Граф коммитов

1228 Коммитов

Автор SHA1 Сообщение Дата
Matt Woodrow 7182f1554c Bug 1109431 - Apply TimestampsFuzzyEqual rounding to the buffered ranges returned from multiple decoders. r=ajones
--HG--
extra : rebase_source : 1ca8f7910db619b90fdc9478bb0dc52a5d16693b
2014-12-11 10:52:57 +13:00
Matt Woodrow c8cf5fc8d0 Bug 1108917 - Part 2: Allow eviction of entire decoders that are ahead of the current play position. r=ajones
--HG--
extra : rebase_source : fc7fbd06e9c0c65de1c5ecb07a7a355fe1f6fbe8
2014-12-11 10:50:41 +13:00
Matt Woodrow 53e63e0a0c Bug 1108917 - Part 1: Attempt to evict data from decoders in start-time order. r=ajones
--HG--
extra : rebase_source : c906ca08b4a984a7b3e2ef1cdcaea810b5f5777d
2014-12-11 10:49:54 +13:00
Bobby Holley 65fb74f0e7 Bug 1108701 - Replace AudioDecodeRendezvous with promise-y goodness. r=cpearce,karlt 2014-12-10 14:03:56 -08:00
Bobby Holley 60c0158fdd Bug 1108701 - Use MediaPromises for RequestAudioData and RequestVideoData. r=cpearce 2014-12-10 14:03:56 -08:00
Bob Owen d921ff4edf Bug 1103946 Part 1: Change --content-sandbox mach / mochitest option to --strict-content-sandbox. r=jmaher
Also, change tests that are skipped using (contentSandbox != off) to (strictContentSandbox).
2014-12-10 09:34:03 +00:00
Chris Pearce e859dcdb44 Bug 1109389 - Don't share task queues with all EMEDecryptors the EME PDM creates. r=kinetik 2014-12-09 20:34:00 +01:00
Jean-Yves Avenard 3b9c59eb6d Bug 1107339 - Add support for MP4 with VP6F video track (FFmpeg only). r=k17e 2014-12-04 11:58:00 -08:00
Karl Tomlinson a5bc939348 bug 1108846 notify MediaSourceDecoder of bytes downloaded to trigger progress events r=kinetik
--HG--
extra : rebase_source : d13f3a9c46cce745cffc86a179875f34e7a52428
2014-12-10 10:45:06 +13:00
Karl Tomlinson d95ab8bc85 bug 1108838 dispatch "stalled" even when no bytes have been received r=cpearce
This is important for MediaSource, where there is no initial request to set up
the stall counter by sending an initial progress event.

For sources using ChannelMediaResource, this means that stalled can now fire
before an HTTP response is received.

Also reset stalled timer on transitions to NETWORK_LOADING,
and don't run the progress timer while stalled.

--HG--
extra : rebase_source : cf2cf8a4de37ca4859761941946393e9747c0706
2014-11-28 17:07:15 +13:00
Karl Tomlinson d21db9fc16 bug 1108838 move stalled/progress timing from MediaDecoder to HTMLMediaElement r=cpearce
This provides that mNetworkState is available for determining whether progress
notification from the resource/MediaCache after stalled should resume progress
events.  The timer will be stopped while stalled in a subsequent patch, and
should not be restarted unless NETWORK_LOADING.

--HG--
extra : rebase_source : e335555de404f6a899762be3c05b8f5444c357e3
2014-11-28 12:17:18 +13:00
Karl Tomlinson 42023d1511 bug 1108787 use NowLoRes() for progress event timing r=cpearce
Precision is not required, and is more costly on some platforms.

--HG--
extra : rebase_source : de36a0d249a89e579a7433f40c89b339f7db5b49
2014-11-21 15:45:29 +13:00
Karl Tomlinson 4834bbd250 bug 1108787 dispatch progress even when timer runs late r=cpearce
avoiding skipping progress events when data is received only soon after the
previous progress event.

