Several issues have been identified that will need to be
resolved carefully before we attempt to make similar changes.
1. Some applications make assumptions about the fact that all
Runnables posted with queueEvent() will be executed before
the next draw. Because the Choreographer may post messages
in the past (due at the frame time), these Runnables were
not always running at the expected time.
2. Some applications apparently hijack the GLThread and try
to run their own Looper on it. This obviously won't work
if we are already running one of our own.
We may need to add new rendering modes or target API checks
to provide an opportunity for applications to opt-in to the
changes in the future.
Bug: 8649804
Change-Id: I6289c3392a98cdbfaf28c7f4f7a8040f2138a3b4
It seems some applications rely on Looper.quit() terminating the
loop immediately without processing all messages. Rather than
risk breaking them, make the safer behavior optional.
Also take care to properly drain the message queue before quitting
so that all of the Message instances are recycled. This may
help release storage sooner in case the Looper doesn't get GC'd
promptly and its remaining queue of undelivered messages sticks
around.
Improve docs on runWithScissors.
Bug: 8596303
Change-Id: I8cbeb6f7a5f6b8e618b5109f87a03defc1486b9f
this could cause a dead-lock if the applicaltion's draw
implementation blocks until something happenson the main
ui thread.
note: we're still doing this synchronous draw in onWindowResize() because
that's what the previous implementation did. Technically, it has the
same problem.
Bug: 8586305
Change-Id: I782568289cc9419346aeea73775d86faa28b3058
GLSurfaceView will now automatically use the Choreographer to
schedule render updates on VSYNC. This happens only in
RENDERMODE_CONTINUOUSLY.
GLSurfaceView was rewriten to use a Looper and a HandlerThread
instead of a custom thread.
The basic mode of operation is unchanged, however the implementation
is largely different.
Removed support for older GPUs which had a limited number of contexts.
Bug: 8485730
Change-Id: I7b270f7ad7bb1259e858429054b1c8c9cc159dc6
Previously we could have returned from createSurface on the main thread
before calling eglCreateWindowSurface on the GLThread. That could lead
to a problem because the surface could be destroyed before
eglCreateWindowSurface got a chance to run.
Bug: 8328715
Change-Id: I273149f7d4b165abbe218a91ff54083f3f498cb0
Un-hide two new methods in MediaCodec, one new constant, and a new
EGL extension.
Bug 7991062
Bug 8191230
Change-Id: I028669132d9ffda1e4b34a561bab3997bbd7dae5
this adds correct versions of the broken GLES20 methods
glGetShaderSource, glGetActiveAttrib and glGetActiveUniform.
the old functions are still there and need to be @hide later.
Bug: 6006380
Change-Id: I8127a77c4b89aa8a9a54bea88774077535e2139d
When pausing we want to do three separate things, in order:
+ release the EGL surface
+ optionally release the EGL context
+ optionally terminate Egl
Previously we would only do these things if we had an EGL surface. But
it is possible that we don't have an EGL Surface, but still have an EGL
context. And in that situation we still want to release the EGL context.
Now we check the preconditions for the three cases separately.
Bug: 6338235
Change-Id: I804683b3d5c136cc98ea3f5051067eea18152ddf
Assume a failure in eglMakeCurrent occurs because the SurfaceView
surface has been destroyed.
See b/6257956 for an example of this failure happening during rotation
stress testing.
Change-Id: I4618703b5291aba3a3f0c6bd83c3435a67b97d33
A careful reading of the EGL spec, as well as experience with many
different EGL drivers, has shown that it is error prone to attempt to
discriminate between different error conditions.
We now treat any error besides EGL_CONTEXT_LOST as an indication that the
EGL context is in a bad state, most likely due to the window manager
having removed the underlying surface flinger surface.
In addition, we changed the way we deal with this kind of error:
Previously we would ignore the error and keep rendering. But if the
EGL context and surface has become invalid, it would be better to
stop drawing. We now stop drawing until the surface view surface is
recreated.
See b/6032663 for an example of this problem affecting the GMM app,
but note that GMM is using their own version of GLSurfaceView, so this
change won't help them directly. They'll have to make a similar change
to their version of GLSurfaceView.
Change-Id: Iffe3e1e3a3c7a91d03140fd34391eadeaecf777e
Signed-off-by: Jack Palevich <jackpal@google.com>
Packages can be enabled/disabled per user.
