Ideally, on a crash, the plugin code should send a PluginCrashed event, one per plugin instance, which is how it's done for the NPAPI plugins. Do these new plugins use nsPluginHost et. al.?
Here's the Runnable that dispatches the event: http://mxr.mozilla.org/mozilla-central/source/content/base/src/nsObjectLoadingContent.cpp#309 ----- Original Message ----- From: "Eric Rescorla" <[email protected]> To: "Benjamin Smedberg" <[email protected]> Cc: [email protected], "Felipe Gomes" <[email protected]>, "Georg Fritzsche" <[email protected]> Sent: Tuesday, July 15, 2014 5:34:23 PM Subject: Re: GMP crash reporting update (schedule slip likely) On Tue, Jul 15, 2014 at 1:19 PM, Benjamin Smedberg <[email protected]> wrote: > The bug for hooking up crash reporting for GMP plugins is likely to slip > past this week, and I'd like to talk about how to reduce risk there. > > Part of the problem is that I may not have broken that bug down > sufficiently. There's really two parts to the problem: associating a GMP > crash with one or more DOM windows, and then hooking up the UI. It's not > hard to find peple on fx-team to take care of the UI hookup, but the other > part is hard. It's still not clear to me how to do that, or whether the > answer differs in the case of a webrtc-GMP and a video-decoder-GMP. > > In any case, I'd like to kick of a discussion about this work, especially > since we're right up against 33 branch, and Felipe is fully booked and > going on vacation later next week. Perhaps we should immediately break this > down into the two bugs and somebody from the media team can help with the > GMP->domwindow hookup? > I've been pondering this a bit... We create the GMP structures in: vcmEnsureExternalCodec(): http://dxr.mozilla.org/mozilla-central/source/media/webrtc/signaling/src/media/VcmSIPCCBinding.cpp#68 Which is called from: http://dxr.mozilla.org/mozilla-central/source/media/webrtc/signaling/src/media/VcmSIPCCBinding.cpp#1556 and http://dxr.mozilla.org/mozilla-central/source/media/webrtc/signaling/src/media/VcmSIPCCBinding.cpp#2394 And in both cases these have a pointer to the PC (via PCWrapper). And the PC has a pointer to the DOMWindow: http://dxr.mozilla.org/mozilla-central/source/media/webrtc/signaling/src/peerconnection/PeerConnectionImpl.h#292 So you should be able to trace all this, modulo lifetime issues -Ekr _______________________________________________ dev-media mailing list [email protected] https://lists.mozilla.org/listinfo/dev-media
