REPORTING BUGS
Bugs can be reported on the help mailing list
sbcl-help@lists.sourceforge.net
or on the development mailing list
sbcl-devel@lists.sourceforge.net
Please include enough information in a bug report that someone reading
it can reproduce the problem, i.e. don't write
Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
PRINT-OBJECT doesn't seem to work with *PRINT-LENGTH*. Is this a bug?
but instead
Subject: apparent bug in PRINT-OBJECT (or *PRINT-LENGTH*?)
In sbcl-1.2.3 running under OpenBSD 4.5 on my Alpha box, when
I compile and load the file
(DEFSTRUCT (FOO (:PRINT-OBJECT (LAMBDA (X Y)
(LET ((*PRINT-LENGTH* 4))
(PRINT X Y)))))
X Y)
then at the command line type
(MAKE-FOO)
the program loops endlessly instead of printing the object.
NOTES:
There is also some information on bugs in the manual page and
in the TODO file. Eventually more such information may move here.
The gaps in the number sequence belong to old bug descriptions which
have gone away (typically because they were fixed, but sometimes for
other reasons, e.g. because they were moved elsewhere).
2:
DEFSTRUCT almost certainly should overwrite the old LAYOUT information
instead of just punting when a contradictory structure definition
is loaded. As it is, if you redefine DEFSTRUCTs in a way which
changes their layout, you probably have to rebuild your entire
program, even if you know or guess enough about the internals of
SBCL to wager that this (undefined in ANSI) operation would be safe.
3: "type checking of structure slots"
a:
ANSI specifies that a type mismatch in a structure slot
initialization value should not cause a warning.
WORKAROUND:
This one might not be fixed for a while because while we're big
believers in ANSI compatibility and all, (1) there's no obvious
simple way to do it (short of disabling all warnings for type
mismatches everywhere), and (2) there's a good portable
workaround, and (3) by their own reasoning, it looks as though
ANSI may have gotten it wrong. ANSI justifies this specification
by saying
The restriction against issuing a warning for type mismatches
between a slot-initform and the corresponding slot's :TYPE
option is necessary because a slot-initform must be specified
in order to specify slot options; in some cases, no suitable
default may exist.
However, in SBCL (as in CMU CL or, for that matter, any compiler
which really understands Common Lisp types) a suitable default
does exist, in all cases, because the compiler understands the
concept of functions which never return (i.e. has return type NIL).
Thus, as a portable workaround, you can use a call to some
known-never-to-return function as the default. E.g.
(DEFSTRUCT FOO
(BAR (ERROR "missing :BAR argument")
:TYPE SOME-TYPE-TOO-HAIRY-TO-CONSTRUCT-AN-INSTANCE-OF))
or
(DECLAIM (FTYPE (FUNCTION () NIL) MISSING-ARG))
(DEFUN REQUIRED-ARG () ; workaround for SBCL non-ANSI slot init typing
(ERROR "missing required argument"))
(DEFSTRUCT FOO
(BAR (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
(BLETCH (REQUIRED-ARG) :TYPE TRICKY-TYPE-OF-SOME-SORT)
(N-REFS-SO-FAR 0 :TYPE (INTEGER 0)))
Such code should compile without complaint and work correctly either
on SBCL or on any other completely compliant Common Lisp system.
b: &AUX argument in a boa-constructor without a default value means
"do not initilize this slot" and does not cause type error. But
an error may be signalled at read time and it would be good if
SBCL did it.
d: (fixed in 0.8.1.5)
33:
And as long as we're wishing, it would be awfully nice if INSPECT could
also report on closures, telling about the values of the bound variables.
Currently INSPECT and DESCRIBE do show the values, but showing the
names of the bindings would be even nicer.
35:
The compiler assumes that any time a function of declared FTYPE
doesn't signal an error, its arguments were of the declared type.
E.g. compiling and loading
(DECLAIM (OPTIMIZE (SAFETY 3)))
(DEFUN FACTORIAL (X) (GAMMA (1+ X)))
(DEFUN GAMMA (X) X)
(DECLAIM (FTYPE (FUNCTION (UNSIGNED-BYTE)) FACTORIAL))
(DEFUN FOO (X)
(COND ((> (FACTORIAL X) 1.0E6)
(FORMAT T "too big~%"))
((INTEGERP X)
(FORMAT T "exactly ~S~%" (FACTORIAL X)))
(T
(FORMAT T "approximately ~S~%" (FACTORIAL X)))))
then executing
(FOO 1.5)
will cause the INTEGERP case to be selected, giving bogus output a la
exactly 2.5
This violates the "declarations are assertions" principle.
According to the ANSI spec, in the section "System Class FUNCTION",
this is a case of "lying to the compiler", but the lying is done
by the code which calls FACTORIAL with non-UNSIGNED-BYTE arguments,
not by the unexpectedly general definition of FACTORIAL. In any case,
"declarations are assertions" means that lying to the compiler should
cause an error to be signalled, and should not cause a bogus
result to be returned. Thus, the compiler should not assume
that arbitrary functions check their argument types. (It might
make sense to add another flag (CHECKED?) to DEFKNOWN to
identify functions which *do* check their argument types.)
(Also, verify that the compiler handles declared function
return types as assertions.)
42:
The definitions of SIGCONTEXT-FLOAT-REGISTER and
%SET-SIGCONTEXT-FLOAT-REGISTER in x86-vm.lisp say they're not
supported on FreeBSD because the floating point state is not saved,
but at least as of FreeBSD 4.0, the floating point state *is* saved,
so they could be supported after all. Very likely
SIGCONTEXT-FLOATING-POINT-MODES could now be supported, too.
61:
Compiling and loading
(DEFUN FAIL (X) (THROW 'FAIL-TAG X))
(FAIL 12)
then requesting a BACKTRACE at the debugger prompt gives no information
about where in the user program the problem occurred.
(this is apparently mostly fixed on the SPARC, PPC, and x86 architectures:
while giving the backtrace the non-x86 systems complains about "unknown
source location: using block start", but apart from that the
backtrace seems reasonable. On x86 this is masked by bug 353. See
tests/debug.impure.lisp for a test case)
64:
Using the pretty-printer from the command prompt gives funny
results, apparently because the pretty-printer doesn't know
about user's command input, including the user's carriage return
that the user, and therefore the pretty-printer thinks that
the new output block should start indented 2 or more characters
rightward of the correct location.
67:
As reported by Winton Davies on a CMU CL mailing list 2000-01-10,
and reported for SBCL by Martin Atzmueller 2000-10-20: (TRACE GETHASH)
crashes SBCL. In general tracing anything which is used in the
implementation of TRACE is likely to have the same problem.
78:
ANSI says in one place that type declarations can be abbreviated even
when the type name is not a symbol, e.g.
(DECLAIM ((VECTOR T) *FOOVECTOR*))
SBCL doesn't support this. But ANSI says in another place that this
isn't allowed. So it's not clear this is a bug after all. (See the
e-mail on cmucl-help@cons.org on 2001-01-16 and 2001-01-17 from WHN
and Pierre Mai.)
(Actually this has changed changed since, and types as above are
now supported. This may be a bug.)
83:
RANDOM-INTEGER-EXTRA-BITS=10 may not be large enough for the RANDOM
RNG to be high quality near RANDOM-FIXNUM-MAX; it looks as though
the mean of the distribution can be systematically O(0.1%) wrong.
Just increasing R-I-E-B is probably not a good solution, since
it would decrease efficiency more than is probably necessary. Perhaps
using some sort of accept/reject method would be better.
85:
Internally the compiler sometimes evaluates
(sb-kernel:type/= (specifier-type '*) (specifier-type t))
(I stumbled across this when I added an
(assert (not (eq type1 *wild-type*)))
in the NAMED :SIMPLE-= type method.) '* isn't really a type, and
in a type context should probably be translated to T, and so it's
probably wrong to ask whether it's equal to the T type and then (using
the EQ type comparison in the NAMED :SIMPLE-= type method) return NIL.
(I haven't tried to investigate this bug enough to guess whether
there might be any user-level symptoms.)
In fact, the type system is likely to depend on this inequality not
holding... * is not equivalent to T in many cases, such as
(VECTOR *) /= (VECTOR T).
98:
In sbcl-0.6.11.41 (and in all earlier SBCL, and in CMU
CL), out-of-line structure slot setters are horribly inefficient
whenever the type of the slot is declared, because out-of-line
structure slot setters are implemented as closures to save space,
so the compiler doesn't compile the type test into code, but
instead just saves the type in a lexical closure and interprets it
at runtime.
To exercise the problem, compile and load
(cl:in-package :cl-user)
(defstruct foo
(bar (error "missing") :type bar))
(defvar *foo*)
(defun wastrel1 (x)
(loop (setf (foo-bar *foo*) x)))
(defstruct bar)
(defvar *bar* (make-bar))
(defvar *foo* (make-foo :bar *bar*))
(defvar *setf-foo-bar* #'(setf foo-bar))
(defun wastrel2 (x)
(loop (funcall *setf-foo-bar* x *foo*)))
then run (WASTREL1 *BAR*) or (WASTREL2 *BAR*), hit Ctrl-C, and
use BACKTRACE, to see it's spending all essentially all its time
in %TYPEP and VALUES-SPECIFIER-TYPE and so forth.
One possible solution would be simply to give up on
representing structure slot accessors as functions, and represent
them as macroexpansions instead. This can be inconvenient for users,
but it's not clear that it's worse than trying to help by expanding
into a horribly inefficient implementation.
