Hi, I'm running gentoo on an AMD 1.4ghz w/ 512 megs of ram. Every time I try to emerge mozilla (or KDE, or OpenOffice, or even gaim) the compile will run and then segfault. I tried changing the -O flags, screwing around with -mcpu /-march and other flags. But no matter what it will always segfault. I tried rebuilding gcc (thinking that it might of gotten screwed up) but it segfaulted during that too! Where gcc segfaults changes depending on what optimizations I use, but it will always segfault before finishing. Here are logs from my past attempts @ mozilla and gaim. (regenerating the KDE/gcc error would take a long long time.) But I could p[rovide them if you wish. Find logs at: http://people.bu.edu/tomk/moz-gaim-emerge-err.tar.gz
Also I'm using linux-kernel 2.4.19-gentoo-r9 it's pretty much a vanilla install, except it's a preemptible kernel and has usb support built in.
Same here, usb uhci/ehci compiled in, from a stage 3 k6-2/i586 install. I tried recompiling the kernel (which failed also) as did gcc. I have a 512m swap space enabled, but it also seems as if I do something that involves heavy disk usage gcc will die then also (eg, cd /usr/portage ; grep -r "gtk+-2.0") It was somehow able to compile GNOME / KDE / mplayer, but gcc / mozilla / kernel fail. Programs run fine once compiled though. gcc version 3.2
Thinking it may have been a filesystem issue (I'm using ReiserFS [15 gigs]) I formated a new partition [using ext3] and mounted it to /var/tmp. Unfortunately it still segfaults. Also I'm using a 1 gig swapspace.
CFLAGS ? -march=pentium4,athlon-tbird,k6 is known to cause segfaults in gcc-3.2
I was using -march=athlon-xp does that have problems also? And do you mean GCC compiled with -march flags causes segfaults, or compiling programs with GCC using the march flag causes segfaults? Thanks ;) -tom.
I mean that passing some args to gcc-3.2 (like -march=pentium4) causes it to bork with some compiles (xfree, mozilla, a few others). It should be fine with -march=athlon-xp. Just verify. Also consider checking memory.
Ran memtest86 on it, all checks out fine. I'm using march=athlon-xp, so it shouldnt have any problems. I switched to mcpu=athlon-xp, but it's still segfaulting. I'm compiling the vanilla linux kernel right now, to see if using a different kernel will solve the problem... We'll see if it does tomorrow. -tom.
SAME HERE... gcc 3.2 segfaults. I'm using a fresh gentoo 1.4 install. When I compile kde 3.1beta2 or mozilla.... I'm using -march=pentium3 -mcpu=pentium3 -O3. I tried to switch this down and could finaly compile kdelibs with -O2. But kdebase doesn't even compile with -march=i686 -mcpu=i686... I tried over and over again... Everything else compiles fine (xfree 4.2.1, gnome, xine, etc...)
Hrm, wonder if it could not be a problem that too many files are open ? Had that today .... anybody can remember how you increase it ?
I have an AMD Athlon 800 here, Gentoo 1.4-rc2, GCC 3.2.1, and it is segfaulting on ebuilds, preventing me from getting a GUI desktop. First Segfault was on QT, which I solved by changing MAKEOPTS="-j1", however Arts 1.1.0 will not compile. I am using CFLAGS="-march=athlon -O3 -pipe", if it matters.
Try gcc-3.2.2_pre20030131 .. it fixes a few problem like that, especially K6 cpu, but maybe it will for you as well ...
After this bug I gave up on gentoo. I recently tried it again, and the same thing happened. A friend suggested I turn down the FSB on my motherboard [the chip|board|mem are advertized to work at 133, which is what it is set to]. Anyways I turned it down to 100mhz and everything compiled just fine. It must have been a hardware issue. In general the computer works fine at 133 FSB, but whenever I have to do large C++ compiles [kde, mozilla, OO, etc] it has to be set at 100. I guess C++ just puts too much of a strain on the sys. Once things are compiled everything works fine. ;)