[Bug tree-optimization/63778] [5 Regression] Segfault with r217178 building 416.gamess from cpu2006

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63778

--- Comment #7 from rguenther at suse dot de <rguenther at suse dot de> ---
On Mon, 10 Nov 2014, hjl.tools at gmail dot com wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63778
> 
> --- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> ---
> (In reply to Richard Biener from comment #3)
> > I can't reproduce it on x86_64-linux, neither with unreduced zheev.F
> > -DSPEC_CPU
> > nor with -march=native added.  Trying a cross to ppc64-linux (BE).
> > 
> > HJ, please provide more detailed instructions (I'm on r217281 now).
> 
> It is fixed by r217260.  Should we add a testcase for this?

Probably - can you reduce sth?  It's fixed by the hunk

Index: genmatch.c
===================================================================
--- genmatch.c  (revision 217259)
+++ genmatch.c  (revision 217260)
@@ -1726,6 +1726,7 @@ dt_node::gen_kids (FILE *f, bool gimple)
   if (exprs_len || fns_len)
     {
       fprintf (f, "case SSA_NAME:\n");
+      fprintf (f, "if (do_valueize (valueize, %s) != NULL_TREE)\n", 
kid_opname)
;
       fprintf (f, "{\n");
       fprintf (f, "gimple def_stmt = SSA_NAME_DEF_STMT (%s);\n", 
kid_opname);