As a workaround for the problem, #'(SETF FOO) expressions
can be replaced with (EFFICIENT-SETF-FUNCTION FOO), where
(defmacro efficient-setf-function (place-function-name)
(or #+sbcl (and (sb-int:info :function :accessor-for place-function-name)
;; a workaround for the problem, encouraging the
;; inline expansion of the structure accessor, so
;; that the compiler can optimize its type test
(let ((new-value (gensym "NEW-VALUE-"))
(structure-value (gensym "STRUCTURE-VALUE-")))
`(lambda (,new-value ,structure-value)
(setf (,place-function-name ,structure-value)
,new-value))))
;; no problem, can just use the ordinary expansion
`(function (setf ,place-function-name))))
100:
There's apparently a bug in CEILING optimization which caused
Douglas Crosher to patch the CMU CL version. Martin Atzmueller
applied the patches to SBCL and they didn't seem to cause problems
(as reported sbcl-devel 2001-05-04). However, since the patches
modify nontrivial code which was apparently written incorrectly
the first time around, until regression tests are written I'm not
comfortable merging the patches in the CVS version of SBCL.
108:
ROOM issues:
a) ROOM works by walking over the heap linearly, instead of
following the object graph. Hence, it report garbage objects that
are unreachable. (Maybe this is a feature and not a bug?)
b) ROOM uses MAP-ALLOCATED-OBJECTS to walk the heap, which doesn't
check all pointers as well as it should, and can hence become
confused, leading to aver failures. As of 1.0.13.21 these (the
SAP= aver in particular) should be mostly under control, but push
ROOM hard enough and it still might croak.
117:
When the compiler inline expands functions, it may be that different
kinds of return values are generated from different code branches.
E.g. an inline expansion of POSITION generates integer results
from one branch, and NIL results from another. When that inline
expansion is used in a context where only one of those results
is acceptable, e.g.
(defun foo (x)
(aref *a1* (position x *a2*)))
and the compiler can't prove that the unacceptable branch is
never taken, then bogus type mismatch warnings can be generated.
If you need to suppress the type mismatch warnings, you can
suppress the inline expansion,
(defun foo (x)
#+sbcl (declare (notinline position)) ; to suppress bug 117 bogowarnings
(aref *a1* (position x *a2*)))
or, sometimes, suppress them by declaring the result to be of an
appropriate type,
(defun foo (x)
(aref *a1* (the integer (position x *a2*))))
This is not a new compiler problem in 0.7.0, but the new compiler
transforms for FIND, POSITION, FIND-IF, and POSITION-IF make it
more conspicuous. If you don't need performance from these functions,
and the bogus warnings are a nuisance for you, you can return to
your pre-0.7.0 state of grace with
#+sbcl (declaim (notinline find position find-if position-if)) ; bug 117..
(see also bug 279)
124:
As of version 0.pre7.14, SBCL's implementation of MACROLET makes
the entire lexical environment at the point of MACROLET available
in the bodies of the macroexpander functions. In particular, it
allows the function bodies (which run at compile time) to try to
access lexical variables (which are only defined at runtime).
It doesn't even issue a warning, which is bad.
The SBCL behavior arguably conforms to the ANSI spec (since the
spec says that the behavior is undefined, ergo anything conforms).
However, it would be better to issue a compile-time error.
Unfortunately I (WHN) don't see any simple way to detect this
condition in order to issue such an error, so for the meantime
SBCL just does this weird broken "conforming" thing.
The ANSI standard says, in the definition of the special operator
MACROLET,
The macro-expansion functions defined by MACROLET are defined
in the lexical environment in which the MACROLET form appears.
Declarations and MACROLET and SYMBOL-MACROLET definitions affect
the local macro definitions in a MACROLET, but the consequences
are undefined if the local macro definitions reference any
local variable or function bindings that are visible in that
lexical environment.
Then it seems to contradict itself by giving the example
(defun foo (x flag)
(macrolet ((fudge (z)
;The parameters x and flag are not accessible
; at this point; a reference to flag would be to
; the global variable of that name.
` (if flag (* ,z ,z) ,z)))
;The parameters x and flag are accessible here.
(+ x
(fudge x)
(fudge (+ x 1)))))
The comment "a reference to flag would be to the global variable
of the same name" sounds like good behavior for the system to have.
but actual specification quoted above says that the actual behavior
is undefined.
(Since 0.7.8.23 macroexpanders are defined in a restricted version
of the lexical environment, containing no lexical variables and
functions, which seems to conform to ANSI and CLtL2, but signalling
a STYLE-WARNING for references to variables similar to locals might
be a good thing.)
135:
Ideally, uninterning a symbol would allow it, and its associated
FDEFINITION and PROCLAIM data, to be reclaimed by the GC. However,
at least as of sbcl-0.7.0, this isn't the case. Information about
FDEFINITIONs and PROCLAIMed properties is stored in globaldb.lisp
essentially in ordinary (non-weak) hash tables keyed by symbols.
Thus, once a system has an entry in this system, it tends to live
forever, even when it is uninterned and all other references to it
are lost.
143:
(reported by Jesse Bouwman 2001-10-24 through the unfortunately
prominent SourceForge web/db bug tracking system, which is
unfortunately not a reliable way to get a timely response from
the SBCL maintainers)
In the course of trying to build a test case for an
application error, I encountered this behavior:
If you start up sbcl, and then lay on CTRL-C for a
minute or two, the lisp process will eventually say:
%PRIMITIVE HALT called; the party is over.
and throw you into the monitor. If I start up lisp,
attach to the process with strace, and then do the same
(abusive) thing, I get instead:
access failure in heap page not marked as write-protected
and the monitor again. I don't know enough to have the
faintest idea of what is going on here.
This is with sbcl 6.12, uname -a reports:
Linux prep 2.2.19 #4 SMP Tue Apr 24 13:59:52 CDT 2001 i686 unknown
I (WHN) have verified that the same thing occurs on sbcl-0.pre7.141
under OpenBSD 2.9 on my X86 laptop. Do be patient when you try it:
it took more than two minutes (but less than five) for me.
145:
a.
ANSI allows types `(COMPLEX ,FOO) to use very hairy values for
FOO, e.g. (COMPLEX (AND REAL (SATISFIES ODDP))). The old CMU CL
COMPLEX implementation didn't deal with this, and hasn't been
upgraded to do so. (This doesn't seem to be a high priority
conformance problem, since seems hard to construct useful code
where it matters.)
[ partially fixed by CSR in 0.8.17.17 because of a PFD ansi-tests
report that (COMPLEX RATIO) was failing; still failing on types of
the form (AND NUMBER (SATISFIES REALP) (SATISFIES ZEROP)). ]
b. (fixed in 0.8.3.43)
146:
Floating point errors are reported poorly. E.g. on x86 OpenBSD
with sbcl-0.7.1,
* (expt 2.0 12777)
debugger invoked on condition of type SB-KERNEL:FLOATING-POINT-EXCEPTION:
An arithmetic error SB-KERNEL:FLOATING-POINT-EXCEPTION was signalled.
No traps are enabled? How can this be?
It should be possible to be much more specific (overflow, division
by zero, etc.) and of course the "How can this be?" should be fixable.
See also bugs #45.c and #183
162:
(reported by Robert E. Brown 2002-04-16)
When a function is called with too few arguments, causing the
debugger to be entered, the uninitialized slots in the bad call frame
seem to cause GCish problems, being interpreted as tagged data even
though they're not. In particular, executing ROOM in the
debugger at that point causes AVER failures:
* (machine-type)
"X86"
* (lisp-implementation-version)
"0.7.2.12"
* (typep 10)
...
0] (room)
...
failed AVER: "(SAP= CURRENT END)"
(Christophe Rhodes reports that this doesn't occur on the SPARC, which
isn't too surprising since there are many differences in stack
implementation and GC conservatism between the X86 and other ports.)
This is probably the same bug as 216
173:
The compiler sometimes tries to constant-fold expressions before
it checks to see whether they can be reached. This can lead to
bogus warnings about errors in the constant folding, e.g. in code
like
(WHEN X
(WRITE-STRING (> X 0) "+" "0"))
compiled in a context where the compiler can prove that X is NIL,
and the compiler complains that (> X 0) causes a type error because
NIL isn't a valid argument to #'>. Until sbcl-0.7.4.10 or so this
caused a full WARNING, which made the bug really annoying because then
COMPILE and COMPILE-FILE returned FAILURE-P=T for perfectly legal
code. Since then the warning has been downgraded to STYLE-WARNING,
so it's still a bug but at least it's a little less annoying.
