On 7/15/2014 6:10 PM, Maire Reavy wrote:
On 7/15/2014 5:55 PM, Randell Jesup wrote:
On 7/15/2014 5:00 PM, Benjamin Smedberg wrote:
On 7/15/2014 4:58 PM, Felipe Gomes wrote:
Sounds good, but then it will always use the infobar. Are they
fundamentally unable to fire at a particular <object>?
Yes. There is no <object>, just a JS API for webrtc.
RTCPeerConnection is new'd, but is connected to an innerWindow ID in
PeerConnection.js. Not certain if that helps, but it may be the key
to figuring out an association that you can use. Connecting a
specific decoder instance to a window ID will require some work, but
I believe is possible.
We'll work with you to figure out what we can do here.
Agreed.
Bsmedberg -- Does this give you what you need, or do you need us to
expose some API? If so, what do you want it to look like?
To round out the loop on this: I filed additional breakdown bugs on the
crash reporting stuff. See
https://bugzilla.mozilla.org/showdependencytree.cgi?id=1009760&hide_resolved=1
I have taken bug 1039575/bug 1039577/bug 1039579 in order to get a crash
dump ID to GMPParent::ActorDestroy and properly annotate the crash. This
turned out to be harder than I expected because GMPParent isn't a
main-thread object and various crashreporter bits were making thread
assumptions. Those patches are now up in Ted's review queue.
I would like the webrtc team to take care of bug 1038961. I really don't
know the object lifetime/hierarchy for all of the webrtc code. I see
that it's currently assigned to Jesup.
I also filed related bug 1039572 about GMPParent::ActorDestroy
re-entering itself, which I intend to fix soon, but it doesn't assert
and probably doesn't block any of the other work.
--BDS
_______________________________________________
dev-media mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-media
