http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55939
--- Comment #5 from Mikael Pettersson <mikpe at it dot uu.se> 2013-01-17 23:20:26 UTC --- Created attachment 29198 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29198 standalone test case Here's a standalone test case, extracted from gmp's t-get_d.c. It's largish (444 lines) but most of that are support functions needed for the program logic but otherwise unrelated to the wrong-code issue. The wrong-code occurs in check_random() as a side-effect of the inlining of my_ldexp().