183: "IEEE floating point issues"
Even where floating point handling is being dealt with relatively
well (as of sbcl-0.7.5, on sparc/sunos and alpha; see bug #146), the
accrued-exceptions and current-exceptions part of the fp control
word don't seem to bear much relation to reality. E.g. on
SPARC/SunOS:
* (/ 1.0 0.0)
debugger invoked on condition of type DIVISION-BY-ZERO:
arithmetic error DIVISION-BY-ZERO signalled
0] (sb-vm::get-floating-point-modes)
(:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
:ROUNDING-MODE :NEAREST
:CURRENT-EXCEPTIONS NIL
:ACCRUED-EXCEPTIONS (:INEXACT)
:FAST-MODE NIL)
0] abort
* (sb-vm::get-floating-point-modes)
(:TRAPS (:OVERFLOW :INVALID :DIVIDE-BY-ZERO)
:ROUNDING-MODE :NEAREST
:CURRENT-EXCEPTIONS (:INEXACT)
:ACCRUED-EXCEPTIONS (:INEXACT)
:FAST-MODE NIL)
188: "compiler performance fiasco involving type inference and UNION-TYPE"
(time (compile
nil
'(lambda ()
(declare (optimize (safety 3)))
(declare (optimize (compilation-speed 2)))
(declare (optimize (speed 1) (debug 1) (space 1)))
(let ((start 4))
(declare (type (integer 0) start))
(print (incf start 22))
(print (incf start 26))
(print (incf start 28)))
(let ((start 6))
(declare (type (integer 0) start))
(print (incf start 22))
(print (incf start 26)))
(let ((start 10))
(declare (type (integer 0) start))
(print (incf start 22))
(print (incf start 26))))))
This example could be solved with clever enough constraint
propagation or with SSA, but consider
(let ((x 0))
(loop (incf x 2)))
The careful type of X is {2k} :-(. Is it really important to be
able to work with unions of many intervals?
191: "Miscellaneous PCL deficiencies"
(reported by Alexey Dejneka sbcl-devel 2002-08-04)
a. DEFCLASS does not inform the compiler about generated
functions. Compiling a file with
(DEFCLASS A-CLASS ()
((A-CLASS-X)))
(DEFUN A-CLASS-X (A)
(WITH-SLOTS (A-CLASS-X) A
A-CLASS-X))
results in a STYLE-WARNING:
undefined-function
SB-SLOT-ACCESSOR-NAME::|COMMON-LISP-USER A-CLASS-X slot READER|
APD's fix for this was checked in to sbcl-0.7.6.20, but Pierre
Mai points out that the declamation of functions is in fact
incorrect in some cases (most notably for structure
classes). This means that at present erroneous attempts to use
WITH-SLOTS and the like on classes with metaclass STRUCTURE-CLASS
won't get the corresponding STYLE-WARNING.
[much later, in 2006-08] in fact it's no longer erroneous to use
WITH-SLOTS on structure-classes. However, including :METACLASS
STRUCTURE-CLASS in the class definition gives a whole bunch of
function redefinition warnings, so we're still not good to close
this bug...
c. (fixed in 0.8.4.23)
201: "Incautious type inference from compound types"
a. (reported by APD sbcl-devel 2002-09-17)
(DEFUN FOO (X)
(LET ((Y (CAR (THE (CONS INTEGER *) X))))
(SETF (CAR X) NIL)
(FORMAT NIL "~S IS ~S, Y = ~S"
(CAR X)
(TYPECASE (CAR X)
(INTEGER 'INTEGER)
(T '(NOT INTEGER)))
Y)))
(FOO ' (1 . 2)) => "NIL IS INTEGER, Y = 1"
b.
* (defun foo (x)
(declare (type (array * (4 4)) x))
(let ((y x))
(setq x (make-array '(4 4)))
(adjust-array y '(3 5))
(= (array-dimension y 0) (eval `(array-dimension ,y 0)))))
FOO
* (foo (make-array '(4 4) :adjustable t))
NIL
205: "environment issues in cross compiler"
(These bugs have no impact on user code, but should be fixed or
documented.)
a. Macroexpanders introduced with MACROLET are defined in the null
lexical environment.
b. The body of (EVAL-WHEN (:COMPILE-TOPLEVEL) ...) is evaluated in
the null lexical environment.
c. The cross-compiler cannot inline functions defined in a non-null
lexical environment.
206: ":SB-FLUID feature broken"
(reported by Antonio Martinez-Shotton sbcl-devel 2002-10-07)
Enabling :SB-FLUID in the target-features list in sbcl-0.7.8 breaks
the build.
207: "poorly distributed SXHASH results for compound data"
SBCL's SXHASH could probably try a little harder. ANSI: "the
intent is that an implementation should make a good-faith
effort to produce hash-codes that are well distributed
within the range of non-negative fixnums". But
(let ((hits (make-hash-table)))
(dotimes (i 16)
(dotimes (j 16)
(let* ((ij (cons i j))
(newlist (push ij (gethash (sxhash ij) hits))))
(when (cdr newlist)
(format t "~&collision: ~S~%" newlist))))))
reports lots of collisions in sbcl-0.7.8. A stronger MIX function
would be an obvious way of fix. Maybe it would be acceptably efficient
to redo MIX using a lookup into a 256-entry s-box containing
29-bit pseudorandom numbers?
211: "keywords processing"
a. :ALLOW-OTHER-KEYS T should allow a function to receive an odd
number of keyword arguments.
212: "Sequence functions and circular arguments"
COERCE, MERGE and CONCATENATE go into an infinite loop when given
circular arguments; it would be good for the user if they could be
given an error instead (ANSI 17.1.1 allows this behaviour on the part
of the implementation, as conforming code cannot give non-proper
sequences to these functions. MAP also has this problem (and
solution), though arguably the convenience of being able to do
(MAP 'LIST '+ FOO '#1=(1 . #1#))
might be classed as more important (though signalling an error when
all of the arguments are circular is probably desireable).
213: "Sequence functions and type checking"
b. MAP, when given a type argument that is SUBTYPEP LIST, does not
check that it will return a sequence of the given type. Fixing
it along the same lines as the others (cf. work done around
sbcl-0.7.8.45) is possible, but doing so efficiently didn't look
entirely straightforward.
c. All of these functions will silently accept a type of the form
(CONS INTEGER *)
whether or not the return value is of this type. This is
probably permitted by ANSI (see "Exceptional Situations" under
ANSI MAKE-SEQUENCE), but the DERIVE-TYPE mechanism does not
know about this escape clause, so code of the form
(INTEGERP (CAR (MAKE-SEQUENCE '(CONS INTEGER *) 2)))
can erroneously return T.
215: ":TEST-NOT handling by functions"
a. FIND and POSITION currently signal errors when given non-NIL for
both their :TEST and (deprecated) :TEST-NOT arguments, but by
ANSI 17.2 "the consequences are unspecified", which by ANSI 1.4.2
means that the effect is "unpredictable but harmless". It's not
clear what that actually means; it may preclude conforming
implementations from signalling errors.
b. COUNT, REMOVE and the like give priority to a :TEST-NOT argument
when conflict occurs. As a quality of implementation issue, it
might be preferable to treat :TEST and :TEST-NOT as being in some
sense the same &KEY, and effectively take the first test function in
the argument list.
c. Again, a quality of implementation issue: it would be good to issue a
STYLE-WARNING at compile-time for calls with :TEST-NOT, and a
WARNING for calls with both :TEST and :TEST-NOT; possibly this
latter should be WARNed about at execute-time too.
216: "debugger confused by frames with invalid number of arguments"
In sbcl-0.7.8.51, executing e.g. (VECTOR-PUSH-EXTEND T), BACKTRACE, Q
leaves the system confused, enough so that (QUIT) no longer works.
It's as though the process of working with the uninitialized slot in
the bad VECTOR-PUSH-EXTEND frame causes GC problems, though that may
not be the actual problem. (CMU CL 18c doesn't have problems with this.)
This is probably the same bug as 162
235: "type system and inline expansion"
a.
(declaim (ftype (function (cons) number) acc))
(declaim (inline acc))
(defun acc (c)
(the number (car c)))
(defun foo (x y)
(values (locally (declare (optimize (safety 0)))
(acc x))
(locally (declare (optimize (safety 3)))
(acc y))))
(foo '(nil) '(t)) => NIL, T.
As of 0.9.15.41 this seems to be due to ACC being inlined only once
inside FOO, which results in the second call reusing the FUNCTIONAL
resulting from the first -- which doesn't check the type.
237: "Environment arguments to type functions"
a. Functions SUBTYPEP, TYPEP, UPGRADED-ARRAY-ELEMENT-TYPE, and
UPGRADED-COMPLEX-PART-TYPE now have an optional environment
argument, but they ignore it completely. This is almost
certainly not correct.
b. Also, the compiler's optimizers for TYPEP have not been informed
about the new argument; consequently, they will not transform
calls of the form (TYPEP 1 'INTEGER NIL), even though this is
just as optimizeable as (TYPEP 1 'INTEGER).
238: "REPL compiler overenthusiasm for CLOS code"
From the REPL,
* (defclass foo () ())
* (defmethod bar ((x foo) (foo foo)) (call-next-method))
causes approximately 100 lines of code deletion notes. Some
discussion on this issue happened under the title 'Three "interesting"
bugs in PCL', resulting in a fix for this oververbosity from the
compiler proper; however, the problem persists in the interactor
because the notion of original source is not preserved: for the
compiler, the original source of the above expression is (DEFMETHOD
BAR ((X FOO) (FOO FOO)) (CALL-NEXT-METHOD)), while by the time the
compiler gets its hands on the code needing compilation from the REPL,
it has been macroexpanded several times.
A symptom of the same underlying problem, reported by Tony Martinez:
* (handler-case
(with-input-from-string (*query-io* " no")
(yes-or-no-p))
(simple-type-error () 'error))
; in: LAMBDA NIL
; (SB-KERNEL:FLOAT-WAIT)
;
; note: deleting unreachable code
; compilation unit finished
; printed 1 note
242: "WRITE-SEQUENCE suboptimality"
(observed from clx performance)
In sbcl-0.7.13, WRITE-SEQUENCE of a sequence of type
(SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) on a stream with element-type
(UNSIGNED-BYTE 8) will write to the stream one byte at a time,
rather than writing the sequence in one go, leading to severe
performance degradation.
