i (and some other ppl on the forum) have come to the conclusion, that some packages built with gcc-r2 will be broken and segfault right away. some examples are xmms, nethack, wine. it seems that not all user configurations are affected (see this thread: http://forums.gentoo.org/viewtopic.php?t=84803).
Confirmed, just emerged the new openssh packages and my SSH client segfaulted. I'm downgrading GCC to 3.2.3-r1 again. Portage 2.0.49-r4 (default-x86-1.4, gcc-3.3.1, glibc-2.3.2-r1, 2.6.0-test5-mm1) ================================================================= System uname: 2.6.0-test5-mm1 i686 AMD Athlon(TM) MP 1800+ ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-O2 -march=athlon-mp -pipe -fomit-frame-pointer" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/share/config /usr/kde/2/share/config /usr/kde/3/share/config /usr/X11R6/lib/X11/xkb" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-O2 -march=athlon-mp -pipe -fomit-frame-pointer" DISTDIR="/usr/portage/distfiles" FEATURES="sandbox ccache autoaddcvs" GENTOO_MIRRORS="ftp://ftp.snt.utwente.nl/pub/os/linux/gentoo" MAKEOPTS="-j3" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://rsync.gentoo.org/gentoo-portage" USE="x86 avi crypt cups encode foomaticdb jpeg kde libg++ libwww mad mmx mpeg ncurses pdflib png quicktime spell truetype xml2 xmms xv zlib alsa gdbm berkdb slang readline ggi tcltk java X sdl tcpd pam ssl perl python esd imlib oggvorbis gnome gtk motif opengl mozilla 3dnow acpi curl dvd ethereal faad gtk2 imap ipv6 jikes lcms ldap md5sum mdb mng moznoirc moznomail pic S3TC slp snmp sse tiff usb v4l wmf xfs xinerama xvid -oss -apm -arts -gif -gpm -mikmod -nls -qt -svga"
You can add kernel menuconfig to the list of programs that segfault. gcc-3.3.1-r1 disappeared, I think someone needs to fetch this from the CVS Attic, wipe -r2, and drop -r1 back in before too many people horribly break their systems.
Confirmed: XMMS (1.2.8-r1 and 1.2.8-r2) is segfaulting here with 3.3.1-r2
I can confirm that two system with gcc 3.3.1-r1 built perfectly fine with openssh-3.7p1 but two other system with gcc 3.3.1-r2 are totally fubar now and neither openssh 3.7p1 or older versions of openssh build at all. You all probably know it already but this bug appears to be just related to 3.3.1-r2 only.
Isn't this more like blocker given that ~x86 defaults to a broken version that produces broken packages?
You can add the compiler itself as a package that won't compile itself again! This was a BAAD idea to replace a reasonably working 3.3.1-r1 and kill our systems. I can't even compile the compiler itself, and I can't boot normally since checkfs fails because e2fsprogs got compiled..... And now -r1 is missing from the tree. What we need is a working binary package of -r1 put to a publically accessable place ASAP!
I've masked -r2. There's a bin pkg of -r1 at http://dev.gentoo.org/~avenj/bins/
I already tried that binary tarball, but it is broken. Seems like the file hasn't been completely uploaded. I wrote an email to avenj to tell him about that but still haven't got answer.
oops ... it was you ... sorry, didn't look at your email in the bugzilla post :(
Actually, I did reply to that email... in any case, it turns out you're right. I'm putting up a new one as we speak.
Here are 2 more Here are 2 more http://forkbomb.dhs.org/gcc-3.3.1-r1.tbz2 http://forkbomb.dhs.org/gcc-3.2.3-r2.tbz2 CFLAGS="-march=pentium3 -O3 -pipe"
*** Bug 28901 has been marked as a duplicate of this bug. ***
Easy fix: # mv $(gcc-config -L)/libgcc_s.a $(gcc-config -L)/libgcc.a and remerge whatever you merged with -r2.
