------- Comment #2 from armin76 at gentoo dot org 2009-09-30 05:34 ------- (In reply to comment #1) > I just rebuilt a vanilla gcc-4.4.1 with itself on armv5tel-linux-gnueabi, and > as usual there were no problems. I even used your LDFLAGS=-Wl,-O1 and > STAGE1_CFLAGS=-O settings. > Weird, guess i'll try with a debian chroot...
> A few notes: > - you seem to be building in the source dir; I don't think that's supported, I > haven't done that in a decade or so No, ls -l /var/tmp/portage/sys-devel/gcc-4.4.1/work/ total 16 drwxr-xr-x 9 root root 4096 Sep 28 11:19 build drwxr-xr-x 26 root root 4096 Sep 28 11:13 gcc-4.4.1 drwxr-xr-x 3 root root 4096 Jul 23 00:34 patch drwxr-xr-x 2 root root 4096 Jul 23 06:00 uclibc > - in the gentoo bug entry there is an indication that you're using > binutils-2.18, I'm using binutils-2.19.1; normally I wouldn't worry but you're > passing -O1 to ld so who knows? Now i'm running 2.19.1 and i've tried 2.20.51.0.1, same result. > - in the gentoo bug entry there is some speculation that gcc runs out of > memory > on 512MB RAM machines; I've rebuilt 4.4.1 with itself on arm boxes with as > little as 64MB without much problems (as long as java is disabled); of course > you need ample swap, at least temporarily It's not the case, the user that said so posted his log error on IRC and he clearly got a killed cc1, not a segfault. Anyway i have 3G of ram. > - I don't think "-pipe" is a good idea on small memory machines > - I do use --enable-checking=release to avoid gcc growing too large and slow, > that may be a factor given the size-related post-4.3 bootstrap failure PR > > I'm not convinced this is a gcc-4.4.1 bug. > I'm thinking this happens when you build gcc-4.4.1 with that make line, and then you build a gcc with it...i'll try. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41500