------- Comment #2 from rguenth at gcc dot gnu dot org 2010-02-15 12:19 ------- (In reply to comment #1) > If you really think this is a purely libstdc++ issue we gonna need *a lot* of > help from the compiler people. Also consider that Jonathan, the author of the > recent improvements to std::bind, will be in vacations for 2 weeks.
Well, usually FAILs at -O0 but not -O2 hint at lifetime problems such as references to local vars being returned. Note that a patch of mine exposes this issue at -O2 ... so it's blocked by this issue. > Anyway, how can this be a [4.5 Regression] if the testcase didn't exist in 4.4 > and, more specifically, uses C++0x features which do not make sense together > with the 4.4 std::bind, which is just was the std::tr1::bind in the std:: > namespace? So the testcase is bogus? Then please remove it. I marked it as a regression prematurely because the patch I really want to apply will expose it at -O2 - which then makes it a regression against an earlier revision of trunk. And I'm quite lost in the myriads of variadic templates when trying to figure out what is going wrong (I tried for several hours already). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43075
