On 11/11/11 20:13, Hans-Peter Nilsson wrote: > AFAICT, your patch has got sufficiently testing now (on three > targets to boot) to be considered safe to check in. Or is > something amiss? > > (If it's the unchecked code quality you mentioned, that can be > just as well dealt with having the tree in a working state; > having the tree broken for some targets accomplishes nothing.)
I briefly looked at that, and while it does tend to move stuff around compared to an unpatched compiler, the effects don't seem to be all that bad. If we find a testcase where it's a real problem, we could do more aggressive reordering after epilogue generation, when we no longer have exit fallthroguh edges. So I'd like to ask for an OK here. Bernd