This requires maintaining stopped/launched states and
enabled / disabled components and packages per user.
Refactored pm.Settings and PackageSettingsBase to keep
track of states per user.
Migrated the stopped-packages.xml to users/<u>/package-restrictions.xml
Changed intent resolution to handle individual user restrictions.
Bunch of IPackageManager calls now have a userId argument.
Make AppWidgetService handle removals of packages.
Added some tests for pm.Settings and PackageManager.
Change-Id: Ia83b529e1df88dbcb3bd55ebfc952a6e9b20e861
This change allows Adreno GPUs to have multiple EGL contexts. We had
to limit this in earlier versions of Android due to limitations in
the Adreno GPU driver (only 8 EGL contexts allowed system wide.)
That brand of GPU has improved its EGL drivers to support multiple
EGL contexts in more recent versions of their drivers used on more
recent versions of Android.
Bug: 6142005
Change-Id: Id3030466be9a3d9fbe728f1785378c1f05da98fe
Certain EGL device drivers are generating this error, so we need to
survive it. (It's not clear from the EGL spec whether this is a
legitimate error code, but since it's being generated we need to
handle it.)
Remove logging of expected errors from elSwapBuffers. We expect these
errors to happen, and are handeling them by ignoring them.
No need to clutter the logs.
Bug: 6083241
Change-Id: If17dcfbb68d3009b92cf95a448728ccb47023b51
This patch adds a function setGlDebugLevel() to libEGL, and
exposes it to the Java layer at android.opengl.GLUtils.enableTracing().
Change-Id: Ia5abb130bc32fcfe3ab25b0a0a5283a54c54f357
This class provides an API for an application to know when it is time to
destroy its EGL context when memory is being trimmed. By having this in
the framework, we can still detect whether it will be useful to destroy
any EGL contexts (because we know if doing so will destroy all of them).
Change-Id: I1eac8d640052778052926b875c7928008f752182
Until now a leak was possible under the following scenario:
Create a GLSurfaceView
Register a renderer (this automatically starts a GLThread).
Discard the GLSurfaceView without installing it in the view system.
This scenario can occur when a device is rotated rapidly from
orientation A to orientation B to orientation C. In that scenario,
orientation B's GLSurfaceView might be discarded without ever being
attached to a window.
If this issue had been identified before GLSurfaceView had clients, one
possible fix would have been to delay the construction of the GLThread
until the GLSurfaceView was attached to a window. Unfortunately, it's
too late, and so making that change would lead to observable changes in
behavior, possibly breaking some clients.
Instead, fixed by making GLThread and EGLHelper static classes that hold
onto a weak reference to the GLSurfaceView. This allows the GLSurfaceView
to be garbage collected when it is no longer used, even if the GLThread
is active. GLSurfaceView's finalize method will manually stop the GLThread
if it is still running when the GLSurfaceView exits.
Part of this change was to remove the Renderer reference from GLThread,
because Renderer is a user-supplied class that could contain a reference
chain that points back to the GLSurfaceView.
Fixes b/5606613 "GLSurfaceView that's never added to a window will
leak threads and views, can leak activities"
Change-Id: Iafdc329eb6e9e40062358e7c119f5547ffe23d5e
This change adds Java definitions for the enums of the
GL_OES_EGL_image_external OpenGL ES extension.
Bug: 3482193
Change-Id: Ib50326f8be9b9cc9021753855c3846ddcdc5eaa2
* commit 'cb0db0306b5849a35d3d99eea1b34ce019c6f0d8':
Make the overridden ImageView#setVisibility remotable
Clamp non-monotonic stats instead of dropping.
DO NOT MERGE. Fix leak in LayoutTransition
Fix lastVisible/global rects
Fix Wimax-less build.
Fix leak in LayoutTransition
Deferring wallpaper update to improve workspace scrolling (issue 5506959)
Terminate EGL when an app goes in the background
boot animation is dithered and scaled
Fix NdefRecord byte-stream constructor.
PopupWindow dismiss() can get into a recursive loop.
Fold WiMAX state into the mobile RSSI.
Remove dedicated wimax icon to fix RSSI layout.
This does not happen on high end gfx devices. This happens
only if only one EGL context is initialized in the current
process.
Change-Id: Ibd1737efdf84eef8a84108b05795440d1ae9964e
