OSDN Git Service

android-x86/external-webkit.git
12 years agoMerge "Revert "Support "Vertical Writing Mode".""
Nicolas Roard [Wed, 12 Oct 2011 21:34:30 +0000 (14:34 -0700)]
Merge "Revert "Support "Vertical Writing Mode".""

12 years agoRevert "Support "Vertical Writing Mode"."
John Reck [Wed, 12 Oct 2011 21:08:26 +0000 (14:08 -0700)]
Revert "Support "Vertical Writing Mode"."

This reverts commit 5403113c5aa5d55a05ea001f562829e907632178

12 years agoam 5ed31e7b: Merge "Ensure the frame inversion flag is immediately effective." into...
Derek Sollenberger [Wed, 12 Oct 2011 20:27:59 +0000 (13:27 -0700)]
am 5ed31e7b: Merge "Ensure the frame inversion flag is immediately effective." into ics-mr0

* commit '5ed31e7bc52e99103e4da5755d944f9ca98305b3':
  Ensure the frame inversion flag is immediately effective.

12 years agoMerge "Ensure the frame inversion flag is immediately effective." into ics-mr0
Derek Sollenberger [Wed, 12 Oct 2011 20:26:49 +0000 (13:26 -0700)]
Merge "Ensure the frame inversion flag is immediately effective." into ics-mr0

12 years agoEnsure the frame inversion flag is immediately effective.
Derek Sollenberger [Wed, 12 Oct 2011 19:22:02 +0000 (15:22 -0400)]
Ensure the frame inversion flag is immediately effective.

Previously the frame inversion state was passed to the UI thread
when the layer tree was synced. This CL removes that restriction
and updates it immediately by storing the state in a cross thread
refcounted object.

bug: 5283034
Change-Id: I58a7ea0f2ed261008b945241bc783cfee4f33466

12 years agoam 4ac06026: Merge "Correctly guard WebFrame methods needed only with the Chromium...
Steve Block [Wed, 12 Oct 2011 19:00:46 +0000 (12:00 -0700)]
am 4ac06026: Merge "Correctly guard WebFrame methods needed only with the Chromium HTTP stack" into ics-mr0

* commit '4ac060268c9283c7c27657f97929fbe99712a14c':
  Correctly guard WebFrame methods needed only with the Chromium HTTP stack

12 years agoMerge "Fallback handle if vertical glyphs are missing."
claireho [Wed, 12 Oct 2011 18:11:14 +0000 (11:11 -0700)]
Merge "Fallback handle if vertical glyphs are missing."

12 years agoMerge "Correctly guard WebFrame methods needed only with the Chromium HTTP stack...
Steve Block [Wed, 12 Oct 2011 18:02:19 +0000 (11:02 -0700)]
Merge "Correctly guard WebFrame methods needed only with the Chromium HTTP stack" into ics-mr0

12 years agoCorrectly guard WebFrame methods needed only with the Chromium HTTP stack
Steve Block [Wed, 12 Oct 2011 17:44:42 +0000 (18:44 +0100)]
Correctly guard WebFrame methods needed only with the Chromium HTTP stack

This fixes the break in builds which use the Android HTTP stack.

Change-Id: I0feb885754c34e96216c8e0c28390808f424534e

12 years agoam d33401fd: Merge "Modify BrowserFrame.requestClientCert() to take the host and...
Steve Block [Wed, 12 Oct 2011 17:00:52 +0000 (10:00 -0700)]
am d33401fd: Merge "Modify BrowserFrame.requestClientCert() to take the host and port as a String" into ics-mr0

* commit 'd33401fdd5a0b8f3d82389f1d7cc5168fa14a122':
  Modify BrowserFrame.requestClientCert() to take the host and port as a String

12 years agoMerge "Modify BrowserFrame.requestClientCert() to take the host and port as a String...
Steve Block [Wed, 12 Oct 2011 16:58:20 +0000 (09:58 -0700)]
Merge "Modify BrowserFrame.requestClientCert() to take the host and port as a String" into ics-mr0

12 years agoModify BrowserFrame.requestClientCert() to take the host and port as a String
Steve Block [Wed, 12 Oct 2011 08:53:09 +0000 (09:53 +0100)]
Modify BrowserFrame.requestClientCert() to take the host and port as a String

Currently we get the host and port from Chromium as a std::string, convert to a
jbyte array to pass over JNI, then convert to String. It's simpler to convert
directly to jstring and to pass that over JNI.

