Re: GCC build failed for native with your patch on

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    That doesn't mean your patch doesn't cause the current bootstrap
    failure.  In fact, I checked, by backing out the patches responsible
    for the previous failure, and your patch definitely causes bootstrap
    to fail. 

DJE came to the opposite conclusion, so something odd is going on here.

    A big clue that your patch was broken should have been the 38
    powerpc-eabisim regressions.  

RTH just fixed those.  Did that fix the bootstrap problem you see?

    Another clue is that the location of the bootstrap failure changed;
    before it was a compare failure, now it's a crash while building
    stage1 libgcc.

Hmm.  It might be a good idea for the regression tester to say that.
When it said "it also failed before you patch", I ignored the failure.
Had it said it was a different failure, I would have treated it differently.

    In fact, this is why I need you to fix your patch ASAP; it's
    preventing me from further work.  What I'd like to do is have the
    tester consider .o file miscomparisons as another kind of testcase
    failure.

I think the tester should consider failures at different points
as different: compilation failures, failures during the build of stage2,
failures during running of stage2, comparison failures, and failures
building other libraries.  This will avoid the sort of problem we had here.