| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The problem was that when attempting to forcefully destroy a texture
in TilesManager::cleanupLayersTextures(), the call to
LayerAndroid::removeTexture() may fail if the texture was busy being
painted -- the call to BackedDoubleBufferedTexture::release() would
return false, and the layer may thus still keep a pointer to the texture.
But the release() call, while indicating it failed, was only delaying
the release -- as soon as the texture was marked as not busy, it
could set its owner to nil.
We could thus have a situation where the layer did not reset its
texture pointers because the owner of the texture was not yet
changed, but the texture would then reset its owner to nil as soon
as it was not busy painting.
In TilesManager::cleanupLayersTexture() the next step before deleting
a texture is to check that the texture does not have an owner -- but
by then the texture could have been marked as not busy, and removed
its owner, letting TilesManager destroying it.
bug:3472320
Change-Id: I3bcf169b30dfacba1773d3b79a3c0d205bf3cbdb
|
|
|
|
|
|
| |
bug:3461349
Change-Id: Id654d176c58027c67be7cb604b87c0ec68984525
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Using existing Skia APIs, FontCache::createFontPlatformData can detect
when the retrieved font is the Skia default and will reject that choice,
allowing WebCore to try the next font on the CSS fallback list. If the
requested family is a generic family, e.g. "serif" or "monotype", the
Skia default will be accepted.
A prior attempt at this fix was not properly handling bold and italic
stylings, which could eventually lead to dangling references to freed
font data used for layout while custom fonts were loading, causing
segfaults in getChild and elsewhere during page rendering.
Bug: 3469204
Bug: 2720133
Change-Id: I3a9e746e18fd27848d0c4b9e1232392df33254a6
|
| |
| |
| |
| |
| | |
issue: 4094300
Change-Id: If8e58daa621e01ac372f2111ebde522a6f90b2e5
|
|/
|
|
|
| |
bug: 4098273
Change-Id: Iaf7eaa13b9d4410dfb984af0e4a83dadedb09648
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We currently rely on the fact that nvidia serializes glFlush calls
to ensure synchronization, but we should be using fences to handle
this synchronization step.
bug: 4098273
Change-Id: I13c4790434dea07d9da68b8ebebbba15b1e085a5
|
|/
|
|
|
| |
bug:4079662 bug:3469243
Change-Id: I2538e33c97e3cf8a937bc310847298b68669a24f
|
|
|
|
|
|
|
|
|
| |
- correct support for the fullscreen button
- change the controls to be 48px high
- auto-hide the controls, touching the video makes them appear again
bug:2126902
Change-Id: Idd2b720034de3d5d432c9ea62d9045934c46f6c1
|
|
|
|
|
|
|
|
|
| |
The VideoLayer is ref counted, so unref at the destrutor is the right way.
Don't do a deletion.
bug: 4072630
Change-Id: I1da4ec138ef119b1d8dd7700fef36bab4df55064
|
|
|
|
|
|
|
|
|
| |
Increasing this size from 8 megs to 32 megs
allows more RAM to be used before
the image is sampled down.
bug:3421165
Change-Id: I73047009522d6e80d1adc95cd12d763307ab1980
|
|
|
|
|
|
|
|
| |
We were trying to update the texture with a bitmap larger than it.
GL did not like this, so the texture was not updated at all...
bug:4078436
Change-Id: I180f4da6991a3db988169d7abd06cc08d5b69a32
|
|
|
|
|
|
|
|
|
| |
Basically we provide a Video Layer for the inline video.
The java side change is at 101311.
bug:3506407,2126902
Change-Id: I9cc1b910f502595d6be7ac81e448343ca6ca9100
|
|
|
|
|
|
|
|
|
|
|
| |
- we were overriding existing animations.
- implement correctly the removeAnimations functions
- refactor AndroidAnimation (share common code)
- fix how we use the timing functions for the animations
- implements timing functions per keyframes
bug:2453890
Change-Id: I367d708338c270171eeaacc7e2fb3729eb78c444
|
|\
| |
| |
| | |
resetting the unusable state (set by the clearView() function) when receiving new content..." into honeycomb-mr1
|
| |
| |
| |
| |
| |
| |
| |
| | |
We were not resetting the unusable state (set by the clearView()
function) when receiving new content...
bug:4080072
Change-Id: Ie1110b511392b0a7fd21bbc23aa1b824eb8a78ad
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- we were replacing animations by new ones regardless of their types, so
when two anims (i.e. transform and opacity) where set on the same
layer, we'd only run the last one
- the selection of the keys for keyframes animations was buggy
bug:2453890
Change-Id: I03da3f6c2ba1f5bf778e099e52d71d2f5e67d27e
|
|\ \
| |/
|/| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Followed model used for transform animations. Tested with
http://www.webkit.org/blog-files/leaves/.
Fixes http://b/2453890.
Change-Id: Ie71217743b6e1ee9ab0e023fdff44e8c65c8d477
|
| |
| |
| |
| |
| | |
bug:3461349 bug:3464483
Change-Id: I627f06d0fe48aaa0adca65cd13dc738af87eeefc
|
|/
|
|
|
|
|
|
|
|
|
| |
Remove the notion of root layers.
