keywords processing"
1045 a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
1046 number of keyword arguments.
1049 (flet ((foo (&key y) (list y)))
1050 (list (foo :y 1 :y 2)))
1052 issues confusing message
1057 ; caught STYLE-WARNING:
1058 ; The variable #:G15 is defined but never used.
1060 212: "Sequence functions and circular arguments"
1061 COERCE, MERGE and CONCATENATE go into an infinite loop when given
1062 circular arguments; it would be good for the user if they could be
1063 given an error instead (ANSI 17.1.1 allows this behaviour on the part
1064 of the implementation, as conforming code cannot give non-proper
1065 sequences to these functions. MAP also has this problem (and
1066 solution), though arguably the convenience of being able to do
1067 (MAP 'LIST '+ FOO '#1=(1 . #1#))
1068 might be classed as more important (though signalling an error when
1069 all of the arguments are circular is probably desireable).
1071 213: "Sequence functions and type checking"
1072 a. MAKE-SEQUENCE, COERCE, MERGE and CONCATENATE cannot deal with
1073 various complicated, though recognizeable, CONS types [e.g.
1074 (CONS * (CONS * NULL))
1075 which according to ANSI should be recognized] (and, in SAFETY 3
1076 code, should return a list of LENGTH 2 or signal an error)
1077 b. MAP, when given a type argument that is SUBTYPEP LIST, does not
1078 check that it will return a sequence of the given type. Fixing
1079 it along the same lines as the others (cf. work done around
1080 sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
1081 entirely straightforward.
1082 c. All of these functions will silently accept a type of the form
1084 whether or not the return value is of this type. This is
1085 probably permitted by ANSI (see "Exceptional Situations" under
1086 ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
1087 know about this escape clause, so code of the form
1088 (INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
1089 can erroneously return T.
1092 SBCL 0.6.12.43 fails to compile
1095 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1096 (flet ((foo (&key (x :vx x-p)) (list x x-p)))
1099 or a more simple example:
1102 (declare (optimize (inhibit-warnings 0) (compilation-speed 2)))
1103 (lambda (x) (declare (fixnum x)) (if (< x 0) 0 (1- x))))
1105 215: ":TEST-NOT handling by functions"
1106 a. FIND and POSITION currently signal errors when given non-NIL for
1107 both their :TEST and (deprecated) :TEST-NOT arguments, but by
1108 ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
1109 means that the effect is "unpredictable but harmless". It's not
1110 clear what that actually means; it may preclude conforming
1111 implementations from signalling errors.
1112 b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
1113 when conflict occurs. As a quality of implementation issue, it
1114 might be preferable to treat :TEST and :TEST-NOT as being in some
1115 sense the same &KEY, and effectively take the first test function in
1117 c. Again, a quality of implementation issue: it would be good to issue a
1118 STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
1119 WARNING for calls with both :TEST and :TEST-NOT; possibly this
1120 latter should be WARNed about at execute-time too.
1122 216: "debugger confused by frames with invalid number of arguments"
1123 In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
1124 leaves the system confused, enough so that (QUIT) no longer works.
1125 It's as though the process of working with the uninitialized slot in
1126 the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
1127 not be the actual problem. (CMU CL 18c doesn't have problems with this.)
1129 217: "Bad type operations with FUNCTION types"
1132 * (values-type-union (specifier-type '(function (base-char)))
1133 (specifier-type '(function (integer))))
1135 #<FUN-TYPE (FUNCTION (BASE-CHAR) *)>
1137 It causes insertion of wrong type assertions into generated
1141 (let ((f (etypecase x
1142 (character #'write-char)
1143 (integer #'write-byte))))
1146 (character (write-char x s))
1147 (integer (write-byte x s)))))
1149 Then (FOO #\1 *STANDARD-OUTPUT*) signals type error.
1151 (In 0.7.9.1 the result type is (FUNCTION * *), so Python does not
1152 produce invalid code, but type checking is not accurate. Similar
1153 problems exist with VALUES-TYPE-INTERSECTION.)
1155 218: "VALUES type specifier semantics"
1156 (THE (VALUES ...) ...) in safe code discards extra values.
1158 (defun test (x y) (the (values integer) (truncate x y)))
1162 Sbcl 0.7.9 fails to compile
1164 (multiple-value-call #'list
1165 (the integer (helper))
1168 Type check for INTEGER, the result of which serves as the first
1169 argument of M-V-C, is inserted after evaluation of NIL. So arguments
1170 of M-V-C are pushed in the wrong order. As a temporary workaround
1171 type checking was disabled for M-V-Cs in 0.7.9.13. A better solution
1172 would be to put the check between evaluation of arguments, but it
1173 could be tricky to check result types of PROG1, IF etc.
1175 228: "function-lambda-expression problems"
1176 in sbcl-0.7.9.6x, from the REPL:
1177 * (progn (declaim (inline foo)) (defun foo (x) x))
1179 * (function-lambda-expression #'foo)
1180 (SB-C:LAMBDA-WITH-LEXENV NIL NIL NIL (X) (BLOCK FOO X)), NIL, FOO
1181 but this first return value is not suitable for input to FUNCTION or
1182 COMPILE, as required by ANSI.
1185 (subtypep 'function '(function)) => nil, t.
1187 233: bugs in constraint propagation
1190 (declare (optimize (speed 2) (safety 3)))
1193 (the double-float x)
1196 (quux y (+ y 2d0) (* y 3d0)))))
1197 (foo 4) => segmentation violation
1199 (see usage of CONTINUATION-ASSERTED-TYPE in USE-RESULT-CONSTRAINTS)
1202 (declaim (optimize (speed 2) (safety 3)))
1204 (if (typep (prog1 x (setq x y)) 'double-float)
1207 (foo 1d0 5) => segmentation violation
1210 (fixed in sbcl-0.7.10.36)
1212 235: "type system and inline expansion"
1214 (declaim (ftype (function (cons) number) acc))
1215 (declaim (inline acc))
1217 (the number (car c)))
1220 (values (locally (declare (optimize (safety 0)))
1222 (locally (declare (optimize (safety 3)))
1225 (foo '(nil) '(t)) => NIL, T.
1227 b. (reported by brown on #lisp 2003-01-21)
1230 (declare (optimize (speed 3) (safety 0)))
1231 (declare (notinline mapcar))
1232 (let ((z (mapcar #'car x)))
1235 Without (DECLARE (NOTINLINE MAPCAR)), Python cannot derive that Z is
1238 DEFUNCT CATEGORIES OF BUGS
1240 These labels were used for bugs related to the old IR1 interpreter.
1241 The # values reached 6 before the category was closed down.