Martin's trick did it! But I think it's important to run ldconfig and to reboot after moving the library, to be sure that it's correctly reloaded, in case it's already cached. Am I wrong? In fact this morning I tried moving out of the way libgcc.a (the script) after that apps segfaulted and without rebooting, and it didn't work. But I've been to lazy to reboot...
I think this is important enough to make an announcement to the list(s) and the forums in an attempt to curb the problems that could potentially occur. No?
No need for ldconfig, as its a static library. Then, we did already mailed to -dev and Ulrich (plate) did post to the forums. Thanks.
*** Bug 28899 has been marked as a duplicate of this bug. ***
+ if [ -f "${D}${LIBPATH}/libgcc.a" ] + then + mv -f ${D}${LIBPATH}/libgcc.a ${D}${LIBPATH}/libgcc_s.a + cat > ${D}${LIBPATH}/libgcc.a <<EOF +/* GNU ld script + Use the shared library, but some functions are only in + the libc.a static library, so try that secondarily. */ +GROUP ( libgcc_s.a /usr/lib/libc.a ) +EOF + fi + This comes from the -r2 ebuild. It is not in the -r1 ebuild, and seen as though that is what Martins little fix reverses, I would think is the cause of the problems. Reversing it though may cause more hassles
*** Bug 28942 has been marked as a duplicate of this bug. ***
No, reversing it with the command given, or taking that out of the build and remerging will have no after effects. Just remerge whatever you merged with it.
Is it worth saying, that although I have downgraded to -r1, I had no problems with anything built with -r2. In case it helps someone debug, my info is: ortage 2.0.49-r4 (default-x86-1.4, gcc-3.3.1, glibc-2.3.2-r1, 2.4.20-gentoo-r6) ================================================================= System uname: 2.4.20-gentoo-r6 i686 AMD Athlon(tm) MP 2400+ ccache version 2.2 [enabled] ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-mcpu=athlon-mp -O2 -pipe" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/kde/2/share/config /usr/kde/3/share/config /usr/X11R6/lib/X11/xkb /usr/kde/3.1/share/config /usr/share/texmf/tex/generic/config/ /usr/share/texmf/tex/platex/config/ /usr/share/config" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-mcpu=athlon-mp -O2 -pipe" DISTDIR="/usr/portage/distfiles" FEATURES="sandbox autoaddcvs ccache fixpackages" GENTOO_MIRRORS="http://www.mirror.ac.uk/sites/www.ibiblio.org http://gentoo.oregonstate.edu http://www.ibiblio.org/pub/Linux/distributions/gentoo" MAKEOPTS="-j3" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://rsync.europe.gentoo.org/gentoo-portage" USE="x86 oss apm avi cups encode foomaticdb gif jpeg libg++ mad mikmod mmx mpeg ncurses nls pdflib png quicktime spell xml2 xmms xv zlib gtkhtml alsa gdbm berkdb slang readline arts tetex bonobo svga tcltk java guile ruby mysql sdl gpm tcpd pam libwww ssl perl python esd imlib oggvorbis qt motif opengl mozilla ldap cdr X kde gtk gtk2 ptl gnome dvd pda crypt truetype" running on a dual Athlon MP system. Let me know if you need specifics, Tom
btw, -r2 is the (failed) try to solve some problems that appear with -r1 but does this by going *BACK* - to branch-update of 20030815 -- while -r1 is already brach-update of 20030904 - so IMHO what we need is a better solution for the problems that appear with -r1 not really a running version based on this r2-ebuild.
I do not see what this have to do with anything - if you work with cvs, you either: 1) backout if a broken patch was commited 2) wait for the fix to be commited 3) fix it yourself I did 1), as I do not know enouth about the parser code to fix it, and do not have the time (and as this is not stable builds ...). You are open to do 3), but that is anyhow fixed with -r3, so please just use that. Anyhow, -r3 is out which fixes this.
Sorry, I didnt mean to insult you - I agree, it was worth trying - unfortunately it didn't work... thanks for r3.