[agb2]

Pi3b.
Since LE #328 and OSMC 17.8-048 Kodi crashes under the following circumstances:
Start playing any video [mine are on a Windows box, shared via SMB].
Bring up the Video OSD by hitting 'm' on keyboard.
Let it continue for a few seconds.
Exit OSD by hitting 'm' again'.

Crash - Kodi crash, not reboot.

Any one else observed this?

Reverting to LE #327 and OSMC 17.8-047 fixes issue.

I will find a crashlog from LE once I've reproduced with stock Estuary [which I don't normally use].
Will also try a local rather than network file.
No reason to suspect skin as it happens with both Estuary V1 and Xonfluence 3.3.17

[Milhouse]

I'm not currently able to reproduce with the latest #0329 build (stock Estuary).

Please see if you can reproduce with stock Estuary, and without any additional addons as this could be add-on related.

A crashlog (with debug logging enabled) would be useful.

[Milhouse]

OK, when I stopped the video (I'd been playing it for about 5 minutes when I posted above) with the OSD on screen, *then* kodi scrashed.

Steps to reproduce:

1) Play a video
2) Open Video OSD
3) Stop video playback -> crash

I'll post a debug-enabled crashlog shortly. This is a normal crashlog for now: http://sprunge.us/AUSc

[asavah]

@Milhouse I ve seen this issue in my build,
but since I use highly experimental 0day stuff (like gcc-linaro-snapshots) I thought it may be related to my builds only.

I've built a Debug Kodi and traced the crash with gdb.
The crash occurs here ->
https://github.com/xbmc/xbmc/blob/master....cpp#L1200
which gets called from ->
https://github.com/xbmc/xbmc/blob/master...SD.cpp#L98

It seems that dynamic_cast is the problem, but only in this case/call.
The regression was introduced here -> https://github.com/xbmc/xbmc/commit/6af4...ff=unified

I can't tell if the issue is arm or linux specific, or if it's another gcc bug, it's above my pay grade and skill level

Dirty patch (partial revert) to fix this:
https://pastebin.com/cje3eGcw


You may want to ask a skilled c++ dev about this.

[Milhouse]

Thanks @asavah

This is an RPi2 LibreELEC debug-enabled crashlog: http://sprunge.us/TVGA - the crash occurs on the same line as yours.

I can't reproduce in LE x86_64 nor latest Kodi master on Ubuntu 16.04 so this is either arm specific or RPi specific (ie. newclock5).

The regression you've identified is part of PR11869 which first appeared in LE test build #0321, however this build isn't crashing - playback ends normally with no issue.

The first build that *is* crashing when playback is stopped while the OSD is enabled, is #0328, so maybe a later change in this build (possibly one of the newclock5 commits) is now playing badly with the earlier PR11869 changes?

[asavah]

The first build that *is* crashing when playback is stopped while the OSD is enabled, is #0328, so maybe a later change in this build (possibly one of the newclock5 commits) is now playing badly with the earlier PR11869 changes?

Can't tell as I don't build every day, there is a 10 days difference between my last working build and the failing one.
I use newclock5 as base, with occational PRs from master which were not merged to popcornmix's branch at the moment of building.

Hmmm may be this https://github.com/popcornmix/xbmc/commi...4294643012 ??

[Milhouse]

Same crash on RPi1 with LE, which might suggest it is less likely to be a gcc/arch-type issue.

[asavah]

This seems to be the fix for the problem https://github.com/xbmc/xbmc/pull/11927

... that's what means to be a true professional programmer

Thanks again @popcornmix

[kl1k]

Nightly osmc updated and another user confirmed its fixed a crashing issue he was having too.

[agb2]

Crash is fixed in OSMC 17.8-051.