media player consistently crashes

[kvdman]

I've been testing mainly wav and mp3s through media player. It crashes since about 1 week.

See attached picture.

[jackburton]

looks like #2759, so it happens with any file, it seems...

[dlmcpaul]

My only thoughts looking at the method shown is why the BAutoLock is casting "this" in PlayerCount when it doesn't in NewWindow which is the 2 places that are protecting fPlayerCount variable.

BWindow* MainApp::NewWindow() {

BAutolock _(this); fPlayerCount++; return new MainWin();

}

int32 MainApp::PlayerCount() const {

BAutolock _(const_cast<MainApp*>(this)); return fPlayerCount;

}

[jackburton]

Replying to dlmcpaul:

My only thoughts looking at the method shown is why the BAutoLock is casting "this" in PlayerCount when it doesn't in NewWindow which is the 2 places that are protecting fPlayerCount variable.

PlayerCount() is const, so you couldn't use BAutoLock without casting.

[kvdman]

'looks like #2759, so it happens with any file, it seems... '

I haven't tried video, but basically any audio file I through at it would crash media player.

[jackburton]

I think I've found the problem, but I don't have any idea on how to solve this correctly. Basically, what happens is that The window thread starts (sometimes) before the MainApp constructor returns, and receives a B_WINDOW_ACTIVATED message from the app_server. Then, the WindowActivated() code calls Controller::PlayerActivated(), which in turn dereferences gMainApp. gMainApp isn't yet initialized, since the MainApp constructor hasn't returned yet. I've fixed the crash by initializing gMainApp in the MainApp constructor, like this:

gMainApp = be_app

and moving the creation of the window after that line. But again, I'm not sure it's a good solution. Maybe we should, in some way, delay the activation of the window after the BApplication has been run.

[mmu_man]

I used the same workaround in hrev27764 but indeed it's not clean, so the ticket should remain open until it's fixed.

[jackburton]

What do we do about this ?

[stippi]

The fix is not actually a work-arround, but is a completely valid fix for a race-condition. There is nothing unclean about it.