rif <[EMAIL PROTECTED]> writes:

> Error in function LISP::ASSERT-ERROR:
>    The assertion (AND
>                   (EQ (C::IR2-CONTINUATION-PRIMITIVE-TYPE C::2CONT)
>                       C::FUNCTION-PTYPE)
>                   (EQ C::CHECK T)) failed.
>    [Condition of type SIMPLE-ERROR]
> 
> The function regress-n is in src/cllib/stat.lisp in clocc, and has not
> changed from the last (working) version of clocc I have (working
> against an older CMUCL 18e binary), leading me to suspect a CMUCL issue
> rather than a clocc issue.  Any ideas?

I think I've found that bug.  It seems to have something to do with
intersections of function types.  As a workaround until this gets
fixed (I'll put a possibly temporary fix in CVS soon), I suggest
taking out the FTYPE declaration of REGRESS-N and maybe the FUNCTION
declaration of the FUNC parameter in REGRESS-N.

But there's another problem revealed by CLOCC that I don't seem to be
able to track down.  CMUCL enters an infinite loop or something while
compiling CLOCC's rng.lisp, which uses #+NEGATIVE-ZERO-IS-NOT-ZERO,
which I think we deleted, RANDOM and a lot of other float stuff, of
course.  Maybe someone with better knowledge of that area can help
out?

> Sorry to be so clueless, I just really want to update my CMUCL but I
> keep having problems.  Thanks to everyone for being so helpful.

To the contrary; thanks for testing the beta!

Reply via email to