We were computing the root layer status incorrectly for some child layers. If an
iframe has a layer but it hasn't been attached, we might assume it is the root
layer of the tree and not produce a recording content.
Bug: 3492471
Change-Id: Ib81fb26d76742d74ebe244c34c1fab353fb9b78d
|
|\
| |
| |
| | |
into honeycomb-mr1
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The initial attempt at fixing this bug was to modify the clip, but
this was not correct and only fixed cases where the plugin was on
the left edge of the screen. This CL properly adjusts the plugin's
boundaries to make sure the content is drawn in the appropriate
position.
bug: 3477581
Change-Id: I5958228a93fa5600cd7e24ed9e7e1e26f6b241bc
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Wait the remaining of the 60FPS cap delay rather than not paint.
Returning true if called faster than 60FPS means
we are not drawing and ask for the framework to call
us again; this works in general because the framework recopy
the previous framebuffer. But in some cases, it didn't, causing
the webview to flicker. A correct fix would be to introduce
the capping in framework rather than try to doing it in the
webview; in the meantime we will sleep the remaining of the delay
as a workaround, so that we still provide the GPU
benefits we wanted (at >60FPS the GPU was being saturated in
some cases).
bug:3500655
Change-Id: Ibaa1d93e0a13433a2c842b19b58538894fdaa7e4
|
|\ |
|
| |
| |
| |
| | |
Change-Id: I21555e923ba6ac6c2bcd291ffa73a1aa087bdb3b
|
|\ \
| |/ |
|
| |
| |
| |
| |
| | |
bug: 3477581
Change-Id: Ia7bbaaca405db33dbefaa8f6f00e9250580e5f7b
|
|\ \
| |/
|/| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The dtor of GLWebViewState is calling the dtor of its TiledPage
variables; the dtor of TiledPage wait until any pending painting
operations running in the TextureGenerator thread are done. The painting
operations ultimately paint using the current GLWebViewState's base
layer. We were doing unref() on the baselayer before destroying the
TiledPage...
bug:3429924
Change-Id: I685e7fca5e5bae796f808debbf7e4ce668b83e6b
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If we detect an OOM from GL, crash the process to stop the
framework falling back to software rendering. The framework will
not enable HW rendering again until the process is killed and
Flash does not work with software rendering.
Bug: 3477085
Change-Id: I8defebbe3ea281c342fc99ace8638c347a3eb246
|
|\ \ \
| |_|/
|/| | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This reverts commit 2879037f2d6c595b8c16e992462ce5ca60d8fdba.
Revert the original work around as the underlying problem is fixed
with I2747dd4630a2a7dc18eae4c5fde36fd14c461747
Change-Id: Ibc1e997bca2aadc27527b6980de38603e6971770
|
| | |
| | |
| | |
| | | |
Change-Id: Id6e0b8199f82c44de0e7b1b7e7eb42bdb96bf908
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Each view will now be capped at 60fps. The previous code would
either starve the drawing of additional webviews or would draw
them at a rate of 60fps/N where N is the number of webviews.
Change-Id: I5698d610b17d4c8cfe137c51c2b978dae04ae1cf
|
| |/
|/|
| |
| | |
Change-Id: I1b6be15a38d2afc55659872b82e13ebcd8fd2ada
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Ensure that we wait in a while loop checking the condition when
we call wait, and only signal after unlocking the Mutex.
Also removes mNewRequest lock and condition in the TexturesGenerator
as it doesn't seem to be needed.
Hopefully this will clear up some ANRs.
Change-Id: I1b8b103a6effd2aa678a304e2519a6f6c484aaad
|
|\ \ \
| |/ /
|/| | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Uses new skia api to identify fixed width fonts. Without this fix, all
Android fonts were reported as variable width, causing layout
adjustment of whitespace. Now that fixed width fonts are properly
reported, space characters preserve the same width as non-space chars.
Change-Id: I7d83481f44e6c34fd8ae5c6638b1625d3a83a637
|
|\ \ \
| | | |
| | | |
| | | | |
flash plugin..."
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
the GPU and the flash plugin...
bug:3471680
Change-Id: I436de6caea4dd5b1171f260d328dfe8170463832
|
|\ \ \ \
| |_|/ /
|/| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
bug: 3451982
Change-Id: If1d7bff729f72c28fec4905c3525b2f2ac1c6475
|
|\ \ \ \
| |/ / /
|/| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
For the example of google news, one of the fixed position layer is under
another fixed layer, and it is not updated for scrolling.
The CSS value did change, but we don't have a updated fixed position call.
Moving the update to sync time fix this issue.
bug:3486524
Change-Id: Iecb65b37d06e65902511e569238546b49b1dcd47
|
|\ \ \ \
| |/ / /
|/| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
bug: 3460361
Change-Id: Ifcf97b4360e0a86cfcc100e6287e1fbce8d2185a
|