| Commit message (Collapse) | Author | Age | Files | Lines |
|\ |
|
| |
| |
| |
| |
| | |
bug:6829376
This reverts commit 0453f79675ffe6fddb59add2dcee0e3f0e7e74b1
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Clear will only happen when the player is stopped and we are moving away from
the current page.
bug:6520904
Change-Id: I8fd872eefdc012aa819280499298e50186a9a20f
|
| | |
| | |
| | |
| | | |
Change-Id: I1d9c361cff6d8b5ff5f1cd198ca97740fdcc2954
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | | |
Change-Id: I71e41f1a76ac3e8e845d9636d57cf3896bb6ec0d
|
|\ \ \ \ |
|
| | | | |
| | | | |
| | | | |
| | | | | |
Change-Id: I8ef6737ef4798d0a0b9100ea60e3a04b8198661c
|
|\ \ \ \ \
| |/ / / /
|/| | | /
| | |_|/
| |/| | |
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
A change in collection, visible rect or scale will cause a re-prepare in invoke
mode. Always re-prepare in draw mode to keep updating tile draw counts - this
way a scrolling webview doesn't steal tiles unjustly from a non-scrolling
webview.
Change-Id: Idf9f88afe36b582a3aa0369cfb76d8e47e5d3da6
|
|\ \ \
| |_|/
|/| | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The previous text bound estimates were incorrect for vertical text, empty
for single-glyph runs, and generally way too large. This change addresses
all of the above issues.
Change-Id: I5f3740a96f0532609f57347299ccb1adf78dc391
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: If46f35856279ab2a356cb7f99671fcd15f4ddf37
(cherry picked from commit 19d56fe70a9ea33c3ce2ad4f0b547e35453aeafd)
|
| | |
| | |
| | |
| | | |
Change-Id: I13145c96ae5911c5ebe6b2006c53f1a05ffddc1e
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| | |
mallinfo will block any malloc/frees while it is running, and
unfortunately it is rather slow - it takes upwards of 70ms. Fortunately,
we don't need all the info mallinfo returns, and can instead query
for the total usage directly. This call merely returns a global value,
so the caching mechanism that is in place is no longer necessary.
Change-Id: Ida3af9202a6825b19b8590d4b40ca72ba2230c90
|
| |
| |
| |
| |
| |
| |
| |
| | |
Make sure to take into account the current matrix to calculate the right
global bounds
Tag operations with the matrix that affected their drawing
Change-Id: Iadda9f3849dc3df1507e63423ff8d3ad0f37d74e
|
|\ \ |
|
| | |
| | |
| | |
| | | |
Change-Id: I7a081b0a99574697a2ea77f85693d7fa65b80743
|
|\ \ \
| |/ /
|/| | |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently, SkPaint::measureText() is very, very slow. Fortunately, we
do not need precise text bounds and can live with approximated bounds
so long as the approximation is not smaller than the glyphs. We err
very much on the side of safety so as to play nicely with calligraphic
fonts and scripts that make heavy use of modifiers.
Change-Id: Id08f0856c0d902386089560fc3cfcdd853a06177
|
|/
|
|
| |
Change-Id: Iaf91d46b448ae49fe2359cfd19694995edec95e5
|
|
|
|
| |
Change-Id: I1380ae53139d5f50a25ea5edb61ec8b6818112ca
|
|\ |
|
| |
| |
| |
| |
| |
| |
| | |
This just keep the state in sync with java side, even though native side didn't
rely on such states.
Change-Id: Ia73b4d09eed7282cdf1687d9ba6e74788749f738
|
|\ \ |
|
| |/
| |
| |
| | |
Change-Id: I967c8e2ea209c848e6a09b44390ab2228b213ca5
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Bug: 6763306
The prior approach queries and deletes only the resources that are
given in the manifest file. There are two problems it seems:
1. It is expensive & complicated. The deletealldata is called by an API
method which clearly states this will clear all javascript storage APIs.
there is an webstorage API for this, why make it complicated?
