http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57018
--- Comment #2 from michael at talamasca dot ocis.net <michael at talamasca dot ocis.net> 2013-04-21 04:36:34 UTC --- I've always been in the habit of specifying --host=i386-... so that my binaries don't vary based on where I compile them. I tried a few -march options to no effect. However, I think I've found the reason. One unusual thing about my GCC is that I compiled with --enable-frame-pointer even though I always manually reverse that from the command line. This is because GCC bloats all omit-frame-pointer code with exception tables otherwise. (I wish GCC had an "-fundebuggable" option to express what I've always meant by -fomit-frame-pointer all these years...) I tried compiling my case with "-Os -fomit-frame-pointer -fasynchronous-unwind-tables" and that suppressed the bug.