This also fixes the fact that WebFrame::requestClientCert() fails to call
ReleaseByteArrayElements().

Requires https://android-git.corp.google.com/g/141238
in frameworks/base.

Bug: 5442710

Change-Id: If81fdef338ecdf2debeece0ebb9926bd261ae219

12 years agoam 47f387fd: (-s ours) Merge "Revert crash on bad scale - do not merge" into ics-mr0
Chris Craik [Wed, 12 Oct 2011 00:32:51 +0000 (17:32 -0700)]
am 47f387fd: (-s ours) Merge "Revert crash on bad scale - do not merge" into ics-mr0

* commit '47f387fdb2c9e81ee710f41b1e79f1419817801a':
  Revert crash on bad scale - do not merge

12 years agoMerge "Revert crash on bad scale - do not merge" into ics-mr0
Chris Craik [Wed, 12 Oct 2011 00:21:03 +0000 (17:21 -0700)]
Merge "Revert crash on bad scale - do not merge" into ics-mr0

12 years agoFallback handle if vertical glyphs are missing.
claireho [Tue, 11 Oct 2011 20:28:10 +0000 (13:28 -0700)]
Fallback handle if vertical glyphs are missing.

Bug 5094208 : browser VerticalWritingMode support.
Some punctuation has to change its glyph in vertical writing mode.
For example, dash in vertical text is a vertical bar. In
current Android system fonts and fallback fonts, none of them have the
substitute tables for vertical writing mode.  We call a conversion
function that converts the punctuation to the Unicode Vertical Form
(0xFE10 - 0xFE19) or to the closest glyphs as alternative when
vertical substitute tables are absent in fonts. Otherwise, we use
Harfbuzz to do the vertical glyph shapping.

Change-Id: If1d31bb72aaaba45ed50bbb75d6246cfa30e3428

12 years agoRevert crash on bad scale - do not merge
Chris Craik [Tue, 11 Oct 2011 23:10:50 +0000 (16:10 -0700)]
Revert crash on bad scale - do not merge

bug:5367327
Change-Id: I94589e007f9669fa1b43ab5cee42bfa44d9e503c

12 years agoam 14492673: Merge "Fixed infinite redraw loop from running out of layer tiles" into...
Chris Craik [Tue, 11 Oct 2011 23:10:24 +0000 (16:10 -0700)]
am 14492673: Merge "Fixed infinite redraw loop from running out of layer tiles" into ics-mr0

* commit '144926737b2b4e6f76f5c6347db3093b8a483e21':
  Fixed infinite redraw loop from running out of layer tiles

12 years agoMerge "Fixed infinite redraw loop from running out of layer tiles" into ics-mr0
Chris Craik [Tue, 11 Oct 2011 23:08:53 +0000 (16:08 -0700)]
Merge "Fixed infinite redraw loop from running out of layer tiles" into ics-mr0

12 years agoam 46f244f2: Merge "Mark tiles discarded by TransferQueue as dirty" into ics-mr0
Chris Craik [Tue, 11 Oct 2011 23:04:22 +0000 (16:04 -0700)]
am 46f244f2: Merge "Mark tiles discarded by TransferQueue as dirty" into ics-mr0

* commit '46f244f2351cbaa1a251d79605ed6d7e21d1d53f':
  Mark tiles discarded by TransferQueue as dirty

12 years agoMerge "Mark tiles discarded by TransferQueue as dirty" into ics-mr0
Chris Craik [Tue, 11 Oct 2011 23:02:01 +0000 (16:02 -0700)]
Merge "Mark tiles discarded by TransferQueue as dirty" into ics-mr0

12 years agoam d253060c: Merge "Fix repaint bug. In some cases when the content size changes...
Nicolas Roard [Tue, 11 Oct 2011 22:44:00 +0000 (15:44 -0700)]
am d253060c: Merge "Fix repaint bug. In some cases when the content size changes, the inval rects we receive from webkit do not fully cover the area that changed. For now, fix this by forcing the invalidation of the entire content (note: we only invalidate the tiles -

