Re: hunt for mainline regressions

[Janis Johnson <janis187 at us dot ibm dot com>]

On Wed, Dec 18, 2002 at 08:16:22AM +0100, Christian Jönsson wrote:
> On Tue, Dec 17, 2002 at 04:44:07PM -0800, Janis Johnson wrote:
> > My regression hunting is pretty well automated now, but I'm still
> > figuring out new little tricks and then updating my scripts and
> > howto document accordingly.  I hope to send those out again soon for
> > review.
> > 
> > I'm selecting regressions to hunt somewhat randomly from among the ones
> > for which Wolfgang Bangerth has identified a half-month range when it
> > occurred.  If you'd like me to search for a particular regression from
> > his list, let me know.
> 
> well, not sure this is on his list, but anyhow, c++/8183 has been a
> problem for a while.

It's not on his list because the bug doesn't show up on i686-linux.
I tried it on ia64-linux and powerpc64-linux and don't see it on either
of those; if I can't reproduce it, I can't hunt for the patch that
caused it to start failing.
 
> Sometime Sept 12 2002 the mainline, then 3.3 (experimental) worked for
> sparc64-{linux,solaris*} targets. Then, sometime, it stopped. I think
> before the end if Sept 2002.
> 
> I really don't know how to hunt the bug down... :-( it might be kernel
> related, it might very well be libc related...

I don't see how an ICE would be related to the kernel or libc.  Can you
reproduce it by doing a simple build (rather than a bootstrap) of G++,
using a recent version of GCC for the build?  If so, you might want to
try my new scripts.

Janis