As of sbcl-0.9.0.36, this is solved for fd-streams, so is less of a
problem in practice. (Fully fixing this would require adding a
ansi-stream-n-bout slot and associated methods to write a byte
sequence to ansi-stream, similar to the existing ansi-stream-sout
slot/functions.)
243: "STYLE-WARNING overenthusiasm for unused variables"
(observed from clx compilation)
In sbcl-0.7.14, in the presence of the macros
(DEFMACRO FOO (X) `(BAR ,X))
(DEFMACRO BAR (X) (DECLARE (IGNORABLE X)) 'NIL)
somewhat surprising style warnings are emitted for
(COMPILE NIL '(LAMBDA (Y) (FOO Y))):
; in: LAMBDA (Y)
; (LAMBDA (Y) (FOO Y))
;
; caught STYLE-WARNING:
; The variable Y is defined but never used.
245: bugs in disassembler
b. On X86 operand size prefix is not recognized.
251:
(defun foo (&key (a :x))
(declare (fixnum a))
a)
does not cause a warning. (BTW: old SBCL issued a warning, but for a
function, which was never called!)
256:
Compiler does not emit warnings for
a. (lambda () (svref (make-array 8 :adjustable t) 1))
b. (fixed at some point before 1.0.4.10)
c. (lambda (x)
(declare (optimize (debug 0)))
(declare (type vector x))
(list (fill-pointer x)
(svref x 1)))
257:
Complex array type does not have corresponding type specifier.
This is a problem because the compiler emits optimization notes when
you use a non-simple array, and without a type specifier for hairy
array types, there's no good way to tell it you're doing it
intentionally so that it should shut up and just compile the code.
Another problem is confusing error message "asserted type ARRAY
conflicts with derived type (VALUES SIMPLE-VECTOR &OPTIONAL)" during
compiling (LAMBDA (V) (VALUES (SVREF V 0) (VECTOR-POP V))).
The last problem is that when type assertions are converted to type
checks, types are represented with type specifiers, so we could lose
complex attribute. (Now this is probably not important, because
currently checks for complex arrays seem to be performed by
callees.)
259:
(compile nil '(lambda () (aref (make-array 0) 0))) compiles without
warning. Analogous cases with the index and length being equal and
greater than 0 are warned for; the problem here seems to be that the
type required for an array reference of this type is (INTEGER 0 (0))
which is canonicalized to NIL.
260:
a.
(let* ((s (gensym))
(t1 (specifier-type s)))
(eval `(defstruct ,s))
(type= t1 (specifier-type s)))
=> NIL, NIL
(fixed in 0.8.1.24)
b. The same for CSUBTYPEP.
262: "yet another bug in inline expansion of local functions"
During inline expansion of a local function Python can try to
reference optimized away objects (functions, variables, CTRANs from
tags and blocks), which later may lead to problems. Some of the
cases are worked around by forbidding expansion in such cases, but
the better way would be to reimplement inline expansion by copying
IR1 structures.
266:
David Lichteblau provided (sbcl-devel 2003-06-01) a patch to fix
behaviour of streams with element-type (SIGNED-BYTE 8). The patch
looks reasonable, if not obviously correct; however, it caused the
PPC/Linux port to segfault during warm-init while loading
src/pcl/std-class.fasl. A workaround patch was made, but it would
be nice to understand why the first patch caused problems, and to
fix the cause if possible.
268: "wrong free declaration scope"
The following code must signal type error:
(locally (declare (optimize (safety 3)))
(flet ((foo (x &optional (y (car x)))
(declare (optimize (safety 0)))
(list x y)))
(funcall (eval #'foo) 1)))
270:
In the following function constraint propagator optimizes nothing:
(defun foo (x)
(declare (integer x))
(declare (optimize speed))
(typecase x
(fixnum "hala")
(fixnum "buba")
(bignum "hip")
(t "zuz")))
273:
Compilation of the following two forms causes "X is unbound" error:
(symbol-macrolet ((x pi))
(macrolet ((foo (y) (+ x y)))
(declaim (inline bar))
(defun bar (z)
(* z (foo 4)))))
(defun quux (z)
(bar z))
(See (COERCE (CDR X) 'FUNCTION) in IR1-CONVERT-INLINE-LAMBDA.)
274:
CLHS says that type declaration of a symbol macro should not affect
its expansion, but in SBCL it does. (If you like magic and want to
fix it, don't forget to change all uses of MACROEXPAND to
MACROEXPAND*.)
275:
The following code (taken from CLOCC) takes a lot of time to compile:
(defun foo (n)
(declare (type (integer 0 #.large-constant) n))
(expt 1/10 n))
(fixed in 0.8.2.51, but a test case would be good)
276:
b. The same as in a., but using MULTIPLE-VALUE-SETQ instead of SETQ.
c. (defvar *faa*)
(defmethod faa ((*faa* double-float))
(set '*faa* (when (< *faa* 0) (- *faa*)))
(1+ *faa*))
(faa 1d0) => type error
279: type propagation error -- correctly inferred type goes astray?
In sbcl-0.8.3 and sbcl-0.8.1.47, the warning
The binding of ABS-FOO is a (VALUES (INTEGER 0 0)
&OPTIONAL), not a (INTEGER 1 536870911)
is emitted when compiling this file:
(declaim (ftype (function ((integer 0 #.most-positive-fixnum))
(integer #.most-negative-fixnum 0))
foo))
(defun foo (x)
(- x))
(defun bar (x)
(let* (;; Uncomment this for a type mismatch warning indicating
;; that the type of (FOO X) is correctly understood.
#+nil (fs-foo (float-sign (foo x)))
;; Uncomment this for a type mismatch warning
;; indicating that the type of (ABS (FOO X)) is
;; correctly understood.
#+nil (fs-abs-foo (float-sign (abs (foo x))))
;; something wrong with this one though
(abs-foo (abs (foo x))))
(declare (type (integer 1 100) abs-foo))
(print abs-foo)))
(see also bug 117)
283: Thread safety: libc functions
There are places that we call unsafe-for-threading libc functions
that we should find alternatives for, or put locks around. Known or
strongly suspected problems, as of 1.0.3.13: please update this
bug instead of creating new ones
284: Thread safety: special variables
There are lots of special variables in SBCL, and I feel sure that at
least some of them are indicative of potentially thread-unsafe
parts of the system. See doc/internals/notes/threading-specials
286: "recursive known functions"
Self-call recognition conflicts with known function
recognition. Currently cross compiler and target COMPILE do not
recognize recursion, and in target compiler it can be disabled. We
can always disable it for known functions with RECURSIVE attribute,
but there remains a possibility of a function with a
(tail)-recursive simplification pass and transforms/VOPs for base
cases.
288: fundamental cross-compilation issues (from old UGLINESS file)
Using host floating point numbers to represent target floating point
numbers, or host characters to represent target characters, is
theoretically shaky. (The characters are OK as long as the characters
are in the ANSI-guaranteed character set, though, so they aren't a
real problem as long as the sources don't need anything but that;
the floats are a real problem.)
289: "type checking and source-transforms"
a.
(block nil (let () (funcall #'+ (eval 'nil) (eval '1) (return :good))))
signals type error.
Our policy is to check argument types at the moment of a call. It
disagrees with ANSI, which says that type assertions are put
immediately onto argument expressions, but is easier to implement in
IR1 and is more compatible to type inference, inline expansion,
etc. IR1-transforms automatically keep this policy, but source
transforms for associative functions (such as +), being applied
during IR1-convertion, do not. It may be tolerable for direct calls
(+ x y z), but for (FUNCALL #'+ x y z) it is non-conformant.
b. Another aspect of this problem is efficiency. [x y + z +]
requires less registers than [x y z + +]. This transformation is
currently performed with source transforms, but it would be good to
also perform it in IR1 optimization phase.
290: Alpha floating point and denormalized traps
In SBCL 0.8.3.6x on the alpha, we work around what appears to be a
hardware or kernel deficiency: the status of the enable/disable
denormalized-float traps bit seems to be ambiguous; by the time we
get to os_restore_fp_control after a trap, denormalized traps seem
to be enabled. Since we don't want a trap every time someone uses a
denormalized float, in general, we mask out that bit when we restore
the control word; however, this clobbers any change the user might
have made.
297:
LOOP with non-constant arithmetic step clauses suffers from overzealous
type constraint: code of the form
(loop for d of-type double-float from 0d0 to 10d0 by x collect d)
compiles to a type restriction on X of (AND DOUBLE-FLOAT (REAL
(0))). However, an integral value of X should be legal, because
successive adds of integers to double-floats produces double-floats,
so none of the type restrictions in the code is violated.
300: (reported by Peter Graves) Function PEEK-CHAR checks PEEK-TYPE
argument type only after having read a character. This is caused
with EXPLICIT-CHECK attribute in DEFKNOWN. The similar problem
exists with =, /=, <, >, <=, >=. They were fixed, but it is probably
less error prone to have EXPLICIT-CHECK be a local declaration,
being put into the definition, instead of an attribute being kept in
a separate file; maybe also put it into SB-EXT?