* commit 'd253060c6edf139871ea630f884cc6631a162854':
  Fix repaint bug. In some cases when the content size changes, the inval rects we receive from webkit do not fully cover the area that changed. For now, fix this by forcing the invalidation of the entire content (note: we only invalidate the tiles -- we do not force an actual regeneration of the pictureset for the entire content)

12 years agoMerge "Fix repaint bug. In some cases when the content size changes, the inval rects...
Nicolas Roard [Tue, 11 Oct 2011 22:42:44 +0000 (15:42 -0700)]
Merge "Fix repaint bug. In some cases when the content size changes, the inval rects we receive from webkit do not fully cover the area that changed. For now, fix this by forcing the invalidation of the entire content (note: we only invalidate the tiles -- we do not force an actual regeneration of the pictureset for the entire content)" into ics-mr0

12 years agoFixed infinite redraw loop from running out of layer tiles
Chris Craik [Tue, 11 Oct 2011 21:35:51 +0000 (14:35 -0700)]
Fixed infinite redraw loop from running out of layer tiles

bug:5349958

Clear a flag when a layer tile can't allocate a texture, and only try and redraw
from tile dirtiness if that flag is set.

Also, don't ask for redraw if offending tiles are offscreen.

Change-Id: Iadb0cb267a9c1f308e5b42a6e0e3b4bc71d18ece

12 years agoam 85e38448: Support spellcheck attribute in input fields
Michael Kolb [Tue, 11 Oct 2011 22:33:34 +0000 (15:33 -0700)]
am 85e38448: Support spellcheck attribute in input fields

* commit '85e3844824444f304a15e57f62d2e5670fb3bf47':
  Support spellcheck attribute in input fields

12 years agoFix repaint bug. In some cases when the content size changes, the
Nicolas Roard [Tue, 11 Oct 2011 22:12:09 +0000 (15:12 -0700)]
Fix repaint bug. In some cases when the content size changes, the
inval rects we receive from webkit do not fully cover the area that
changed. For now, fix this by forcing the invalidation of the entire
content (note: we only invalidate the tiles -- we do not force an
actual regeneration of the pictureset for the entire content)

bug:5361852
Change-Id: I682766d0101d4c8b8a38357dc7846c0030c1f9dc

12 years agoMark tiles discarded by TransferQueue as dirty
Chris Craik [Mon, 10 Oct 2011 22:58:27 +0000 (15:58 -0700)]
Mark tiles discarded by TransferQueue as dirty

bug:5409902

Tiles were being discarded from the queue (and simply unsuccessfully
added). This caused them to get stuck in the 'ValidatedUntransferred' state.

Now if a tile isn't added successfully, or if it's discarded, it removes its
painting texture and will have to repaint from scratch.

Change-Id: I551e00fb8a6be3b0f3cabeabaa91e8b8b30019d5

12 years agoSupport spellcheck attribute in input fields
Michael Kolb [Tue, 11 Oct 2011 21:05:45 +0000 (14:05 -0700)]
Support spellcheck attribute in input fields

    Bug: 5267547

    When spellcheck is set to false, disable the
    IME suggestions and the spelling error highlighting

Change-Id: Ib62bc760334159aa086214cc6f1388d1abe89cb4

12 years agoMerge "Support "Vertical Writing Mode"."
claireho [Tue, 11 Oct 2011 19:31:25 +0000 (12:31 -0700)]
Merge "Support "Vertical Writing Mode"."

12 years agoMerge "Delay sending Plugin Draw event until the surface is ready."
Derek Sollenberger [Mon, 10 Oct 2011 20:28:07 +0000 (13:28 -0700)]
Merge "Delay sending Plugin Draw event until the surface is ready."

12 years agoDelay sending Plugin Draw event until the surface is ready.
Derek Sollenberger [Mon, 22 Aug 2011 17:46:32 +0000 (13:46 -0400)]
Delay sending Plugin Draw event until the surface is ready.

bug: 5436630
Change-Id: I058a0d750f185c56c0fe70ac0bff7c9781c35dc8

12 years agoMerge "Fixed URL lookup for data URLs."
George Mount [Mon, 10 Oct 2011 20:06:48 +0000 (13:06 -0700)]
Merge "Fixed URL lookup for data URLs."

