[Bug rtl-optimization/38245] [4.4 Regression] stack corruption when a call is removed but not the outgoing argument pushes

[steven at gcc dot gnu dot org]

------- Comment #14 from steven at gcc dot gnu dot org  2008-12-18 22:35 -------
Created an attachment (id=16939)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16939&action=view)
make all functions with nonzero crtl->outgoing_args_size non-leaf

The result of this patch is that DCE of dead const/pure calls still happens,
but the argument pushes will stay around.  Removing the pushes too is just hard
and not really worth it.

A nice-to-have would be some mechanism to detect this situation where we remove
a call but not the pushes, because it probably will be the result of a missed
optimization on GIMPLE...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38245