| Commit message (Collapse) | Author | Age | Files | Lines |
|\
| |
| |
| |
| | |
* commit '538b01d6410e7c7a5b2faabe7b84c80ddc32d5f3':
Cherry-pick WebKit r100677 to fix a rendering crash
|
| |
| |
| |
| |
| |
| |
| |
| | |
This fixes a crash from positioned generated content under run-in.
See http://trac.webkit.org/changeset/100677.
Bug: 6079158
Change-Id: I3d2012c58f47e71ae500e33551dfab5587b84534
|
|\ \
| |/
|/|
| |
| | |
* commit '2857538e8cd8c752b06ee75dd039791a186455b3':
Skia API changes as a result of an update to the Skia library.
|
| |
| |
| |
| |
| |
| | |
These changes are required to work with r3022 of Skia
Change-Id: I4d4a13bad9acc54934bbb46bdfc08ee424c7a33a
|
| |
| |
| |
| |
| | |
bug:5820635
Change-Id: I24b6eae6c76c29e44106b4ec87e74d945aad7b1b
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
bug:5759299
Webkit invals aren't correct when the content size changes (even simply growing
vertically), for now just repaint from scratch in that case.
Change-Id: I408b289bfee1139d6c8f3c7babd963d217cb5bc0
|
|\ \ |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Cherry pick Ic6b6860ca4bfe891e84fb0445d67c7ea728c6fe3
The HTML5 History API is disabled at runtime by default in WebKit.
Configure it to be enabled.
Bug: 5767904
Change-Id: Ib7374e0145c581588990baec6e67025c1b7ef693
|
|\ \ |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Cherry-pick from master
When we have composited layers inside iframes/frames, the layers
hierarchy is not always up to date at the time of the layerSync()
call. If some of those layers are scheduled to be repainted, the
repaint operation will triggers the update of the composited layers
tree -- possibly resulting in the deallocation of the very same
GraphicsLayer we were painting from, and thus leading to a crash.
The fix consist in gathering all the root RenderLayer (for each
frame containing composited layers) and explicitely asking
RenderLayerCompositor to check if the composited tree needs to be
updated, before we traverse the tree to paint the elements.
bug:5695185
Change-Id: I33a00b847eb19c9aa4b68f0ac3adbe36709ed00b
|
|\ \
| | |
| | |
| | | |
ics-mr1
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| | |
cherry-pick of https://android-git.corp.google.com/g/#/c/155078/ into MR1 branch
bug:5660814
We want to avoid swapping in and out of singlesurface mode when we have enough
textures to hold the content, but not while the content is updating.
Change-Id: I3c45f2b9f3c8e06b397b701fb53ac6ac5ff8efdd
|
|\ \ |
|
| |/
| |
| |
| |
| |
| | |
cherry-pick of https://android-git.corp.google.com/g/#/c/155121/2 into MR1 branch
bug:5757950
Change-Id: I34042e8b1f1adf1fc1826d7fde5500708bef3b60
|
|/
|
|
|
|
|
| |
With this partial revert of ececd484e5832707be4a73b70da37f862eb14c35 , single core devices can't have tile generation starved by javascript/plugins.
bug:5746432
Change-Id: I42b5799f4ba136e1c6f9812219824d21390bace2
|
|\ |
|
| |
| |
| |
| | |
Bug: 5712065
This reverts commit 35e0b8dd7902e92f50ede4df2e4bad4ed581b770
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
In some cases we have to deal with very large layers (e.g. 130k x 56k).
We do clip them at draw time, but at prepare time we will generate *all* the
needed BaseTile objects. Those are small, but when you have to allocated
100k of them and then iterate, it does not make things fast, and we can
ANR (and sometimes recover later).
bug:5466840
Change-Id: I01c64ed1014fa719b619609fd2bd24126dc9056e
|
|\ \ \ |
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
bug:5726709
Framework invals are not valid when zooming completes because much content
outside the rect will have changed from zooming.
Change-Id: I72e4509e13bf97ea90f09716699e66c1c3f9acd8
|
|\ \ \
| |/ /
|/| | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Bug: 4586251
Explicitly manage ref counter for WebUrlLoaderClient. This will
prevent early release of the object.