--HG--
extra : rebase_source : 5108ae7d91bac613ed67f85c0963c6ca020bee07
2014-11-21 17:37:44 +13:00
Bobby Holley 312651fc98 Bug 1108707 - Make reader shutdown asynchronous. r=cpearce (relanding on a CLOSED TREE because it wasn't the source of the leaks) 2014-12-09 11:43:21 -08:00
Ryan VanderMeulen 073573c2a9 Backed out changeset 18c302f80bce (bug 1108707) for nsTArray_base leaks.
CLOSED TREE
2014-12-09 15:49:38 -05:00
Chris Pearce b392b7de2a Bug 1077159 - Use macros for GMP API names, to make them implicitly versionable. r=jesup 2014-12-10 09:35:26 +13:00
Bobby Holley fc5e0383a6 Bug 1108707 - Make reader shutdown asynchronous. r=cpearce 2014-12-09 11:43:21 -08:00
Paul Adenot 1a9942004c Bug 1065924 - Make sure we correctly update the variables tracking the number of samples written. r=jesup
--HG--
extra : rebase_source : 0f42c4a6d612a98f56edf93811cccca6d210795e
2014-12-09 14:11:32 +01:00
Byron Campen [:bwc] 9f12762413 Bug 1091242 - Part 7: Wiring the build system together. r=jesup, r=ted 2014-11-17 17:03:28 -08:00
Byron Campen [:bwc] 3af28ea884 Bug 1091242 - Part 6: Wiring the new JSEP handler code in. See https://github.com/unicorn-wg/gecko-dev/tree/multistream_rebase for more history. r=jesup, r=smaug 2014-11-19 16:16:29 -08:00
Byron Campen [:bwc] 0019770636 Bug 1091242 - Part 3: Mochitest work. See https://github.com/unicorn-wg/gecko-dev/tree/multistream_rebase for more history. r=jib 2014-11-19 16:18:10 -08:00
Ehsan Akhgari d60d0c93f5 Bug 1108687 - Fix the MOZ_STACK_CLASS annotation on DecodersToInitialize; r=jyavenard
The attribute syntax should appear after class.

--HG--
extra : rebase_source : 2793470bf5bcd8b701729906987ce9958016b4e7
2014-12-08 23:24:55 -05:00
Phil Ringnalda 5fe54ea0fd Backed out 10 changesets (bug 1091242) for Android/b2g non-unified build bustage
CLOSED TREE

Backed out changeset 7f72b55c5de7 (bug 1091242)
Backed out changeset f1501aa24397 (bug 1091242)
Backed out changeset 7fde5994aee5 (bug 1091242)
Backed out changeset 59b415714087 (bug 1091242)
Backed out changeset dadb65fedc08 (bug 1091242)
Backed out changeset 21be81424e4e (bug 1091242)
Backed out changeset 498fb1dafba5 (bug 1091242)
Backed out changeset 8d0653eb85ab (bug 1091242)
Backed out changeset c82d484e135a (bug 1091242)
Backed out changeset 3e0c8932f1b1 (bug 1091242)
2014-12-08 20:53:07 -08:00
Chris Peterson e424225899 Bug 1107814 - Part 2: Mark some directories as FAIL_ON_WARNINGS conditionally for some compilers. r=gps 2014-12-04 01:11:14 -08:00
Chris Peterson 22d0c4b343 Bug 1107814 - Part 1: Mark more directories as FAIL_ON_WARNINGS for all compilers. r=gps 2014-12-04 16:24:03 -08:00
Bobby Holley 9ef2bbc703 Bug 1108767 - Add the ability to chain same-type promises. r=cpearce 2014-12-08 17:19:05 -08:00
Bobby Holley a143dac86d Bug 1108767 - Allow promises to be resolved without a MediaPromiseHolder. r=cpearce
It turns out that there are use-cases where this restriction is clunky.
2014-12-08 17:19:05 -08:00
Bobby Holley dc54f0958e Bug 1108767 - Add MediaPromise::{Resolve,Reject}IfExists. r=cpearce 2014-12-08 17:19:05 -08:00
Bobby Holley 0f8a47ff4f Bug 1108767 - Stop including MediaTaskQueue.h from MediaPromise.h. r=cpearce
We need this so that MediaTaskQueue can use promises.
2014-12-08 17:19:05 -08:00
Bobby Holley 2a3ae3c23c Bug 1108767 - Separate ThenValueRunnable so that the constructors don't collide when the resolve value is the same as the reject value. r=cpearce 2014-12-08 17:19:05 -08:00
Wes Kocher 126b4f108e Backed out changeset bf25101e66cf (bug 1095098) for build bustage 2014-12-08 16:27:12 -08:00
Denis Volk 0f94bddfc8 Bug 1095098: move do_QueryObject templates into their own header r=froydnj 2014-11-20 12:20:10 +01:00
Bobby Holley 138a71a52a Bug 1097823 - Implement Media Promises. r=cpearce 2014-12-08 14:45:37 -08:00
Bobby Holley 143aa4f9db Bug 1097823 - Implement MediaTaskQueue::ForceDispatch. r=cpearce
This is necessary to have strong guarantees that promises will be resolved.
While we're flushing the task queue, normal dispatch starts to fail,
meaning that we can't dispatch promise resolution. We have 3 options to handle
this:

(A) Never respond to the promise.
(B) Invoke the Resolve/Reject callback synchronously if dispatch fails.
(C) Prevent dispatch from failing.

