Mythbackend deadlock with slave backend connected

[redpepperracing@…]

This appears to be different from 9721, 9704 and was not fixed by commit 3e2375faa09956dd3f32d862a156d57fc4127668.

I can reliably reproduce the problem with current trunk by starting a backend, a slave backend, and a remote frontend, and trying to enter live tv. The backtraces from the two BE and BE and remote FE logs are attached.

All systems are running freshly compiled trunk:

Please attach all output as a file in bug reports. MythTV Version : v0.25pre-1825-g6496880 MythTV Branch : master Network Protocol : 65 Library API : 0.25.20110421-1 QT Version : 4.6.3 Options compiled in:

linux debug use_hidesyms using_alsa using_oss using_backend using_bindings_perl using_bindings_python using_bindings_php using_dvb using_frontend using_hdhomerun using_hdpvr using_iptv using_ivtv using_joystick_menu using_lirc using_mheg using_opengl_video using_qtdbus using_qtwebkit using_v4l using_x11 using_xrandr using_bindings_perl using_bindings_python using_bindings_php using_mythtranscode using_opengl using_vdpau using_ffmpeg_threads using_live using_mheg

All systems are also running up to date FC13 installs. The master has 2x HDHR tuners (4 total) and 2x PVR-x50 tuners, the slave has 1x HD-PVR via USB, the slave records to an NFS mounted storage group from the MBE.

If you need any more info or debug testing, let me know.

[redpepperracing@…]

Master back end backtrace

[redpepperracing@…]

Slave back end backtrace

[redpepperracing@…]

Master back end log

[redpepperracing@…]

Slave back end log

[redpepperracing@…]

Frontend log

[sphery]

In 3e2375faa:

Fix a potential deadlock in AutoExpire?

With danielk's changes that I merged in, he has us use Sleep() internally almost exclusively. However, Sleep() requires the instance_lock to be pre-locked, and in this place, it wasn't. This can cause problems.

[zeloise@…]

Hello

the problem is still there even after commit 3e2375faa.

I hope this helps zeloise

[Github]

Initialize update_thread in ctor

This seems to help with avoiding the deadlocks when slave backends attach, and should be otherwise non-harmful.

Fixes #9745

Signed-off-by: Gavin Hurlbut <ghurlbut@…>

Changeset: f33e9a69c254d879be6d42e766755493519c2839

[beirdo]

If this did not in fact fix it, please re-open.

[zeloise@…]

Problem solved with commit f33e9a

thanks zeloise