When I try to emerge or unmerge an application, it says (for example): >>> emerge dev-lang/python-2.2.1 to / and then locks up. This shows my process list (with sandbox using about 98% cpu): 7813 pts/1 00:00:00 sandbox 7814 pts/1 00:00:00 ebuild.sh <defunct> I then have to kill sandbox manually with kill -9 Updating is still possible with 'emerge rsync' and '--pretend' works also fine I tried to use the portage rescue image (to no avail, but it did complain about not being able to unlink directories). I also did a emerge --clean I do think that it could have started when an earlier emerge failed, cause I remember seeing some kind of error (on not being able to compile maybe). Sorry I can't be more specific the bug is discussed in the forums: http://forums.gentoo.org/viewtopic.php?t=1419 Portage 1.8.19 Athlon CPU (compiled with P6 instruction set) kernel 2.4.18-xfs
Have you tried to disable the sandbox and see if the problem goes away? put "-sandbox" in your FEATURES in /etc/make.conf. If the problem is still there, then that is not a sandbox problem. If it goes away when you disable the sandbox, then upgrade your portage, re-enable the sandbox and see if the problem still exists.
I disabled sandbox in make.conf, resulting in the following errors: ACCESS DENIED open_wr: /usr/portage/python-2.2.1 Sandbox error : the SANDBOX_DENY environmental variable should be defined. Sandbox error : the SANDBOX_READ environmental variable should be defined. Sandbox error : the SANDBOX_WRITE environmental variable should be defined. Sandbox error : the SANDBOX_PREDICT environmental variable should be defined. And this keeps looping. Also, (again) this happens with every application, so python above is an example. Appearantly it still is a sandbox problem ??
Some progress. I installed binary packages of portage (version 1.9.10) python and dependencies (debianutils, fileutils and python-fcksum). That didn't help. But then I disabled sandbox again, and now emerge works! So it is definately sandbox related.
I did some more testing with the ebuild script: ebuild clean, unpack, compile and install all do their job after that sandbox is started automatically (don't know why) and locks up ebuild merge locks up before doing anything I checked that the ebuild script gets run all the way (except maybe for merge), and it does. So the problem does not originate in the ebuild script. I finally decided to bootstrap my system again to see if that would help, but the problem stayed (yes, i thought that was strange too). Because I didn't want to waste more time on it, I then reinstalled my system.
Bert, Still an issue?
No problems anymore :) But only after I reinstalled my system. I have never seen the problem again. Still I think its weird.
I had a very similar thing happen on my system with portage 2.0.1 - the machine crashed during an emerge, and then whenever I tried to run emerge without --pretend, sandbox would get stuck in a loop at 99% CPU use. The problem was (I think) in this bit of code in sandbox.c: while (EOF != fscanf(pids_stream, "%d\n", &tmp_pid)) { if (0 == kill(tmp_pid, 0)) { ... After the reboot, I had a stale /tmp/sandboxpids.tmp laying around, and somehow sandbox got stuck in this loop. ltrace showed kill() being called, and that's the only kill in sandbox.c. Removing sandboxpids.tmp fixed everything.
Does this still happen with 2.0.9?
ugh oops? this isn't my bug? Heh, sorry, carry on ;o)
