4 GCC regressions, 3 new, with your patch on 2001-06-04T18:55:45Z.

[Geoff Keating]

> Date: Mon, 4 Jun 2001 15:58:22 -0400
> From: Phil Edwards <pedwards@disaster.jaj.com>
> Cc: brendan@zen.org, jh@suse.cz, lars@nocrew.org, pme@sources.redhat.com,
>         theonetruekenny@yahoo.com, bkoz@redhat.com

> On Mon, Jun 04, 2001 at 08:58:08PM +0000, GCC regression checker wrote:
> > With your recent patch, GCC has some regression test failures, which
> > used to pass.  There are 3 new failures, and 1
> > failures that existed before and after that patch; 0 failures
> > have been fixed.
> > 
> > The new failures are:
> > powerpc-eabisim libstdc++-v3.sum 21_strings/element_access.cc
> > powerpc-eabisim libstdc++-v3.sum 26_numerics/complex_value.cc
> > native g++.sum g++.pt/crash67.C
> > 
> > The old failures, which were not fixed or introduced by your patch, are:
> > native libstdc++-v3.sum 26_numerics/complex_value.cc
> > 
> > For more information, see < http://www.cygnus.com/~geoffk/gcc-regression/ >.
> 
> > --- /maat/heart/tbox/changelog_pass/gcc/ChangeLog	Thu May 10 15:20:14 2001
> > +++ gcc/ChangeLog	Fri Jun  1 12:03:02 2001
> [snip 1005 lines of ChangeLog never mentioning libstdc++]

It gets truncated to avoid sending huge mail messages.  I don't have a
good solution to this, other than to have people fix the tree so that
it never gets that long.  It's really bad that the tree has been
broken for so long; I didn't expect that this sort of thing would
happen when I was writing the tester, I had hoped that when people saw
that their patches caused problems they would fix them, not leave bugs
in the tree for weeks.

> I'm a little confused on how it determined that our patches were to blame,
> if it hasn't passed since May 10th.  Our changes aren't even listed in
> the changelog diff included here (or on the web version).

The ChangeLog is since it last passed, but the list of people it sends
mail to is those that have changed the tree since the tester last ran.

I'm open to suggestions on how to improve this.  Would it help if the
message listed the changes since it last ran instead since it last passed?

> Scratch "a little", I'm /very/ confused.  I saw no regressions locally.
> Can someone confirm the powerpc-eabisim thing?  Can't build one of those
> here.

The first two new failures seem to be a bug.  When the tester sees a
build failure, it forgets about the previous regressions it saw.  This
has now been fixed (I hope!).

The third one, crash67.C, is certainly new.

-- 
Geoff Keating <geoffk@redhat.com>