On Mon, 19 Dec 2011, Jakub Jelinek wrote:
> Hi!
>
> We were ignoring maybe_clean_or_replace_eh_stmt return value,
> which tells us if stmt previously could throw but doesn't any longer.
> I was a little bit worried that gimple_purge_dead_eh_edges from within
> cfg cleanup wouldn't work well, but at least bootstrap/regtested
> on x86_64-linux and i686-linux didn't show any issues. Ok for trunk?
Ok.
Thanks,
Richard.
> 2011-12-19 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/51596
> * tree-cfg.c (replace_uses_by): Call gimple_purge_dead_eh_edges
> when needed.
>
> * g++.dg/opt/pr51596.C: New test.
>
> --- gcc/tree-cfg.c.jj 2011-12-16 17:34:07.000000000 +0100
> +++ gcc/tree-cfg.c 2011-12-19 11:47:12.554921183 +0100
> @@ -1627,7 +1627,8 @@ replace_uses_by (tree name, tree val)
> if (fold_stmt (&gsi))
> stmt = gsi_stmt (gsi);
>
> - maybe_clean_or_replace_eh_stmt (orig_stmt, stmt);
> + if (maybe_clean_or_replace_eh_stmt (orig_stmt, stmt))
> + gimple_purge_dead_eh_edges (gimple_bb (stmt));
>
> update_stmt (stmt);
> }
> --- gcc/testsuite/g++.dg/opt/pr51596.C.jj 2011-12-19 11:49:43.548046185
> +0100
> +++ gcc/testsuite/g++.dg/opt/pr51596.C 2011-12-19 11:48:49.000000000
> +0100
> @@ -0,0 +1,39 @@
> +// PR tree-optimization/51596
> +// { dg-do compile }
> +// { dg-options "-O -fnon-call-exceptions" }
> +
> +struct A { float v[2]; };
> +struct B { int v[2]; };
> +
> +struct C
> +{
> + B c;
> + C f ()
> + {
> + B b;
> + for (int i = 0; i < 2; i++)
> + b.v[i] = c.v[i];
> + return *this;
> + }
> +};
> +
> +struct D
> +{
> + A d;
> + D (B x)
> + {
> + for (int i = 0; i < 2; i++)
> + d.v[i] = x.v[i];
> + }
> +};
> +
> +int bar ();
> +
> +C i;
> +
> +void
> +foo ()
> +{
> + while (bar ())
> + D (i.f ().c);
> +}
>
> Jakub
>
>
--
Richard Guenther <rguent...@suse.de>
SUSE / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer
