[PATCH] Fix eh ICE with -foptimize-sibling-calls

[Richard Henderson]

On Fri, Jan 11, 2002 at 05:13:24PM +0100, Jakub Jelinek wrote:
> ftp://people.redhat.com/gcc/testsuite/eh_sibcall.C

-rw-------    1 2140     2140      1301421 Jan 11  2002 eh_sibcall.C

chmod please?

> The problem is that when expand_call in pass 0 actually emits any eh
> regions, sibcall_failure is set and that entire insn chain is thrown away.
> Unfortunately, this throwing away is not communicated back to except.c,
> which still includes it in exception_handler_labels and its other data
> structures.

Huh?  except.c shouldn't care what you do during initial rtl 
generation.  exception_handler_labels is recomputed from scratch
at the beginning of finish_eh_regions.

I think your patch is on the wrong track here.


r~