Firefox 3.0 compiled with -O2/-O3 will segfault on ppc32 and ppc64/32ul if compiled with the follwoing packages: www-client/mozilla-firefox-3.0-r1 sys-devel/gcc-4.3.1 sys-libs/glibc-2.8_p20080602 Compiling with -O1 fixes the problem. I'll attach a backtrace from my G4. [ first mentioned at http://forums.gentoo.org/viewtopic-t-697303.html ]
Created attachment 158041 [details] backtrace from powerbook g4
*** Bug 228999 has been marked as a duplicate of this bug. ***
Backtrace with debug symbols in attachment 158061 [details]
I'm not sure your root cause is quite correct. I still have gcc 4.2.4 on my system, so I recompiled xulrunner, mozilla-firefox, and even glibc using gcc 4.2.4, but firefox still crashes on startup. Going to give a -O1 compile a try next. I hesitate to try a glibc downgrade, as I have no desire to accidentally trash my system.
So everyone here has glibc-2.8? Because i don't get that on glibc-2.6.1 and gcc-4.1.2.
(In reply to comment #5) > So everyone here has glibc-2.8? Correct. > Because i don't get that on glibc-2.6.1 and gcc-4.1.2. I'm currently on the road and cannot test firefox-3.0 on my "almost stable" system at home using gcc-4.3.1 or gcc-4.2.3, but I'll do once I'm home again.
(In reply to comment #5) > So everyone here has glibc-2.8? Right. > Because i don't get that on glibc-2.6.1 and gcc-4.1.2. Not surprised. This bug seems to be related to some combination of glibc-2.8, gcc-4.3.x, and -O2. I tried rebuilding with -O1 last night (though I realised I forgot to switch back to gcc 4.3, so this was with 4.2), and I get no segfault. I'll try a couple more combinations to see what comes out of this, but don't expect much in the way of results before the weekend. I'll undoubtedly slow down a bit since I have something that works now...
its not gcc related, occurs the same with 4.1.2 && glibc-2.8. However it doesn't occur with gcc-4.1.2 && glibc-2.6, glibc-2.7.
Can i get `emerge --verbose --info | grep ARCH` from a ppc64/32bitul? Thanks
I've added the workaround for this in the mozcoreconf-2.eclass Thanks to Brent Baude for giving me the output.
(In reply to comment #10) > I've added the workaround for this in the mozcoreconf-2.eclass Isn't that backwards? elif [[ ${ARCH} == ppc ]] && ! has_version '>=sys-libs/glibc-2.8'; then mozconfig_annotate "more than -O1 segfaults on ppc with glibc-2.8" \ --enable-optimize=-O1 Why the negation before has_version?
(In reply to comment #11) > (In reply to comment #10) > > I've added the workaround for this in the mozcoreconf-2.eclass > > Isn't that backwards? > > elif [[ ${ARCH} == ppc ]] && ! has_version '>=sys-libs/glibc-2.8'; then > mozconfig_annotate "more than -O1 segfaults on ppc with glibc-2.8" \ > --enable-optimize=-O1 > > Why the negation before has_version? > Yeah, fixed now, i just put that to test it.
