Backed out 4 changesets (bug 1210852 (
), bug 1103188 (
), bug 1070216 (
)) for various crashes with mediamanager.cpp assertions a=backout
Backed out changeset 3705005d2190 (bug 1070216 (
))
Backed out changeset c458ad434a12 (bug 1210852 (
))
Backed out changeset 1babaebeccc6 (bug 1103188 (
))
Backed out changeset b199ea9f265c (bug 1103188 (
))
Bug 1070216 (
) - Properly manage lifetime of allocated CaptureDevices. r=jib a=lizzard
We currently avoid Deallocating a CaptureDevice used for multiple gUMStreams when one of them calls Deallocate() by keeping track of how many called Start().
The normal lifetime sequence however, is:
Allocate()
Start()
Stop()
Deallocate()
This patches fixes the lifetime management by keeping track of how many users of the CaptureDevice called Allocate().
Bug 1210852 (
) - do SelectSettings of device capabilities on media thread. a=lizzard
Bug 1103188 (
) - Keep track of stopped tracks in gUM stream listener. r=jib a=lizzard
This is needed to avoid something like:
[old stream] stop track 1 -> deallocate MediaDevice for track 1
[new stream] gUM() -> allocate MediaDevice for track 1
[old stream] stop stream -> deallocate MediaDevice for track 1
[new stream] gUM() -> start MediaDevice for track 1 (oops, MediaDevice was no more!)
Bug 1103188 (
) - Keep track of capture stop only in gUM stream listener. r=jib a=lizzard
Bug 1199131 (
) - Crash when click on a SELECT wich has at least 1 OPTION with background-attachment:fixed and a background. r=mstange, a=lizzard
Bug 1156238 (
). At minimum always use the root reference frame as the "stop at" ancestor when finding the animated geometry root.
Bug 1156238 (
). Skip setting async scroll clips if we aren't painting to the window because they are useless then. r=mstange
Displayports only get acted upon when painting to the window, and the async scroll clips only get computed when we use a displayport.
In addition we change an assert because if we are painting to the window then our root reference frame is either a root frame, or a popup frame. In either case we should not be able to get to out of flows outside of the frame subtree rooted at the root reference frame by following placeholders.
Bug 1219134 (
): P4. Properly assign the decoded sample duration. r=edwin a=sylvestre
We default to the previous logic if for some unlikely condition we couldn't find the frame's duration (using the last input frame's duration)
Bug 1219134 (
): P3. Properly set the keyframe flag. r=edwin a=sylvestre
We were incorrectly setting the keyframe flag of the last input sample rather than of the decoded one.
Bug 1219134 (
): P1. Fix pts calculation in FFmpeg video decoder. r=edwin a=sylvestre
This is the primary reason why we got no pts returned (pts were set to 0) when using early version of LibAV. Apparently you are expected to set the pts when allocating the buffer of a frame.
This is undocumented, but both LibAV and FFmpeg do so internally. So do the same.
Bug 1220037 (
) - test convolution. r=padenot, a=sylvestre
This is in the mochitest suite so that Android and B2G tests can run it, but designed so that it can be moved to web-platform-tests when they run on all platforms.
Bug 1220037 (
) - pack and unpack Nyquist for MOZ_LIBAV_FFT. r=padenot, a=sylvestre
BufferComplexMultiply knows nothing about this format and so ends up corrupting the DC coefficient if packed Nyquists are multiplied.
Bug 1211708 (
) - Allow themes to specify XBL bindings even in unprivileged documents. r=sicking, a=sylvestre
Bug 1219833 (
) - Respect composition underline color; r=masayuki, a=sylvestre
nsTextFrame didn't use the composition underline color if the composition didn't have a foreground color defined. This patch makes it use the underline color if foreground color is not defined.
Bug 1219833 (
) - Respect composition underline color; r=masayuki, a=sylvestre
nsTextFrame didn't use the composition underline color if the composition didn't have a foreground color defined. This patch makes it use the underline color if foreground color is not defined.