This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug ipa/60871] [4.9/4.10 Regression] internal compiler error: in possible_polymorphic_call_targets, at ipa-devirt.c:1510
- From: "hubicka at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 01 May 2014 08:19:19 +0000
- Subject: [Bug ipa/60871] [4.9/4.10 Regression] internal compiler error: in possible_polymorphic_call_targets, at ipa-devirt.c:1510
- Auto-submitted: auto-generated
- References: <bug-60871-4 at http dot gcc dot gnu dot org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60871
--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
This is yet another case where get_binfo_at_offset incorrectly returns NULL
(wich is in there for long time, but before we started sanity check it just led
to missed optimizations).
In class D there is base C7 and A (that is virtual base of C). They share vtbl
pointer, but get_binfo_at_offset is organized in an assumption that sharing of
pointer doesn't happen over sibling bases.
I guess easiest fix would be to recursively dive into all bases with a given
offset, but I will try to look through C++ FE implementation first to see if
there is more elegant way around.
This may affect ipa-devirt, too, that may need to push vtable into the stack
even in the sibling walk.
Honza