(C) seems like the option least likely to violate invariants if we can get away
with it. Promise resolution is unlikely to be a heavyweight task in the way that
a decode task might be, so this should hopefully be ok.

Note that this still doesn't help for bonafide task queue shutdown. It's up to
consumers to tear down their MediaPromiseHolders before the task queues are shut
down.
2014-12-08 14:45:37 -08:00
Bobby Holley 8c9d29f7fe Bug 1097823 - Move NotDecodedReason from RequestSampleCallback to MediaDecoderReader. r=cpearce 2014-12-08 14:45:36 -08:00
Byron Campen [:bwc] c29c3be2eb Bug 1091242 - Part 7: Wiring the build system together. r=jesup, r=ted 2014-11-17 17:03:28 -08:00
Byron Campen [:bwc] bb404e8254 Bug 1091242 - Part 6: Wiring the new JSEP handler code in. See https://github.com/unicorn-wg/gecko-dev/tree/multistream_rebase for more history. r=jesup, r=smaug 2014-11-19 16:16:29 -08:00
Byron Campen [:bwc] 711f732dd9 Bug 1091242 - Part 3: Mochitest work. See https://github.com/unicorn-wg/gecko-dev/tree/multistream_rebase for more history. r=jib 2014-11-19 16:18:10 -08:00
Ryan VanderMeulen 529edd40b5 Merge inbound to m-c. a=merge 2014-12-08 15:46:14 -05:00
Ryan VanderMeulen 0aae88bf85 Backed out changesets ae63e2dda6e0, a9ca0113db3c, and d0a6757e24bd (bug 1097823) for bustage.
DONTBUILD CLOSED TREE
2014-12-08 15:00:02 -05:00
Bobby Holley 1a45088976 Bug 1097823 - Implement Media Promises. r=cpearce 2014-12-08 11:34:50 -08:00
Bobby Holley 881af4ca94 Bug 1097823 - Implement MediaTaskQueue::ForceDispatch. r=cpearce
This is necessary to have strong guarantees that promises will be resolved.
While we're flushing the task queue, normal dispatch starts to fail,
meaning that we can't dispatch promise resolution. We have 3 options to handle
this:

(A) Never respond to the promise.
(B) Invoke the Resolve/Reject callback synchronously if dispatch fails.
(C) Prevent dispatch from failing.

(C) seems like the option least likely to violate invariants if we can get away
with it. Promise resolution is unlikely to be a heavyweight task in the way that
a decode task might be, so this should hopefully be ok.

Note that this still doesn't help for bonafide task queue shutdown. It's up to
consumers to tear down their MediaPromiseHolders before the task queues are shut
down.
2014-12-08 11:34:50 -08:00
Bobby Holley 7aa1467674 Bug 1097823 - Move NotDecodedReason from RequestSampleCallback to MediaDecoderReader. r=cpearce 2014-12-08 11:34:50 -08:00
Jay Wang 56bf9455a1 Bug 1105452 - Need to use new Audio system APIs for audio offload playback. r=roc, r=jld, r=ggrisco
Resolve the build failure caused by API changes

There are some changes in Audio APIs in Android version
21. Modifying the code to use the new APIs.

Change-Id: I24fdeb20f8f957d05fb6c0c317de0a6f0769c347

Resolve seccomp violation caused by syscall 256

Modify the filter to allow syscall 256 (set_tid_address).

Change-Id: I49461770c4c5e70bf68462d34321381b0b7ead0a
2014-12-02 17:10:00 -05:00
Andreas Pehrson 27ed59b426 Bug 1106963 - Resync media stream clock before destroying decoded stream. r=roc 2014-12-04 19:29:00 +01:00
Jan-Ivar Bruaroey afad2eeeee Bug 1053407 - return DOMError rather than string in peerConnection callbacks. r=bz, r=jesup 2014-11-26 15:26:37 -05:00
Bobby Holley aa27b00ccd Backed out changeset c405b4271e64 (bug 1097823) 2014-12-07 17:09:12 -08:00
Bobby Holley 4867c99ce8 Backed out changeset 8dfbe042e946 (bug 1097823) 2014-12-07 17:09:11 -08:00
Bobby Holley b55769853e Backed out changeset b56e5e762329 (bug 1097823) 2014-12-07 17:09:09 -08:00
Bobby Holley 8327697375 Backed out changeset b76636e9f49a (bug 1097823) 2014-12-07 17:09:08 -08:00