Solaris 2.6 bootstrap failure: /usr/include/stdlib.h:111: arguments given to macro `abort'

[Zack Weinberg]

Jeffrey A Law wrote:
>   In message < 199908261643.JAA08021@zack.bitmover.com >you write:
>   > > When I bootstrap on Solaris2.7 and SunOS4 using cc in both cases, it
>   > > successfully builds stage1, but then inifinite loops on the first file
>   > > in stage2.
>   > 
>   > I have seen similar things on x86.  I *think* this is due to bugs in
>   > the recent resource and flow changes, nothing to do with me.
> Can you double check this after Bernd's latest patch to gcse.c?  I thought
> I had installed a particular change, but didn't.  The net result is it was
> possible for gcse.c to go into an infinite loop looking for copy propagations

Bootstrap succeeds now, but I'm still seeing core dumps in the test
suite.  noncompile/920923-1.c and noncompile/971104-1.c.  920923-1 is
because expand_end_case has been called when case_stack is NULL.
971104-1 is because make_label_edge has been called with a label for
which INSN_UID is nonzero but BLOCK_FOR_INSN is NULL.

zw