When a new IME is attached it is not enough to remove the
WindowManager messages from the local queue, but the ones in
the previous IME queue must also be removed.
Fixes bug 8263462.
Change-Id: I9e916c6052a83dc7691bcba0b6ab8328b9b7cc36
A previous fix made initializing GL work better by calling eglMakeCurrent()
prior to querying the max texture size. However, that fix worked by creating an eglSurface
earlier than we did before, which for some reason causes problems later if wallpaper creation
fails and we back off to a software solution.
This new fix creates a temporary pbuffer surface instead, which still allows us to make the
call to eglMakeCurrent() prior to querying the max texture size, but does not result in the
later canvas lock failure if the wallpaper creation fails anyway.
Issue #8319960 sluggish yakju animations over launcher
Change-Id: I394d672549260a354f03ad9fd1b9e1f9a161a371
ImageWallpaper was sometimes querying GL for a max texture size too early
(before the first call to eglMakeCurrent()). This problem caused the wallpaper
to then get created in software, resulting in noticeably slower performance
when the wallpaper was visible.
This fix ensures that the
makeCurrent happens before the query, ensuring that wallpapers of the right
size actually get created instead of failing due to this error at creation time.
Issue #8319960 sluggish yakju animations over launcher
Change-Id: I12a3eba9f1818bdf544691e0727fe12f7e820651
new approach:
* for each app
* initial launch
* sleep 7.5s
* do 10 iterations: launch app with force stop + sleep in between
* report average of 10 launches
Change-Id: I9e68975325aa83af35620d727823f5c072aac488
In general, calling requestLayout() during layout is a Bad Idea.
However, ListView does this during the normal course of its layout, as it
removes and adds views during layout. However, it handles this properly,
ensuring that the views in its hierarchy are all measured and laid out
properly by the time its layout process is done.
A previous fix to the request-during-layout issue attempted to distinguish
the correct from incorrect behavior by checking whether views had been properly
laid out since the requestLayout() call, and making sure the views were
visible in the hierarchy (parented, attached, and !GONE), since otherwise
the views would not be laid out, the flags wouldn't be cleared, and requests
are superfluous anyway. However, this logic only checked whether the
requesting views were GONE, whereas the check should include the entire
parent hierarchy of the views (since a view with a GONE parent is still not
visible to the user).
This fix adds that additional check and cleans up other parts of the previous
code, such as not bothering to post() requests that occur during the second
layout pass unless those requests are also valid (coming from visible views).
Issue #8370042 Path seems to be in an infinite layout loop
Change-Id: I7aaf701229adfeee349a9a7c9ec14585735ba9f6
Show notification when a bugreport is finished, letting the user
launch a SEND_MULTIPLE intent to share them. Add dialog that warns
user about contents before sharing. Since bugreports are now stored
in private app data of the Shell app, use FileProvider to build Uris
that we can grant others access to.
Define BUGREPORT_FINISHED as being a protected broadcast. Delete
older bugreports automatically to reclaim disk space. Migrate any
Intent extras to ClipData when building PendingIntents.
Add --receiver-permission support to am shell command.
Bug: 7005318
Change-Id: If6c607dbcf137362d5887eac482ff7391563890f
The renderer used to pre-cache glyphs at record time. This then changed
to pre-caching at the beginning of every frame. This unfortunately entails
a lot of duplicate work on every frame, which amounts to 0.5 to 1ms in
some stock applications.
This change is somewhere in the middle: pre-caching happens the first
time a DrawTextOp is deferred or every time the screen-space transform
is different from the last pre-caching operation.
Change-Id: Id6d9e2599d90a5b75010b0f0a28746befbf3c205
If the IGraphicBufferProducers are the same, it's really the same
window, so we keep the same native object.
Bug: 8322406
Change-Id: I96d55a90e6656a05d4e32ae22945226c350b5747
Fix a few things found in our "Constants" section.
- Close unclosed links.
- Avoid periods inside parens for summary sentences.
- Lowercasing in a few places for consistency.
Change-Id: I9aa689fd980b373614dae7c4f8257e0786d2340a
Fix for failing android.speech.tts.cts.TextToSpeechServiceTest#testSynthesizeToFile.
In test env, ParcelFileDescriptor instance may be EXACTLY the same one that client uses.
And if it's closed by a client, then service is prevented from writing anything to the output.
Bug: 8377754
Change-Id: I7f95aae1b877e543ab02d3c548b29537aa852a89
This fixes a bug where keyguard was created without setting
the preferred orientation. The result is it would temporarily
assume the orientation of the device after being hidden.
The fix sets the flag when the keyguard view hierarchy is created
rather than when visibility changes.
Fixes bug 8124979
Change-Id: Ic54a185e4f51f79f2bb7fbc5ffc6992641e89132
Add a minimum size for the event pools for devices which don't have sensors.
Otherwise the system server crash loops on boot.
Change-Id: Ic51c6fc26c8779d9f435f358d4274148a2ddbfb3
