(Sorry, I hit "send" way to early in the previous mail)
"Robert E. Brown" wrote:
> I just reported these bugs to a CMU CL mailing list, but they apply to sbcl
> as well, so I'm reporting them here too. I would have taken a stab at
> fixing these myself, but I'm having trouble building the HEAD revision of
> sbcl from the CVS repository. Transcript of the breakage is attached below.
> I think I've found an off-by-one bug in the code to print floating point
> numbers. Evaluating (format t "~,9,,3,,,EE" 0.000000001) produces
> 1000.0000000e-12E instead of 100.00000000e-12E.
I confirm that (in SBCL-0.6.12.48), but it should probably be
> Also, I think there's a bug in the handling of t or * in cons type
> specifiers. Both of the following evaluate to (values nil nil). I expect
> (values t t).
> (subtypep '(cons (member foo) t) '(cons (member foo) *))
> (subtypep '(cons (member foo) *) '(cons (member foo) t))
Both of these do evaluate to (values t t) in SBCL-0.6.12.48.
Which version of SBCL are you using?
(Onto the breakage):
> obj/from-host/src/compiler/generic/early-objdef.lisp-obj-tmp written
> compilation finished in 0:00:00
> creating directory: obj/from-host/src/compiler/target/
> compiling file "/arm/software/cvs-trees/sbcl/src/compiler/x86/parms.lisp"
> (written 11 JUN 2001 09:44:07 AM):
> debugger invoked on SB-DEBUG:*DEBUG-CONDITION* of type UNBOUND-VARIABLE:
> error in SB-KERNEL::UNBOUND-SYMBOL-ERROR-HANDLER: The variable
> #:EXPR-TMP-11 is unbound.
> 0: [ABORT] Return to toplevel.
> Within the debugger, you can type HELP for help.
> (EVAL #:EXPR-TMP-11)
This looks like a bug in the interaction with CMUCL again.
Your output looks like you are using CMUCL.
contains a hack that made it possible to build SBCL with CMUCL, because
there were problems with the DEFCONSTANT-EQX macro. CMUCL obviously has
a bug in EVAL-WHEN handling, and code src/compiler/x86/parms.lisp
has some DEFCONSTANT-EQX stuff in it, that probably causes your
You should probably use SBCL to build from CVS, which worked for me.
I've tried to rebuild the current HEAD branch with CMUCL (18c), but if
failed when compiling/loading src/compiler/main.lisp.
Need to look into that.
Has anybody successfully built SBCL with CMUCL recently?
Martin Atzmueller <martin@...>