Ubuntu
xfce4-appfinder package

  1. Bug #1048805
  2. Comment #13

Comment 13 for bug 1048805

Revision history for this message
Snild (snild) wrote : Re: xfrun4 launches very slowly

It happened again.

The last note in my previous comment was wrong; there _is_ an appfinder instance running:
$ ps -Tg 4030
  PID SPID TTY TIME CMD
 4030 4030 ? 00:00:00 xfce4-appfinder
 4030 4031 ? 00:00:00 gdbus
 4030 4032 ? 00:00:00 xfce4-appfinder

Like before, DBus has another PID (4029) for the owner of "org.xfce.Appfinder", which is probably the case because of daemonization.

Attached to it with gdb.

(gdb) info threads
  Id Target Id Frame
  3 Thread 0x7faf4897d700 (LWP 4031) "gdbus" 0x00007faf5c8c83cd in poll ()
   from /lib/x86_64-linux-gnu/libc.so.6
  2 Thread 0x7faf43fff700 (LWP 4032) "xfce4-appfinder" 0x00007faf5cbb182c in __lll_lock_wait () from /lib/x86_64-linux-gnu/libpthread.so.0
* 1 Thread 0x7faf5f0189c0 (LWP 4030) "xfce4-appfinder" 0x00007faf5cbb182c in __lll_lock_wait () from /lib/x86_64-linux-gnu/libpthread.so.0

Two "normal" threads, and one gdbus thread. The two first threads are waiting on a lock (or a condition?). The gdbus thread, which is currently polling for an event, would have to be the one to awaken at least one of these threads. That's reasonable -- no reason to do stuff until an event comes in.

I set a breakpoint on the next instruction in poll (after the syscall returns). It appears it doesn't recieve any events; starting another appfinder instance doesn't cause any breaks in the gdb-attached process. Hm. Looks like I may have to learn some DBus. :)

Installed d-feet, tried to look at the org.xfce.Appfinder bus. It shows nothing for a while, then pops up an error message saying "org.xfce.Appfinder : Timeout was reached". I guess that rules out any problem in the "client" code (well, maybe it could be made more robust by trying to check for this condition first, but it shouldn't be needed).

At this point, I wasn't sure what to do next, so I figured I'd kill the running appfinder daemon and see what happens.

Well, the next start was fine. It daemonized successfully, too. And all following starts worked fine as well. So at least the problem isn't persistent in the dbus-daemon. It could still be a dbus problem, of course, but it seems unlikely. So, what causes the appfinder service to stop responding? I still have no idea.