301: ARRAY-SIMPLE-=-TYPE-METHOD breaks on corner cases which can arise
in NOTE-ASSUMED-TYPES
In sbcl-0.8.7.32, compiling the file
(defun foo (x y)
(declare (type integer x))
(declare (type (vector (or hash-table bit)) y))
(bletch 2 y))
(defun bar (x y)
(declare (type integer x))
(declare (type (simple-array base (2)) y))
(bletch 1 y))
gives the error
failed AVER: "(NOT (AND (NOT EQUALP) CERTAINP))"
303: "nonlinear LVARs" (aka MISC.293)
(defun buu (x)
(multiple-value-call #'list
(block foo
(multiple-value-prog1
(eval '(values :a :b :c))
(catch 'bar
(if (> x 0)
(return-from foo
(eval `(if (> ,x 1)
1
(throw 'bar (values 3 4)))))))))))
(BUU 1) returns garbage.
The problem is that both EVALs sequentially write to the same LVAR.
306: "Imprecise unions of array types"
a. fixed in SBCL 0.9.15.48
b.(subtypep
'array
`(or
,@(loop for x across sb-vm:*specialized-array-element-type-properties*
collect `(array ,(sb-vm:saetp-specifier x)))))
=> NIL, T (when it should be T, T)
309: "Dubious values for implementation limits"
(reported by Bruno Haible sbcl-devel "Incorrect value of
multiple-values-limit" 2004-04-19)
(values-list (make-list 1000000)), on x86/linux, signals a stack
exhaustion condition, despite MULTIPLE-VALUES-LIMIT being
significantly larger than 1000000. There are probably similar
dubious values for CALL-ARGUMENTS-LIMIT (see cmucl-help/cmucl-imp
around the same time regarding a call to LIST on sparc with 1000
arguments) and other implementation limit constants.
314: "LOOP :INITIALLY clauses and scope of initializers"
reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
test suite, originally by Thomas F. Burdick.
;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html>
;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is
;; initialized to the result of evaluating form1. 6.1.7.2 says that
;; initially clauses are evaluated in the loop prologue, which precedes all
;; loop code except for the initial settings provided by with, for, or as.
(loop :for x = 0 :then (1+ x)
:for y = (1+ x) :then (ash y 1)
:for z :across #(1 3 9 27 81 243)
:for w = (+ x y z)
:initially (assert (zerop x)) :initially (assert (= 2 w))
:until (>= w 100) :collect w)
Expected: (2 6 15 38)
Got: ERROR
318: "stack overflow in compiler warning with redefined class"
reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
test suite.
(defstruct foo a)
(setf (find-class 'foo) nil)
(defstruct foo slot-1)
This used to give a stack overflow from within the printer, which has
been fixed as of 0.8.16.11. Current result:
; caught ERROR:
; can't compile TYPEP of anonymous or undefined class:
; #<SB-KERNEL:STRUCTURE-CLASSOID FOO>
...
debugger invoked on a TYPE-ERROR in thread 19973:
The value NIL is not of type FUNCTION.
CSR notes: it's not really clear what it should give: is (SETF FIND-CLASS)
meant to be enough to delete structure classes from the system?
319: "backquote with comma inside array"
reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP
test suite.
(read-from-string "`#1A(1 2 ,(+ 2 2) 4)")
gives
#(1 2 ((SB-IMPL::|,|) + 2 2) 4)
which probably isn't intentional.
324: "STREAMs and :ELEMENT-TYPE with large bytesize"
In theory, (open foo :element-type '(unsigned-byte <x>)) should work
for all positive integral <x>. At present, it only works for <x> up
to about 1024 (and similarly for signed-byte), so
(open "/dev/zero" :element-type '(unsigned-byte 1025))
gives an error in sbcl-0.8.10.
325: "CLOSE :ABORT T on supeseding streams"
Closing a stream opened with :IF-EXISTS :SUPERSEDE with :ABORT T leaves no
file on disk, even if one existed before opening.
The illegality of this is not crystal clear, as the ANSI dictionary
entry for CLOSE says that when :ABORT is T superseded files are not
superseded (ie. the original should be restored), whereas the OPEN
entry says about :IF-EXISTS :SUPERSEDE "If possible, the
implementation should not destroy the old file until the new stream
is closed." -- implying that even though undesirable, early deletion
is legal. Restoring the original would none the less be the polite
thing to do.
326: "*PRINT-CIRCLE* crosstalk between streams"
In sbcl-0.8.10.48 it's possible for *PRINT-CIRCLE* references to be
mixed between streams when output operations are intermingled closely
enough (as by doing output on S2 from within (PRINT-OBJECT X S1) in the
test case below), so that e.g. the references #2# appears on a stream
with no preceding #2= on that stream to define it (because the #2= was
sent to another stream).
(cl:in-package :cl-user)
(defstruct foo index)
(defparameter *foo* (make-foo :index 4))
(defstruct bar)
(defparameter *bar* (make-bar))
(defparameter *tangle* (list *foo* *bar* *foo*))
(defmethod print-object ((foo foo) stream)
(let ((index (foo-index foo)))
(format *trace-output*
"~&-$- emitting FOO ~D, ambient *BAR*=~S~%"
index *bar*)
(format stream "[FOO ~D]" index))
foo)
(let ((tsos (make-string-output-stream))
(ssos (make-string-output-stream)))
(let ((*print-circle* t)
(*trace-output* tsos)
(*standard-output* ssos))
(prin1 *tangle* *standard-output*))
(let ((string (get-output-stream-string ssos)))
(unless (string= string "(#1=[FOO 4] #S(BAR) #1#)")
;; In sbcl-0.8.10.48 STRING was "(#1=[FOO 4] #2# #1#)".:-(
(error "oops: ~S" string)))))
It might be straightforward to fix this by turning the
*CIRCULARITY-HASH-TABLE* and *CIRCULARITY-COUNTER* variables into
per-stream slots, but (1) it would probably be sort of messy faking
up the special variable binding semantics using UNWIND-PROTECT and
(2) it might be sort of a pain to test that no other bugs had been
introduced.
328: "Profiling generic functions", transplanted from #241
(from tonyms on #lisp IRC 2003-02-25)
In sbcl-0.7.12.55, typing
(defclass foo () ((bar :accessor foo-bar)))
(profile foo-bar)
(unintern 'foo-bar)
(defclass foo () ((bar :accessor foo-bar)))
gives the error message
"#:FOO-BAR already names an ordinary function or a macro."
Problem: when a generic function is profiled, it appears as an ordinary
function to PCL. (Remembering the uninterned accessor is OK, as the
redefinition must be able to remove old accessors from their generic
functions.)
329: "Sequential class redefinition"
reported by Bruno Haible:
(defclass reactor () ((max-temp :initform 10000000)))
(defvar *r1* (make-instance 'reactor))
(defvar *r2* (make-instance 'reactor))
(slot-value *r1* 'max-temp)
(slot-value *r2* 'max-temp)
(defclass reactor () ((uptime :initform 0)))
(slot-value *r1* 'uptime)
(defclass reactor () ((uptime :initform 0) (max-temp :initform 10000)))
(slot-value *r1* 'max-temp) ; => 10000
(slot-value *r2* 'max-temp) ; => 10000000 oops...
Possible solution:
The method effective when the wrapper is obsoleted can be saved
in the wrapper, and then to update the instance just run through
all the old wrappers in order from oldest to newest.
332: "fasl stack inconsistency in structure redefinition"
(reported by Tim Daly Jr sbcl-devel 2004-05-06)
Even though structure redefinition is undefined by the standard, the
following behaviour is suboptimal: running
(defun stimulate-sbcl ()
(let ((filename (format nil "/tmp/~A.lisp" (gensym))))
;;create a file which redefines a structure incompatibly
(with-open-file (f filename :direction :output :if-exists :supersede)
(print '(defstruct astruct foo) f)
(print '(defstruct astruct foo bar) f))
;;compile and load the file, then invoke the continue restart on
;;the structure redefinition error
(handler-bind ((error (lambda (c) (continue c))))
(load (compile-file filename)))))
(stimulate-sbcl)
and choosing the CONTINUE restart yields the message
debugger invoked on a SB-INT:BUG in thread 27726:
fasl stack not empty when it should be
336: "slot-definitions must retain the generic functions of accessors"
reported by Tony Martinez:
(defclass foo () ((bar :reader foo-bar)))
(defun foo-bar (x) x)
(defclass foo () ((bar :reader get-bar))) ; => error, should work
Note: just punting the accessor removal if the fdefinition
is not a generic function is not enough:
(defclass foo () ((bar :reader foo-bar)))
(defvar *reader* #'foo-bar)
(defun foo-bar (x) x)
(defclass foo () ((bar :initform 'ok :reader get-bar)))
(funcall *reader* (make-instance 'foo)) ; should be an error, since
; the method must be removed
; by the class redefinition
Fixing this should also fix a subset of #328 -- update the
description with a new test-case then.
