The adjusted margin-end needs to be applied to the volume controls when the fullscreen button is hidden due to fullscreen being unavailable, such as within an iframe that is lacking the allowfullscreen attribute. Previously the margin-end was only applied when the video was determined to be audio-only.
acorn is now at db59bd0296c16796a07e82d2dc284861f464603e (untagged commit after
0.11.0 release).
pretty-fast is now at abc8c1133cc5443fb3bb50c426731821d27300a6 (tagged 0.2.0).
test_cancel_restart needs the progress notification to show reliably so it can
cancel the pending install. To ensure this a sjs script is used to
asynchronously deliever the XPI data. It starts responding but sends no data
until a second request is made to tell it to complete. So the test can start
the install, then do what it likes with the progress dialog before finally
telling the server to complete the download for the install.
--HG--
extra : rebase_source : 9bd5c617a28ee4edaa8e18599ad8eca0b52f7133
This test includes a race condition between showing the install progress
doorhanger and the XPI download completing. When the download completes before
the popup finishes showing (there is some animating going on there) the popup
is hidden and never shows properly, the test promptly breaks. Something about
the e10s support changes made this far more frequent.
This change makes us only listen for the popupshowing event for progress
notifications and then move on to wait for the following notification or dialog
that we care about after that.
test_cancel_restart is the exception that actually needs to interact with the
progress dialog so it is disabled here and will be fixed in a separate change.
--HG--
extra : rebase_source : 2a1d8c1b89a88275e37f6ecb6e13afeac0e98f8f
In some cases removing an element from a document doesn't call its XBL
constructor. tabbrowser knows this and so calls a destroy method on browser
elements when removing them.
When remote-browser was added we forgot to add a specific destroy method so the
base browser binding's would be used. This would mean remote-browsers aren't
being destroyed correctly.
Also when we switch from remote to non-remote browsers or vice versa we don't
call the destroy method at all so sometimes the browser isn't destroyed properly
and observer notifications get fired on dead browser elements.
--HG--
extra : rebase_source : d54ca85d912fdd736495c500910ff51fb5e15b6a