12 years agoMerge "Double buffering for layers"
Chris Craik [Mon, 10 Oct 2011 20:05:29 +0000 (13:05 -0700)]
Merge "Double buffering for layers"

12 years agoDouble buffering for layers
Chris Craik [Wed, 5 Oct 2011 20:19:22 +0000 (13:19 -0700)]
Double buffering for layers

bug:5309934

Defer picture changes and invalidations when:
1) a tiled texture hasn't completed painting
2) the user is scrolling (done in PaintedSurface)

Remaining issues:
Should block updates on zooming

Change-Id: I60ee722e651f96e95a4ac0e819999fc1589037c3

12 years agoAdding support for multiple video surfaces per plugin.
Derek Sollenberger [Wed, 14 Sep 2011 18:11:23 +0000 (14:11 -0400)]
Adding support for multiple video surfaces per plugin.

This CL refactors our implementation to support multiple
non-overlapping video surfaces per plugin instance. This
will allow plugins to hardware decode simulatenous video
streams using hardware.

Additionally, this CL improves plugin startup latency by
constructing the primary plugin surface as soon as possible
in the UI thread.

bug: 5318290
Change-Id: I53ceb8b6696159ecc4765c922704727a9a75a17d

12 years agoThe new FAST_PICTURESET code path seems to trigger repaint
Nicolas Roard [Sat, 8 Oct 2011 01:31:58 +0000 (18:31 -0700)]
The new FAST_PICTURESET code path seems to trigger repaint
bugs in some cases. The current performance improvement of this
code path is reasonably small compared to the default PICTURESET.
This CL turns off FAST_PICTURESET for now (we will revisit the
issue in the future)

bug:5406701
Change-Id: I7ab7addf14d46b28ebf6bd4ff7f2b3b4ee044cc4

12 years agoFixed URL lookup for data URLs.
George Mount [Fri, 7 Oct 2011 22:41:57 +0000 (15:41 -0700)]
Fixed URL lookup for data URLs.

 Bug 5383517
 Data URLs were being stripped, preventing long click from doing anything.

Change-Id: I5eba97725375a155d39f2489f97714a2b8bc8189

12 years agoMerge "Switch find on page to GL"
John Reck [Fri, 7 Oct 2011 20:10:24 +0000 (13:10 -0700)]
Merge "Switch find on page to GL"

12 years agoSwitch find on page to GL
John Reck [Thu, 6 Oct 2011 23:17:19 +0000 (16:17 -0700)]
Switch find on page to GL

 Bug: 5354191
 Refactor how extras are drawn in GL

Change-Id: Ifb61c73abecd2d074903a7fb44fe4b82aacc64ac

12 years agoMerge "Fix scrolling offset of ScrollableLayer when we touch them. With the new direc...
Nicolas Roard [Fri, 7 Oct 2011 19:04:05 +0000 (12:04 -0700)]
Merge "Fix scrolling offset of ScrollableLayer when we touch them. With the new direct update, we are setting directly the layers position; but for ScrollableLayer, we already set their position in the UI, so for them their webkit counterpart is actually older/wrong. The fix consists simply in not updating the positioning of ScrollableLayer."

12 years agoFix scrolling offset of ScrollableLayer when we touch them.
Nicolas Roard [Fri, 7 Oct 2011 18:54:14 +0000 (11:54 -0700)]
Fix scrolling offset of ScrollableLayer when we touch them.
With the new direct update, we are setting directly the layers position;
but for ScrollableLayer, we already set their position in the UI, so for
them their webkit counterpart is actually older/wrong.
The fix consists simply in not updating the positioning of
ScrollableLayer.

bug:5421867
Change-Id: Ia14e41aff641173a45ed898cd3b3eceab3f7bd7a

12 years agoMerge "Fix inverted mode rendering"
Nicolas Roard [Fri, 7 Oct 2011 17:40:12 +0000 (10:40 -0700)]
Merge "Fix inverted mode rendering"

12 years agoFix WebViewCore uninitialised data.
Ben Murdoch [Fri, 7 Oct 2011 11:58:57 +0000 (12:58 +0100)]
Fix WebViewCore uninitialised data.

Improve WebViewCore initialisation by using the
initialiser list rather than assignment in the
constructor. Ensure that *all* members get an initial
value.

Also remove the WebViewCore::reset(bool) function, that
was only being called from the constructor and so is
now unnecessary.