339: "DEFINE-METHOD-COMBINATION bugs"
(reported by Bruno Haible via the clisp test suite)
a. Syntax checking laxity (should produce errors):
i. (define-method-combination foo :documentation :operator)
ii. (define-method-combination foo :documentation nil)
iii. (define-method-combination foo nil)
iv. (define-method-combination foo nil nil
(:arguments order &aux &key))
v. (define-method-combination foo nil nil (:arguments &whole))
vi. (define-method-combination foo nil nil (:generic-function))
vii. (define-method-combination foo nil nil (:generic-function bar baz))
viii. (define-method-combination foo nil nil (:generic-function (bar)))
ix. (define-method-combination foo nil ((3)))
x. (define-method-combination foo nil ((a)))
b. define-method-combination arguments lambda list badness
i. &aux args are currently unsupported;
ii. default values of &optional and &key arguments are ignored;
iii. supplied-p variables for &optional and &key arguments are not
bound.
c. (fixed in sbcl-0.9.15.15)
344: more (?) ROOM T problems (possibly part of bug 108)
In sbcl-0.8.12.51, and off and on leading up to it, the
SB!VM:MEMORY-USAGE operations in ROOM T caused
unhandled condition (of type SB-INT:BUG):
failed AVER: "(SAP= CURRENT END)"
Several clever people have taken a shot at this without fixing
it; this time around (before sbcl-0.8.13 release) I (WHN) just
commented out the SB!VM:MEMORY-USAGE calls until someone figures
out how to make them work reliably with the rest of the GC.
(Note: there's at least one dubious thing in room.lisp: see the
comment in VALID-OBJ)
346: alpha backtrace
In sbcl-0.8.13, all backtraces from errors caused by internal errors
on the alpha seem to have a "bogus stack frame".
349: PPRINT-INDENT rounding implementation decisions
At present, pprint-indent (and indeed the whole pretty printer)
more-or-less assumes that it's using a monospace font. That's
probably not too silly an assumption, but one piece of information
the current implementation loses is from requests to indent by a
non-integral amount. As of sbcl-0.8.15.9, the system silently
truncates the indentation to an integer at the point of request, but
maybe the non-integral value should be propagated through the
pprinter and only truncated at output? (So that indenting by 1/2
then 3/2 would indent by two spaces, not one?)
352: forward-referenced-class trouble
reported by Bruno Haible on sbcl-devel
(defclass c (a) ())
(setf (class-name (find-class 'a)) 'b)
(defclass a () (x))
(defclass b () (y))
(make-instance 'c)
Expected: an instance of c, with a slot named x
Got: debugger invoked on a SIMPLE-ERROR in thread 78906:
While computing the class precedence list of the class named C.
The class named B is a forward referenced class.
The class named B is a direct superclass of the class named C.
[ Is this actually a bug? DEFCLASS only replaces an existing class
when the class name is the proper name of that class, and in the
above code the class found by (FIND-CLASS 'A) does not have a
proper name. CSR, 2006-08-07 ]
353: debugger suboptimalities on x86
On x86 backtraces for undefined functions start with a bogus stack
frame, and backtraces for throws to unknown catch tags with a "no
debug information" frame. These are both due to CODE-COMPONENT-FROM-BITS
(used on non-x86 platforms) being a more complete solution then what
is done on x86.
On x86/linux large portions of tests/debug.impure.lisp have been commented
out as failures. The probable culprit for these problems is in x86-call-context
(things work fine on x86/freebsd).
More generally, the debugger internals suffer from excessive x86/non-x86
conditionalization and OAOOMization: refactoring the common parts would
be good.
354: XEPs in backtraces
Under default compilation policy
(defun test ()
(throw :unknown t))
(test)
Has the XEP for TEST in the backtrace, not the TEST frame itself.
(sparc and x86 at least)
Since SBCL 0.8.20.1 this is hidden unless *SHOW-ENTRY-POINT-DETAILS*
is true (instead there appear two TEST frames at least on ppc). The
underlying cause seems to be that SB-C::TAIL-ANNOTATE will not merge
the tail-call for the XEP, since Python has by that time proved that
the function can never return; same happens if the function holds an
unconditional call to ERROR.
356: PCL corruption
(reported by Bruno Haible)
After the "layout depth conflict" error, the CLOS is left in a state where
it's not possible to define new standard-class subclasses any more.
Test case:
(defclass prioritized-dispatcher ()
((dependents :type list :initform nil)))
(defmethod sb-pcl:validate-superclass ((c1 sb-pcl:funcallable-standard-class)
(c2 (eql (find-class 'prioritized-dispatcher))))
t)
(defclass prioritized-generic-function (prioritized-dispatcher standard-generic-function)
()
(:metaclass sb-pcl:funcallable-standard-class))
;; ERROR, Quit the debugger with ABORT
(defclass typechecking-reader-class (standard-class)
())
Expected: #<STANDARD-CLASS TYPECHECKING-READER-CLASS>
Got: ERROR "The assertion SB-PCL::WRAPPERS failed."
[ This test case does not cause the error any more. However,
similar problems can be observed with
(defclass foo (standard-class) ()
(:metaclass sb-mop:funcallable-standard-class))
(sb-mop:finalize-inheritance (find-class 'foo))
;; ERROR, ABORT
(defclass bar (standard-class) ())
(make-instance 'bar)
]
357: defstruct inheritance of initforms
(reported by Bruno Haible)
When defstruct and defclass (with :metaclass structure-class) are mixed,
1. some slot initforms are ignored by the DEFSTRUCT generated constructor
function, and
2. all slot initforms are ignored by MAKE-INSTANCE. (This can be arguably
OK for initforms that were given in a DEFSTRUCT form, but for those
given in a DEFCLASS form, I think it qualifies as a bug.)
Test case:
(defstruct structure02a
slot1
(slot2 t)
(slot3 (floor pi)))
(defclass structure02b (structure02a)
((slot4 :initform -44)
(slot5)
(slot6 :initform t)
(slot7 :initform (floor (* pi pi)))
(slot8 :initform 88))
(:metaclass structure-class))
(defstruct (structure02c (:include structure02b (slot8 -88)))
slot9
(slot10 t)
(slot11 (floor (exp 3))))
;; 1. Form:
(let ((a (make-structure02c)))
(list (structure02c-slot4 a)
(structure02c-slot5 a)
(structure02c-slot6 a)
(structure02c-slot7 a)))
Expected: (-44 nil t 9)
Got: (SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..
SB-PCL::..SLOT-UNBOUND.. SB-PCL::..SLOT-UNBOUND..)
;; 2. Form:
(let ((b (make-instance 'structure02c)))
(list (structure02c-slot2 b)
(structure02c-slot3 b)
(structure02c-slot4 b)
(structure02c-slot6 b)
(structure02c-slot7 b)
(structure02c-slot8 b)
(structure02c-slot10 b)
(structure02c-slot11 b)))
Expected: (t 3 -44 t 9 -88 t 20)
Got: (0 0 0 0 0 0 0 0)
359: wrong default value for ensure-generic-function's :generic-function-class argument
(reported by Bruno Haible)
ANSI CL is silent on this, but the MOP's specification of ENSURE-GENERIC-FUNCTION says:
"The remaining arguments are the complete set of keyword arguments
received by ENSURE-GENERIC-FUNCTION."
and the spec of ENSURE-GENERIC-FUNCTION-USING-CLASS:
":GENERIC-FUNCTION-CLASS - a class metaobject or a class name. If it is not
supplied, it defaults to the class named STANDARD-GENERIC-FUNCTION."
This is not the case in SBCL. Test case:
(defclass my-generic-function (standard-generic-function)
()
(:metaclass sb-pcl:funcallable-standard-class))
(setf (fdefinition 'foo1)
(make-instance 'my-generic-function :name 'foo1))
(ensure-generic-function 'foo1
:generic-function-class (find-class 'standard-generic-function))
(class-of #'foo1)
; => #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
(setf (fdefinition 'foo2)
(make-instance 'my-generic-function :name 'foo2))
(ensure-generic-function 'foo2)
(class-of #'foo2)
Expected: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS STANDARD-GENERIC-FUNCTION>
Got: #<SB-MOP:FUNCALLABLE-STANDARD-CLASS MY-GENERIC-FUNCTION>
362: missing error when a slot-definition is created without a name
(reported by Bruno Haible)
The MOP says about slot-definition initialization:
"The :NAME argument is a slot name. An ERROR is SIGNALled if this argument
is not a symbol which can be used as a variable name. An ERROR is SIGNALled
if this argument is not supplied."
Test case:
(make-instance (find-class 'sb-pcl:standard-direct-slot-definition))
Expected: ERROR
Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION NIL>
363: missing error when a slot-definition is created with a wrong documentation object
(reported by Bruno Haible)
The MOP says about slot-definition initialization:
"The :DOCUMENTATION argument is a STRING or NIL. An ERROR is SIGNALled
if it is not. This argument default to NIL during initialization."
Test case:
(make-instance (find-class 'sb-pcl:standard-direct-slot-definition)
:name 'foo
:documentation 'not-a-string)
Expected: ERROR
Got: #<SB-MOP:STANDARD-DIRECT-SLOT-DEFINITION FOO>
369: unlike-an-intersection behavior of VALUES-TYPE-INTERSECTION
In sbcl-0.8.18.2, the identity $(x \cap y \cap y)=(x \cap y)$
does not hold for VALUES-TYPE-INTERSECTION, even for types which
can be intersected exactly, so that ASSERTs fail in this test case:
(in-package :cl-user)
(let ((types (mapcar #'sb-c::values-specifier-type
'((values (vector package) &optional)
(values (vector package) &rest t)
(values (vector hash-table) &rest t)
(values (vector hash-table) &optional)
(values t &optional)
(values t &rest t)
(values nil &optional)
(values nil &rest t)
(values sequence &optional)
(values sequence &rest t)
(values list &optional)
(values list &rest t)))))
(dolist (x types)
(dolist (y types)
(let ((i (sb-c::values-type-intersection x y)))
(assert (sb-c::type= i (sb-c::values-type-intersection i x)))
(assert (sb-c::type= i (sb-c::values-type-intersection i y)))))))
370: reader misbehaviour on large-exponent floats
(read-from-string "1.0s1000000000000000000000000000000000000000")
causes the reader to attempt to create a very large bignum (which it
will then attempt to coerce to a rational). While this isn't
completely wrong, it is probably not ideal -- checking the floating
point control word state and then returning the relevant float
(most-positive-short-float or short-float-infinity) or signalling an
error immediately would seem to make more sense.
