Merge lp://staging/~darkxst/unity-settings-daemon/lp1377847 into lp://staging/unity-settings-daemon

Thanks, small nitpick in the inline comments.

Are you sure nothing is using the removed symbols? I know they've only been there for a short time.

Also would you please remove the d/changelog editing? We'll upload this through CI train which will do that automatically from the commit message and I'm not sure if it'll add the bug reference with the manually edited version.

Laney, existing programs access the idle monitor through gnome-desktop (or gnome-session). gnome-shell does although its been using the new api for ages, and besides that is provided by mutter. which only leaves u-s-d (previously before roberts patches it would have called through via gnome-desktop as well), but there is little point for it to call out over dbus only to get itself.

Using this version I just had a segfault in gsd_idle_monitor_handle_xevent_all()

since the ppa doesn't have ddebs I can get a debug backtrace, but it segfaults in a call to g_hash_table_contains

what I did
- using unity, lock the screen manually
- walk away from the computer for half an hour
- come back and move the mouse to get the lock screen/unlock

It doesn't seem to happen every time with those steps though (or the time without touching the computer matters, I tried just waiting for the screen to dim/turn off when I did testing)

That was the bug with the previous version, can you try building the package yourself and see if you can reproduce it using this branch? I didn't get a crash all day yesterday. At least if you do your own build you'll get a ddeb for a better backtrace.

The build is the current version from https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/landing-026

Got the ddeb from http://ddebs.ubuntu.com/queue/utopic/

the bt is

"#0 0x206b6e61 in ?? ()
#1 0xb6d687f8 in g_hash_table_lookup_node (hash_return=<synthetic pointer>,
    key=0xe00015, hash_table=0x94b5b90)
    at /build/buildd/glib2.0-2.42.0/./glib/ghash.c:368
#2 g_hash_table_contains (hash_table=0x94b5b90, key=0xe00015)
    at /build/buildd/glib2.0-2.42.0/./glib/ghash.c:1267
#3 0xb76b01f4 in gsd_idle_monitor_handle_xevent (alarm_event=0xbfcdd9ec,
    alarm_event=0xbfcdd9ec, monitor=0x94a46a0) at gsd-idle-monitor.c:225
#4 gsd_idle_monitor_handle_xevent_all (xevent=0xbfcdd9ec)
    at gsd-idle-monitor.c:248
..."

On Thu, Oct 09, 2014 at 09:14:57AM -0000, Sebastien Bacher wrote:
> The build is the current version from https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/landing-026
>
> Got the ddeb from http://ddebs.ubuntu.com/queue/utopic/
>
> the bt is
>
> "#0 0x206b6e61 in ?? ()
> #1 0xb6d687f8 in g_hash_table_lookup_node (hash_return=<synthetic pointer>,
> key=0xe00015, hash_table=0x94b5b90)
> at /build/buildd/glib2.0-2.42.0/./glib/ghash.c:368
> #2 g_hash_table_contains (hash_table=0x94b5b90, key=0xe00015)
> at /build/buildd/glib2.0-2.42.0/./glib/ghash.c:1267
> #3 0xb76b01f4 in gsd_idle_monitor_handle_xevent (alarm_event=0xbfcdd9ec,
> alarm_event=0xbfcdd9ec, monitor=0x94a46a0) at gsd-idle-monitor.c:225
> #4 gsd_idle_monitor_handle_xevent_all (xevent=0xbfcdd9ec)
> at gsd-idle-monitor.c:248
> ..."

It's the same as what we had before with earlier PPA versions - I think
that somehow the device monitor wasn't being disposed properly so we
were accessing some random bit of memory which was crashing inside glib.

With the current version I can't get this crash any more. Previously if
I used the mouse it was quite likely to crash while fading, but now this
never happens for me.

Have you managed to make it happen again?

--
Iain Lane [ <email address hidden> ]
Debian Developer [ <email address hidden> ]
Ubuntu Developer [ <email address hidden> ]

no, it's a one time thing from this morning, I tried to run under valgrind as well, no error there

On Thu, Oct 09, 2014 at 10:29:01AM -0000, Sebastien Bacher wrote:
> no, it's a one time thing from this morning, I tried to run under valgrind as well, no error there

I'll keep trying all day. Not sure what we should do, the bug this is
fixing is important but this crash might happen more often across all
users and will be hard to fix if it's very difficult to reproduce. :(

--
Iain Lane [ <email address hidden> ]
Debian Developer [ <email address hidden> ]
Ubuntu Developer [ <email address hidden> ]

seems to trigger when going to unity-greeter (e.g try selecting another user from the indicator)

Different problem, but I think I fixed this in https://code.launchpad.net/~laney/unity-settings-daemon/lp1377847-2/+merge/237846

hopefully... it's hard to gdb these refcounting bugs, I should learn to use refdbg. :)

thanks for the update, triggering a silo rebuild with r4060.

The changes look fine, small nitpick, some g_object_(un)ref calls lack a space before the '('

fixed the spaces

not sure why that merge request is still open but the change have been merged previous cycle