------- Comment #62 from fxcoudert at gcc dot gnu dot org 2008-02-27 22:28 ------- That new testcase fails under Darwin at -O0 (because of a faulty libm) and passes with optimization turned on (because the middle-end is very clever). So, there two solutions are: either have 1 FAIL (-O0) and the rest PASS, like I had before; or have 1XFAIL and the rest XPASS, which is was we have after my last commit.
The first situation was clearly better, but I was at work when Dominique emailed my after my first commit, saying the new testcase FAILed, and I got confused. I think I'll revert to the old situation, and we'll live with that one FAIL until Apple fixes the bug or the universe stops expanding and condenses into a single point of pure energy, whichever happens first. (My money is on the second guess.) -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25477
