[Bug lto/57602] [4.9 Regression] Runfails for several C/C++ benchmarks from spec2000 for i686 with -flto after r199422

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57602

--- Comment #8 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> ---
And a testcase from PR57879 fails with -O2 / -O1 and -flto, it happens when
building GCC itself.