Re: [Sbcl-devel] Re: compiler inconsistency

[William H. N. <wil...@ai...>]

On Wed, Jan 30, 2002 at 08:47:18AM +0300, Alexey Dejneka wrote:
> Another way to see it is to set SB-C::*CHECK-CONSISTENCY* to T.

As far as I could tell, the *CHECK-CONSISTENCY* code was sadly
out-of-date in the original cmucl-ca.-2.4.8 code that I forked from,
and I never used it after a few failed attempts. Now I'm sure it's
more out of date. I'd love to get it back in shape, and/or move the
not-too-expensive assertions into ordinary (not conditional on
*CHECK-CONSISTENCY*) compiler code.
 
But I've spent hardly any time experimenting with it, and mostly just
assumed that getting the *CHECK-CONSISTENCY* stuff to run cleanly
would be a major project. Have you been having some success in getting
it to run in general? Maybe it might not be too difficult to make it
run always?

> > But the word DESIRABLE embarrasses me. Should this function link if
> > it is "wanted" (in this case the AVER is right and there is a bug
> > upstream) or "wanted and possible"?
> 
> There is only one client of TARGET-IF-DESIRABLE now and it was
> broken. So I think the first variant is right. I've fixed the bug and
> added some checks and cross-references.

OK, thank you. Pending things for me to deal with are now, IIRC,
  1. this patch
  2. your new test case for "The value NIL is not of type SB-C:TN",
     which should at least be logged in a BUGS
  3. your new-in-0.7.0 compiler bug report, which I hope I might
     be able to fix
I'm currently moving to a new apartment, so I might be more
scatterbrained and unpredictable than usual, but I expect I'll deal
with the first two in the next day or two, and I still hope to put
some time into the third too.

-- 
William Harold Newman <wil...@ai...>
"Look on my works, ye Mighty, and despair!" -- Ozymandias, King of Kings 
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C