372: floating-point overflow not signalled on ppc/darwin
The following assertions in float.pure.lisp fail on ppc/darwin
(Mac OS X version 10.3.7):
(assert (raises-error? (scale-float 1.0 most-positive-fixnum)
floating-point-overflow))
(assert (raises-error? (scale-float 1.0d0 (1+ most-positive-fixnum))
floating-point-overflow)))
as the SCALE-FLOAT just returns
#.SB-EXT:SINGLE/DOUBLE-FLOAT-POSITIVE-INFINITY. These tests have been
disabled on Darwin for now.
377: Memory fault error reporting
On those architectures where :C-STACK-IS-CONTROL-STACK is in
*FEATURES*, we handle SIG_MEMORY_FAULT (SEGV or BUS) on an altstack,
so we cannot handle the signal directly (as in interrupt_handle_now())
in the case when the signal comes from some external agent (the user
using kill(1), or a fault in some foreign code, for instance). As
of sbcl-0.8.20.20, this is fixed by calling
arrange_return_to_lisp_function() to a new error-signalling
function, but as a result the error reporting is poor: we cannot
even tell the user at which address the fault occurred. We should
arrange such that arguments can be passed to the function called from
arrange_return_to_lisp_function(), but this looked hard to do in
general without suffering from memory leaks.
379: TRACE :ENCAPSULATE NIL broken on ppc/darwin
See commented-out test-case in debug.impure.lisp.
380: Accessor redefinition fails because of old accessor name
When redefining an accessor, SB-PCL::FIX-SLOT-ACCESSORS may try to
find the generic function named by the old accessor name using
ENSURE-GENERIC-FUNCTION and then remove the old accessor's method in
the GF. If the old name does not name a function, or if the old name
does not name a generic function, no attempt to find the GF or remove
any methods is made.
However, if an unrelated GF with an incompatible lambda list exists,
the class redefinition will fail when SB-PCL::REMOVE-READER-METHOD
tries to find and remove a method with an incompatible lambda list
from the unrelated generic function.
382: externalization unexpectedly changes array simplicity
COMPILE-FILE and LOAD
(defun foo ()
(let ((x #.(make-array 4 :fill-pointer 0)))
(values (eval `(typep ',x 'simple-array))
(typep x 'simple-array))))
then (FOO) => T, NIL.
Similar problems exist with SIMPLE-ARRAY-P, ARRAY-HEADER accessors
and all array dimension functions.
383: ASH'ing non-constant zeros
Compiling
(lambda (b)
(declare (type (integer -2 14) b))
(declare (ignorable b))
(ash (imagpart b) 57))
on PPC (and other platforms, presumably) gives an error during the
emission of FASH-ASH-LEFT/FIXNUM=>FIXNUM as the assembler attempts to
stuff a too-large constant into the immediate field of a PPC
instruction. Either the VOP should be fixed or the compiler should be
taught how to transform this case away, paying particular attention
to side-effects that might occur in the arguments to ASH.
384: Compiler runaway on very large character types
(compile nil '(lambda (x)
(declare (type (member #\a 1) x))
(the (member 1 nil) x)))
The types apparently normalize into a very large type, and the compiler
gets lost in REMOVE-DUPLICATES. Perhaps the latter should use
a better algorithm (one based on hash tables, say) on very long lists
when :TEST has its default value?
A simpler example:
(compile nil '(lambda (x) (the (not (eql #\a)) x)))
(partially fixed in 0.9.3.1, but a better representation for these
types is needed.)
385:
(format nil "~4,1F" 0.001) => "0.00" (should be " 0.0");
(format nil "~4,1@F" 0.001) => "+.00" (should be "+0.0").
386: SunOS/x86 stack exhaustion handling broken
According to <http://alfa.s145.xrea.com/sbcl/solaris-x86.html>, the
stack exhaustion checking (implemented with a write-protected guard
page) does not work on SunOS/x86.
388:
(found by Dmitry Bogomolov)
(defclass foo () ((x :type (unsigned-byte 8))))
(defclass bar () ((x :type symbol)))
(defclass baz (foo bar) ())
causes error
SB-PCL::SPECIALIZER-APPLICABLE-USING-TYPE-P cannot handle the second argument
(UNSIGNED-BYTE 8).
[ Can't trigger this any more, as of 2006-08-07 ]
389:
(reported several times on sbcl-devel, by Rick Taube, Brian Rowe and
others)
ROUND-NUMERIC-BOUND assumes that float types always have a FORMAT
specifying whether they're SINGLE or DOUBLE. This is true for types
computed by the type system itself, but the compiler type derivation
short-circuits this and constructs non-canonical types. A temporary
fix was made to ROUND-NUMERIC-BOUND for the sbcl-0.9.6 release, but
the right fix is to remove the abstraction violation in the
compiler's type deriver.
393: Wrong error from methodless generic function
(DEFGENERIC FOO (X))
(FOO 1 2)
gives NO-APPLICABLE-METHOD rather than an argument count error.
395: Unicode and streams
One of the remaining problems in SBCL's Unicode support is the lack
of generality in certain streams.
a. FILL-POINTER-STREAMs: SBCL refuses to write (e.g. using FORMAT)
to streams made from strings that aren't character strings with
fill-pointers:
(let ((v (make-array 5 :fill-pointer 0 :element-type 'standard-char)))
(format v "foo")
v)
should return a non-simple base string containing "foo" but
instead errors.
(reported on sbcl-help by "tichy")
396: block-compilation bug
(let ((x 1))
(dotimes (y 10)
(let ((y y))
(when (funcall (eval #'(lambda (x) (eql x 2))) y)
(defun foo (z)
(incf x (incf y z))))))
(defun bar (z)
(foo z)
(values x)))
(bar 1) => 11, should be 4.
397: SLEEP accuracy
The more interrupts arrive the less accurate SLEEP's timing gets.
(time (sb-thread:terminate-thread
(prog1 (sb-thread:make-thread (lambda ()
(loop
(princ #\!)
(force-output)
(sb-ext:gc))))
(sleep 1))))
398: GC-unsafe SB-ALIEN string deporting
Translating a Lisp string to an alien string by taking a SAP to it
as done by the :DEPORT-GEN methods for C-STRING and UTF8-STRING
is not safe, since the Lisp string can move. For example the
following code will fail quickly on both cheneygc and pre-0.9.8.19
GENCGC:
(setf (bytes-consed-between-gcs) 4096)
(define-alien-routine "strcmp" int (s1 c-string) (s2 c-string))
(loop
(let ((string "hello, world"))
(assert (zerop (strcmp string string)))))
(This will appear to work on post-0.9.8.19 GENCGC, since
the GC no longer zeroes memory immediately after releasing
it after a minor GC. Either enabling the READ_PROTECT_FREE_PAGES
#define in gencgc.c or modifying the example so that a major
GC will occasionally be triggered would unmask the bug.)
On cheneygc the only solution would seem to be allocating some alien
memory, copying the data over, and arranging that it's freed once we
return. For GENCGC we could instead try to arrange that the string
from which the SAP is taken is always pinned.
For some more details see comments for (define-alien-type-method
(c-string :deport-gen) ...) in host-c-call.lisp.
403: FORMAT/PPRINT-LOGICAL-BLOCK of CONDITIONs ignoring *PRINT-CIRCLE*
In sbcl-0.9.13.34,
(defparameter *c*
(make-condition 'simple-error
:format-control "ow... ~S"
:format-arguments '(#1=(#1#))))
(setf *print-circle* t *print-level* 4)
(format nil "~@<~A~:@>" *c*)
gives
"ow... (((#)))"
where I (WHN) believe the correct result is "ow... #1=(#1#)",
like the result from (PRINC-TO-STRING *C*). The question of
what the correct result is is complicated by the hairy text in
the Hyperspec "22.3.5.2 Tilde Less-Than-Sign: Logical Block",
Other than the difference in its argument, ~@<...~:> is
exactly the same as ~<...~:> except that circularity detection
is not applied if ~@<...~:> is encountered at top level in a
format string.
But because the odd behavior happens even without the at-sign,
(format nil "~<~A~:@>" (list *c*)) ; => "ow... (((#)))"
and because something seemingly similar can happen even in
PPRINT-LOGICAL-BLOCK invoked directly without FORMAT,
(pprint-logical-block (*standard-output* '(some nonempty list))
(format *standard-output* "~A" '#1=(#1#)))
(which prints "(((#)))" to *STANDARD-OUTPUT*), I don't think
that the 22.3.5.2 trickiness is fundamental to the problem.
My guess is that the problem is related to the logic around the MODE
argument to CHECK-FOR-CIRCULARITY, but I haven't reverse-engineered
enough of the intended meaning of the different MODE values to be
confident of this.