Change-Id: I4b8e41b470eaacb594ce85afdd650e7827ab4a6d
|
|\ \ \
| | | |
| | | |
| | | | |
ics-mr1
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Bug: 5455817 (Text reflow not working in browser)
Change-Id: I2a62a7fb4c7acde640cd87b8d0f03200b8f7fc2f
|
|\ \ \ \ |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
We only keep the scrollable layers invalidating their entire area
instead of incurring this penalty for *all* layers.
Also remove unnecessary repaint when scrolling (we have the entire
content on the UI side already).
While the entire scrollable area will be invalidated and marked
as dirty, the existing code in PaintedSurface already only look at
the visible tiles of scrollable layers to consider the layer's content
as being ready to display, so the real world penalty (while far from
optimal) is limited.
Implementing the correct approach (only invalidating what really
changed on scrollable layers) would sadly be a lot more complex,
as currently webkit will *not* send us the repaint invals if they
are on a currently clipped area, as webkit's default behaviour
to implement scrolling of such element is to repaint them anyway...
bug:5721618
Change-Id: Ia470157c716fa1c557e4a196ba014296ad9e627a
|
|\ \ \ \ \ |
|
| |/ / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Bug: 5727901
Change-Id: Ic2323eca6bc6e77fe428f3602800799e7b996cce
|
|\ \ \ \ \
| |/ / / /
|/| | | | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
bug:5724482 bug:5660963
we were calling setViewport and setViewRect out of order, so we were clipping
with the last frame's (or last webview's, if >1 were visible) projection matrix
Change-Id: Ifcbfa0021cb33e3400ac9eaf64fc2235ee8afe50
|
|\ \ \ \ \
| |_|/ / /
|/| | | | |
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
bug:5704511
Dirty tiles use two textures - one for display, and one for painting.
This fixes an issue where the number of textures requested did not meet the
needs of a double buffered tree with most content invalidated. The webview would
give up drawing, and thus not paint the invalidated portions of the layers.
Change-Id: Icd9b9e0b53cf82415074dac4338a8c48de880364
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
bug:5720732
Change-Id: Iaf3a41adaa0799867a615b42d1d603aca3c8e35d
|
|\ \ \ \
| |/ / /
|/| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Bug: 5719053
Change-Id: I5f69e411d2ade2eaaf5c9e9e39dc8dc7d23425f1
|
|\ \ \ \
| | | | |
| | | | |
| | | | | |
ics-mr1
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Bug: 5699085
Previously, animations were copied along with layer trees from the webkit to UI
thread. This changes that to instead use a single object per animation with
refcounting so that animations are kept consistent.
Additionally, animations are now run on both the painting and drawing trees in
the tree manager so that animated content clipped correctly during a tree swap.
Change-Id: I79f0c0e47b717f9fdddf303eb7ec29efc4950aaf
|
|/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
bug:5713701
This fixes a potential deadlock between TextureGenerator and ImagesManager
Change-Id: I57503d3b73aff571629826f2278491232b37cb67
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This reverts commit 2d19ea7c0d192c0c1eda1caed7e25143bbe22c20.
bug:5689093
|
|\ \ \ \
| | | | |
| | | | |
| | | | | |
into ics-mr1
|
| |/ / /
| | | |
| | | |
| | | |
| | | | |
bug:5704428
Change-Id: I1d1d1f214642b6ff868e8425778565686bfe8b69
|
|/ / /
| | |
| | |
| | |
| | | |
bug:5699003
Change-Id: I4c35008f1324e0f9e83b7dd9e72e09a7e09ae7a9
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
bug:5699531
Partial layer inval isn't working yet, disabling for correctness.
Change-Id: I4838db41ea24e07f7b30c44ddb8321f598087d2a
|
|\ \ \ \
| |/ / /
|/| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
bug:5666027
Previously, only the most recent version of the scrollablelayerandroid would
have its position updated. This caused issues with position inconsistency
between the painting and drawing version of the layer, if both existed.
Change-Id: Ife29ae4e2cb00fbaa2f6fc95d9914b3434862f10
|
|\ \ \ \ |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
- asynchronous loading
- support for tiled content using TiledTexture
- fix repaint/duplication bugs
- share same textures as layers
- disambiguate between images by computing a CRC code
- added better debugging in TilesManager/ClassTracker to track
memory usage
bug:5661120 bug:5572134 bug:5521718
Change-Id: Id422fb991d6233bbe4dc6e5c3c7409468b7dca98
|