------- Comment #11 from doko at debian dot org 2009-05-13 16:28 ------- Subject: Re: [4.4/4.5 regression] symbols ce...@glibcxx_3.4.3 not exported
bkoz at gcc dot gnu dot org schrieb: > ------- Comment #10 from bkoz at gcc dot gnu dot org 2009-05-12 23:01 ------- > > Sorry for the delay Paolo, this fix looks fine. > > I see tanhl missing too in that log, yet the gnu.ver exports have it and so > does src/math_compatibility_long_double.cc, although depending on > _GLIBCXX_HAVE_TANHL. What gives? Is this some kind of temporary build issue or > is this still indicative of configure/build issues? no that was accidentally removed in your suggested workaround for PR39491. > Please note that there are no arm ABI baselines. Is this due to > multilib/configure issues and or large numbers of variants? Can Debian > volunteers fix this or elaborate? Hopefully the necessity of this is now > apparent. I'll ask the Debian ARM port maintainers. would welcome feedback from GCC ARM maintainers as well. > I appreciate the current efforts to get Debian builds back on track, but am a > bit alarmed at the releases w/ known ABI fails on secondary arches. What can > be > done to improve this situation? baseline files would help. Debian itself does include symbol files as well, but in this case I didn't update those when adding the new armeabi support. But I only will notice such regressions when building the package for Debian or Ubuntu, which I did for arm not before the 4.4 release. Matthias -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40038
