Just updated baselayout from 1.8.6.11 to 1.8.6.12 and at the end it runs grpconv which took up 512MB ram and 256MB swap and was then OOM killed. It left group.lock and gshadow.lock in /etc and changed the modified time on group but made no other changes I can see. Reproducible: Didn't try Steps to Reproduce: 1.update baselayout 2. 3. Actual Results: ... * Removing invalid backup copies of critical config files... * Updating module dependencies... /usr/sbin/ebuild.sh: line 1325: 32588 Terminated /usr/sbin/grpconv * Please be sure to update all pending '._cfg*' files in /etc are updated, * else things might break at your next reboot! You can use 'etc-update' * to accomplish this: ... dmesg shows: Out of Memory: Killed process 32588 (grpconv). Expected Results: not use all my ram and bring the system to it's knees with a swap storm Portage 2.0.49-r15 (default-1.0, gcc-2.95.3, glibc-2.2.5-r8, 2.4.20-gentoo-r8a) ================================================================= System uname: 2.4.20-gentoo-r8a i686 AMD Athlon(tm) processor Gentoo Base System version 1.4.3.12 ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-march=i686 -O3 -pipe" CHOST="i686-pc-linux-gnu" COMPILER="" 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="-march=i686 -O3 -pipe" DISTDIR="/mnt/linuxback/portage/distfiles" FEATURES="sandbox ccache autoaddcvs digest userpriv usersandbox strict" GENTOO_MIRRORS="http://linux.oit.umass.edu/gentoo/gentoo http://www.ibiblio.org/pub/Linux/distributions/gentoo" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/mnt/linuxback/portage" PORTDIR_OVERLAY="/mnt/linuxback/portage-local" SYNC="rsync://rsync.us.gentoo.org/gentoo-portage" USE="x86 oss 3dnow apm avi crypt cups jpeg libg++ mikmod mmx mpeg ncurses pdflib spell truetype xml2 xmms xv berkdb bonobo cdr directfb esd gdbm ggi gif gnome gnome-libs gpm gtk gtkhtml guile imlib libwww motif mozilla nls oggvorbis opengl pam perl png python readline sdl slang snmp ssl svga tcltk tcpd tiff X -kde -qt -arts -java dvd encode gtk2 -quicktime"
If you run it from console, does this still happens ?
yup
Could you merge 'strace' and 'ltrace' and attach logs for both of '[sl]trace -ff grpconv' ?
Created attachment 20843 [details] strace.gz strace...once it gets into the brk(*) part it just keeps going incrementing until it's killed.
Created attachment 20844 [details] ltrace.gz ltrace loops...2 full loops and the start of a third are captured
Try to update to -r9 of glibc ?
still does it w/ glibc-2.2.5-r9
Same here...
from the grpconv manual page ... BUGS Errors in the password or group files (such as invalid or duplicate entries) may cause these programs to loop forever or fail in other strange ways. Please run pwck and grpck to correct any such errors before converting to or from shadow passwords or groups. ... running grpck prior to grpconv fixes the problem for me
grpck fixed it for me. I had duplicate nogroup, sshd, and apache entries.
*** Bug 40898 has been marked as a duplicate of this bug. ***
So basically this should fix it?: -- Index: baselayout-1.8.6.13.ebuild =================================================================== RCS file: /home/cvsroot/gentoo-x86/sys-apps/baselayout/baselayout-1.8.6.13.ebuild,v retrieving revision 1.3 diff -u -r1.3 baselayout-1.8.6.13.ebuild --- baselayout-1.8.6.13.ebuild 9 Feb 2004 04:12:24 -0000 1.3 +++ baselayout-1.8.6.13.ebuild 15 Feb 2004 21:28:28 -0000 @@ -703,9 +703,18 @@ # Enable shadow groups (we need ROOT=/ here, as grpconv only # operate on / ...). if [ "${ROOT}" = "/" -a \ - ! -f /etc/gshadow -a -x /usr/sbin/grpconv ] + ! -f /etc/gshadow -a -x /usr/sbin/grpck -a -x /usr/sbin/grpconv ] then - /usr/sbin/grpconv + /usr/sbin/grpck &>/dev/null + if [ "$?" -eq 0 ] + then + /usr/sbin/grpconv + else + echo + ewarn "Running 'grpck' returned errors. Please run it by hand, and then" + ewarn "run 'grpconv' afterwards!" + echo + fi fi # Simple Release version for testing of features that *should* be -- Anyhow, commited to cvs. Please reopen if there are any more issues.