Re: [Sbcl-devel] sparc breakage

[Nathan F. <fr...@cs...>]

On Wed, Aug 11, 2004 at 03:55:48PM +0100, Christophe Rhodes wrote:
> In line with the "many eyes makes bugs shallow" philosophy, I warmly
> encourage as many as are able to examine the 0.8.13.47 patch (findable
> on the sbcl-commits list, or from 
>   cvs diff -u -D'2004-08-10 00:18 UTC' -D'2004-08-10 00:22 UTC'
> ) to identify the problem(s).  In line with my own schedule, I'm
> afraid I am leaving my electronic toys behind until Sunday night, and
> as such I cannot assist with the audit. :-/
> 
> So I hope that the culprit (the code culprit, that is, not the human
> one: we know who /that/ is :-) is found long before I return... but if
> not, I shall then attempt to help.

Looking at CVS history makes clear that .47 is my fault--perhaps I
was a little overzealous in making things work for 32-bit and 64-bit
ports.

I've looked at the diff and I can't see much wrong with it.  I even
toyed with the idea that my changes to NWORDS might have an effect,
but the attached C test file for experimenting with my idea shows
there shouldn't be any difference between the two NWORDS implementations.
The C runtime changes suggest themselves as the source of the
difficulty (as the Lisp changes are more-or-less easily verified to
be harmless), but the only big change there is to NWORDS...

I'm building everything on Solaris now, so perhaps the badness will
be found soon.
-- 
Nathan | From Man's effeminate slackness it begins.  --Paradise Lost