404: nonstandard DWIMness in LOOP with unportably-ordered clauses
In sbcl-0.9.13, the code
(loop with stack = (make-array 2 :fill-pointer 2 :initial-element t)
for length = (length stack)
while (plusp length)
for element = (vector-pop stack)
collect element)
compiles without error or warning and returns (T T). Unfortunately,
it is inconsistent with the ANSI definition of the LOOP macro,
because it mixes up VARIABLE-CLAUSEs with MAIN-CLAUSEs. Furthermore,
SBCL's interpretation of the intended meaning is only one possible,
unportable interpretation of the noncompliant code; in CLISP 2.33.2,
the code compiles with a warning
LOOP: FOR clauses should occur before the loop's main body
and then fails at runtime with
VECTOR-POP: #() has length zero
perhaps because CLISP has shuffled the clauses into an
ANSI-compliant order before proceeding.
405: a TYPE-ERROR in MERGE-LETS exercised at DEBUG 3
In sbcl-0.9.16.21 on linux/86, compiling
(declaim (optimize (debug 3)))
(defstruct foo bar)
(let ()
(flet ((i (x) (frob x (foo-bar foo))))
(i :five)))
causes a TYPE-ERROR
The value NIL is not of type SB-C::PHYSENV.
in MERGE-LETS.
406: functional has external references -- failed aver
Given the following food in a single file
(eval-when (:compile-toplevel :load-toplevel :execute)
(defstruct foo3))
(defstruct bar
(foo #.(make-foo3)))
as of 0.9.18.11 the file compiler breaks on it:
failed AVER: "(NOT (FUNCTIONAL-HAS-EXTERNAL-REFERENCES-P CLAMBDA))"
Defining the missing MAKE-LOAD-FORM method makes the error go away.
407: misoptimization of loop, COERCE 'FLOAT, and HANDLER-CASE for bignums
(reported by Ariel Badichi on sbcl-devel 2007-01-09)
407a: In sbcl-1.0.1 on Linux x86,
(defun foo ()
(loop for n from (expt 2 1024) do
(handler-case
(coerce n 'single-float)
(simple-type-error ()
(format t "Got here.~%")
(return-from foo)))))
(foo)
causes an infinite loop, where handling the error would be expected.
407b: In sbcl-1.0.1 on Linux x86,
(defun bar ()
(loop for n from (expt 2 1024) do
(handler-case
(format t "~E~%" (coerce n 'single-float))
(simple-type-error ()
(format t "Got here.~%")
(return-from bar)))))
fails to compile, with
Too large to be represented as a SINGLE-FLOAT: ...
from
0: ((LABELS SB-BIGNUM::CHECK-EXPONENT) ...)
1: ((LABELS SB-BIGNUM::FLOAT-FROM-BITS) ...)
2: (SB-KERNEL:%SINGLE-FLOAT ...)
3: (SB-C::BOUND-FUNC ...)
4: (SB-C::%SINGLE-FLOAT-DERIVE-TYPE-AUX ...)
408: SUBTYPEP confusion re. OR of SATISFIES of not-yet-defined predicate
As reported by Levente M\'{e}sz\'{a}ros sbcl-devel 2006-02-20,
(aver (equal (multiple-value-list
(subtypep '(or (satisfies x) string)
'(or (satisfies x) integer)))
'(nil nil)))
fails. Also, beneath that failure lurks another failure,
(aver (equal (multiple-value-list
(subtypep 'string
'(or (satisfies x) integer)))
'(nil nil)))
Having looked at this for an hour or so in sbcl-1.0.2, and
specifically having looked at the output from
laptop$ sbcl
* (let ((x 'string)
(y '(or (satisfies x) integer)))
(trace sb-kernel::union-complex-subtypep-arg2
sb-kernel::invoke-complex-subtypep-arg1-method
sb-kernel::type-union
sb-kernel::type-intersection
sb-kernel::type=)
(subtypep x y))
my (WHN) impression is that the problem is that the semantics of TYPE=
are wrong for what the UNION-COMPLEX-SUBTYPEP-ARG2 code is trying
to use it for. The comments on the definition of TYPE= probably
date back to CMU CL and seem to define it as a confusing thing:
its primary value is something like "certainly equal," and its
secondary value is something like "certain about that certainty."
I'm left uncertain how to fix UNION-COMPLEX-SUBTYPEP-ARG2 without
reducing its generality by removing the TYPE= cleverness. Possibly
the tempting TYPE/= relative defined next to it might be a
suitable replacement for the purpose. Probably, though, it would
be best to start by reverse engineering exactly what TYPE= and
TYPE/= do, and writing an explanation which is so clear that one
can see immediately what it's supposed to mean in odd cases like
(TYPE= '(SATISFIES X) 'INTEGER) when X isn't defined yet.
409: MORE TYPE SYSTEM PROBLEMS
Found while investigating an optimization failure for extended
sequences. The extended sequence type implementation was altered to
work around the problem, but the fundamental problem remains, to wit:
(sb-kernel:type= (sb-kernel:specifier-type '(or float ratio))
(sb-kernel:specifier-type 'single-float))
returns NIL, NIL on sbcl-1.0.3.
(probably related to bug #408)
410: read circularities and type declarations
Consider the definition
(defstruct foo (a 0 :type (not symbol)))
followed by
(setf *print-circle* t) ; just in case
(read-from-string "#1=#s(foo :a #1#)")
This gives a type error (#:G1 is not a (NOT SYMBOL)) because of the
implementation of read circularity, using a symbol as a marker for
the previously-referenced object.
415: Issues creating large arrays on x86-64/Linux and x86/Darwin
(make-array (1- array-dimension-limit))
causes a GC invariant violation on x86-64/Linux, and
an unhandled SIGILL on x86/Darwin.
416: backtrace confusion
(defun foo (x)
(let ((v "foo"))
(flet ((bar (z)
(oops v z)
(oops z v)))
(bar x)
(bar v))))
(foo 13)
gives the correct error, but the backtrace shows
1: (SB-KERNEL:FDEFINITION-OBJECT 13 NIL)
as the second frame.
418: SUBSEQ on lists doesn't support bignum indexes
LIST-SUBSEQ* now has all the works necessary to support bignum indexes,
but it needs to be verified that changing the DEFKNOWN doesn't kill
performance elsewhere.
Other generic sequence functions have this problem as well.
419: stack-allocated indirect closure variables are not popped
(locally (declare (optimize speed (safety 0)))
(defun bug419 (x)
(multiple-value-call #'list
(eval '(values 1 2 3))
(let ((x x))
(declare (dynamic-extent x))
(flet ((mget (y)
(+ x y))
(mset (z)
(incf x z)))
(declare (dynamic-extent #'mget #'mset))
((lambda (f g) (eval `(progn ,f ,g (values 4 5 6)))) #'mget #'mset))))))
(ASSERT (EQUAL (BUG419) '(1 2 3 4 5 6))) => failure
420: The MISC.556 test from gcl/ansi-tests/misc.lsp fails hard.
In sbcl-1.0.13 on Linux/x86, executing
(FUNCALL
(COMPILE NIL
'(LAMBDA (P1 P2)
(DECLARE
(OPTIMIZE (SPEED 1) (SAFETY 0) (DEBUG 0) (SPACE 0))
(TYPE (MEMBER 8174.8604) P1) (TYPE (MEMBER -95195347) P2))
(FLOOR P1 P2)))
8174.8604 -95195347)
interactively causes
SB-SYS:MEMORY-FAULT-ERROR: Unhandled memory fault at #x8.
The gcl/ansi-tests/doit.lisp program terminates prematurely shortly after
MISC.556 by falling into gdb with
fatal error encountered in SBCL pid 2827: Unhandled SIGILL
unless the MISC.556 test is commented out.
Analysis: + and a number of other arithmetic functions exhibit the
same behaviour. Here's the underlying problem: On x86 we perform
single-float + integer normally using double-precision, and then
coerce the result back to single-float. (The FILD instruction always
gives us a double-float, and unless we do MOVE-FROM-SINGLE it remains
one. Or so it seems to me, and that would also explain the observed
behaviour below.)
During IR1 we derive the types for both
(+ <single> <integer>) ; uses double-precision
(+ <single> (FLOAT <integer> <single>)) ; uses single-precision
and get a mismatch for a number of unlucky arguments. This leads to
derived result type NIL, and ends up flushing the whole whole
operation -- and finally we generate code without a return sequence,
and fall through to whatever.
The use of double-precision in the first case appears to be an
(un)happy accident -- interval arithmetic gives us the
double-precision result because that's what the backend does.
(+ 8172.0 (coerce -95195347 'single-float)) ; => -9.518717e7
(+ 8172.0 -95195347) ; => -9.5187176e7
(coerce (+ 8172.0 (coerce -95195347 'double-float)) 'single-float)
; => -9.5187176e7
Which should be fixed, the IR1, or the backend?
421: READ-CHAR-NO-HANG misbehaviour on Windows Console:
It seems that on Windows READ-CHAR-NO-HANG hangs if the user
has pressed a key, but not yet enter (ie. SYSREAD-MAY-BLOCK-P
seems to lie if the OS is buffering input for us on Console.)
reported by Elliot Slaughter on sbcl-devel 2008/1/10.
422: out-of-extent return not checked in safe code
(declaim (optimize safety))
(funcall (catch 't (block nil (throw 't (lambda () (return))))))
behaves ...erratically. Reported by Kevin Reid on sbcl-devel
2007-07-06. (We don't _have_ to check things like this, but we
generally try to check returns in safe code, so we should here too.)