2. It seems a bug somewhere actually prevents cleaning disk storage.
(see the bug above for experiments). This can be a bug of the old approach,
or this can be a webkit bug (less likely).
Change-Id: I7cf984efce5117191ca0e691007e2e0a11d75e93
|
|\ \ \
| |_|/
|/| | |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Bug: 5869022
When appcache reached max size, a callback is triggered. The
second parameter of this callback was returning the used Web SQL
quota, which is a rather different concept then AppCache. Fix this
to return max database size.
Change-Id: I5ece83934d5a43b6927bc8fc25bcae1354c49149
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Makes more sense to have PlatformGraphicsContextRecording
do the optimizations as it already has all of the interesting data
rather than have GraphicsOperationCollection "deep inspect" the
stream
Change-Id: I0dba6531a352d178afbee2c5eea7d67b82c28df5
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
WebKit will sometimes produce graphics context calls that result in
GC::Save, GC::SetClip, GC::Restore - as they don't do any drawing, we
can detect and filter those out which saves on the amount of state
management done.
Change-Id: I61715b0a14adb39521c605e2562fc4ac4e70a327
|
| |
| |
| |
| | |
Change-Id: I086a3ff7b6191bf5e20f8b5f6f74bcc7d2e867c2
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Basically, player is responsible to enter fullscreen mode such that we have
the proper proxy updated on the Java side.
bug:5710646
framework change:
https://android-git.corp.google.com/g/#/c/202139/
Change-Id: Ie7adf620f2e9889142ef153e12d676a5f288bc0a
|
|\ \ |
|
| | |
| | |
| | |
| | | |
Change-Id: I39a170ce120e4ce5237e1cfc2b0e94cdb42c6124
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Swap the SkBitmap instead of a copy to save the memory bandwidth.
Change-Id: I0998248a9a87ce4058898e66a4739995da9c003c
|
|\ \ \ \
| |/ / /
|/| | |
| | | | |
* commit 'ceeb43093b8e721dfe6671d7da44c1d1fd6f6ed8':
|
| |\ \ \
| | | | |
| | | | |
| | | | | |
Change-Id: I4bc0b90d3078b9eaebe8fc08eda767767dc68dd9
|
| | |\ \ \
| | |/ / /
| |/| | |
| | | | | |
Change-Id: I7c9f8c596ec815bc110922acccbb346aa7cbb2bf
|
| |_|/ /
|/| | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Don't record state changes that didn't actually change state
Due to GraphicsContextAndroid::syncPlatformContext these 4 states
are set before every draw call. This is cheap for PlatformContextSkia,
as it detects that nothing has changed and so doesn't do anything. However,
PlatformGraphicsContextRecording would always record these same 4 commands
over and over again, which isn't efficient
Change-Id: I386bcb91c7851824ed01de1856ce6f6a358f8eb4
|
| | | |
| | | |
| | | |
| | | | |
Change-Id: Ifc514b1cfc12298c65a6c553c749a7b35520de7b
|
|\ \ \ \
| |/ / /
| | | |
| | | |
| | | | |
* commit 'ff80940ae8358f49efbbb2d989b02e4b23dd1930':
Accept-Language header is not updated when phone language changes
|
| |\ \ \
| | |/ /
| | | |
| | | | |
Change-Id: I9b3379ec26c3a0fb208b8cdd66a0b60088c1b01d
|
| | |\ \ |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The Java class WebSettings has accessor methods for User Agent
and Accept-Language strings. These methods ensure that the values
are updated if the Locale has changed. However, WebSettings.cpp
in WebKit reads the private fields in WebSettings.java directly,
and thereby fails to trigger the update.
The correction is to use the appropriate accessor methods from
WebSettings.cpp.
Change-Id: Id1aa1f4de38d2a73cad9660afc35944c663dbd4a
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Performance seems comparable to SkPicture already, and this
makes it easier to iterate on fixing some of the problem areas
Change-Id: I2e7b6b282e6b53ba2dc68d8a62b3b982f2b5838d
|
|\ \ \ \ \ |
|