[Bug lto/47274] [4.6 regression] ICE in lto_varpool_replace_node, at lto-symtab.c:306

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47274

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-01-17 13:49:47 UTC ---
I can't reproduce this with the testcase from comment #3 on x86_64-linux
with either gold or GNU ld from stock binutils 2.21 nor without using
the linker plugin.

The only call to lto_varpool_replace_node sees

Breakpoint 5, lto_varpool_replace_node (vnode=0x7ffff5d041a0, 
    prevailing_node=0x7ffff5d040d0)
    at /space/rguenther/src/svn/trunk/gcc/lto-symtab.c:304
304       if (vnode->needed)
(gdb) n
306           gcc_assert (!vnode->analyzed || prevailing_node->analyzed);
(gdb) p vnode->analyzed
$1 = 0
(gdb) p prevailing_node->analyzed
$2 = 1

Note that the assert probably should do &&, not ||.  Or it doesn't
make sense at all.  Asserting that the prevailing_node is analyzed
should be enough.

Honza?

People, you have to show us what goes wrong for you.