stale lock prevents apport runs

bah, the lock file was not dated Sep 5th... it was Aug 5th. Sorry.

I have the same issue. It seems like a X server crash left the stale lock file around. (I'm using Option "NoTrapSignals" "true" in xorg.conf, so apport gets to know about the X server crashes).

This debdiff makes apport use /var/lock/apport as a lockfile instead.
It also makes it exit with code 1 in case the lockfile could not get created (bug 147237).

I've also added a "/bin/rm -f /var/lock/apport" to the init script, in case apport gets restarted.

In fact, the lockfile would not have to be moved to /var/lock: deleting it in the init script's "start" action would be enough. But it seems like having lockfiles in /var/lock sounds reasonable.

The .lock file is not stale at all. It does not matter if it exists, since the lock is done using flock(2), not by merely testing the existence of that file. So the patch would not help in any way. So the reason for this must be entirely different. Can you please attach your /var/log/apport.log?

I guess then that the reason has been that the lockfile from "root" was left and then already the
fd = os.open(lockfile, os.O_WRONLY|os.O_CREAT|os.O_NOFOLLOW)
failed.

I've just had X crashing again and the lock file left, owned by root.

Daniel, no, that's not it. First, apport is always run as root. Second, if the os.open() fails, apport writes "cannot create lock file" into apport.log and exits. If you got "another apport instance is already running, aborting", then the flock() call failed. If it usually works, then I guess there really was another apport instance running at that time. If it generally fails, it should be reproducible with

  python -c "import os, fcntl; fd = os.open('/var/lock/.crash', os.O_WRONLY|os.O_CREAT|os.O_NOFOLLOW); fcntl.lockf(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)"

Does that work for you?

Whoops, sorry. Of course I meant

  sudo python -c "import os, fcntl; fd = os.open('/var/crash/.lock', os.O_WRONLY|os.O_CREAT|os.O_NOFOLLOW); fcntl.lockf(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)"

Yes, the code snippet works for me and I believe I've created confusion here - not knowing about flock(2) and that apport is always being run as root, sorry.

hggdh, can you give more information? Has the problem happened to you again?

Since then I have been monitoring the logs and /var/crash. I never saw it happening again.

So, sorry, no new data. I guess we could close it invalid then.

Thanks, hggdh, for reporting back. Let's close this then until it happens again.