[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-bugs-dist
Subject:    [Bug 297671] ktorrent UI freezes, on start of app. and during download of torrents
From:       Michael Pyne <mpyne () kde ! org>
Date:       2012-04-10 1:02:36
Message-ID: bug-297671-17878-tFiBfWrpIi () http ! bugs ! kde ! org/
[Download RAW message or body]

https://bugs.kde.org/show_bug.cgi?id=297671

--- Comment #4 from Michael Pyne <mpyne@kde.org> ---
So I seem to be in a KDE session not affected by this issue after some
cleaning/sanitization...

Some of the debugging I've done to this point is to strace processes that are
waiting on a reply from DBus (kontact, klauncher, kded4 especially). There was
nothing especially illuminating unfortunately (and was done at a Linux TTY and
I forgot to log the strace output). The strace output consisted mostly of
poll()/recvmsg() syscalls, where poll() would timeout (once a second) and
recvmsg() would give EAGAIN.

Looking at the fd that recvmsg() was being called in (under /proc/$PID/fd and
/proc/$PID/fdinfo), the recvmsg() was being called on a Unix domain socket,
with flags of close-on-exec, read/write, and non-blocking set.

So in this case EAGAIN is actually innocuous: It just means that there is
nothing to receive (similar to EWOULDBLOCK).

Based on reports of this same type of issue (even from years past), and based
on the file descriptors open for affected processes according to /proc/$PID/fd,
I tried unlinking $KDEHOME/cache-$HOSTNAME/{libphonon/*,ksycoca} while outside
of KDE and restarting. That didn't help.

However (and I forget what prompted me to look at this, I think I saw
permission denied warnings), running the ipcs command to list open SysV IPC
primitives showed some semaphores and shared memory segments opened to other
users from previous logins. I used ipcrm to remove those stole IPC primitives
and *so far* everything has been working properly in this desktop session.

If anyone has any clue why SysV IPC would affect some DBus connections
(permissions?) I'd be happy to try and reproduce. This is the kind of thing
that even if it's an "installation issue" is nearly impossible to debug and so
should be identified by the runtime (or better yet, made such that it's not a
problem at all).

-- 
You are receiving this mail because:
You are watching all bug changes.
[prev in list] [next in list] [prev in thread] [next in thread]