Additionally remove two completely unused members and one
(m_now) that was being written once but never read (and only
ever behind a debug #ifdef).

Bug: 5399726

Change-Id: I29216737c2406eed1cdc426d011bca0728fdf232

12 years agoFix inverted mode rendering
Nicolas Roard [Fri, 7 Oct 2011 01:50:02 +0000 (18:50 -0700)]
Fix inverted mode rendering

The GPU texture upload workaround broke it, as we are not using the
Shaders when doing the copy. Adding some code to instead use a final
shader that can be either normal or inverted, on demand.

bug:5424589

Change-Id: If845233e12c65738925fbc81b21f1dc0fd38546e

12 years agoFix repaint request logic for the image layer code path
Nicolas Roard [Fri, 7 Oct 2011 00:10:55 +0000 (17:10 -0700)]
Fix repaint request logic for the image layer code path

bug:5425608 bug:5218173

Change-Id: Ib14a5e53466b05781a842e24443e89a4779dba68

12 years agoFix repaint issues with overflow layers
Nicolas Roard [Wed, 5 Oct 2011 23:14:36 +0000 (16:14 -0700)]
Fix repaint issues with overflow layers

bug:5297412
Change-Id: I6b483e174339bc854e5d301c5c1f2e5851e57772

12 years agoMerge "Streamline the layers update codepath. Directly update the layers transform...
Nicolas Roard [Wed, 5 Oct 2011 18:35:24 +0000 (11:35 -0700)]
Merge "Streamline the layers update codepath. Directly update the layers transform and position. This makes updates faster and less dependent on other webkit work."

12 years agoStreamline the layers update codepath.
Nicolas Roard [Wed, 5 Oct 2011 01:01:40 +0000 (18:01 -0700)]
Streamline the layers update codepath.
Directly update the layers transform and position.
This makes updates faster and less dependent on other webkit work.

counterpart java CL: https://android-git.corp.google.com/g/#/c/139853/

bug:5218173

Change-Id: I03a76ab853e81f0f12177fb785707ffb8dace330

12 years agoMerge "Remove m_contentsImage usage"
John Reck [Wed, 5 Oct 2011 00:41:15 +0000 (17:41 -0700)]
Merge "Remove m_contentsImage usage"

12 years agoRemove m_contentsImage usage
John Reck [Tue, 4 Oct 2011 23:04:01 +0000 (16:04 -0700)]
Remove m_contentsImage usage

 Bug: 5410522

Change-Id: Id8457715665dbd51761900011beba57b7f634fff

12 years agoFix ring clip
John Reck [Tue, 4 Oct 2011 16:12:27 +0000 (09:12 -0700)]
Fix ring clip

 Bug: 5406350

Change-Id: I95507ee163a2cb41cc605efaabf32b7cad814533

12 years agoallow for compositing of 1x1 plugins
Chris Craik [Mon, 3 Oct 2011 21:38:23 +0000 (14:38 -0700)]
allow for compositing of 1x1 plugins

bug:5382635

Change-Id: I0da21245ec7808da452941a45ae5008e278f3007

12 years agoHW accelate button focus rings
John Reck [Mon, 3 Oct 2011 16:26:49 +0000 (09:26 -0700)]
HW accelate button focus rings

 Bug: 5353510

Change-Id: I1bb306381864357dab4c38fd5c182fda2dd08d8c

12 years agoMerge "Fine-tune the number of buckets in PictureSet add some clearing logic as well"
Nicolas Roard [Sat, 1 Oct 2011 01:49:21 +0000 (18:49 -0700)]
Merge "Fine-tune the number of buckets in PictureSet add some clearing logic as well"

12 years agoFine-tune the number of buckets in PictureSet
Nicolas Roard [Sat, 1 Oct 2011 01:24:04 +0000 (18:24 -0700)]
Fine-tune the number of buckets in PictureSet
add some clearing logic as well

bug:5145259 bug:5255262
Change-Id: I764e2b67ff15893e853e36185cd610e7293d0129

12 years agoMerge "Turn on the new Gpu Upload code path"
Teng-Hui Zhu [Fri, 30 Sep 2011 22:34:25 +0000 (15:34 -0700)]
Merge "Turn on the new Gpu Upload code path"

12 years agoMerge "Fix low-res tiles rendering too early"
Chris Craik [Fri, 30 Sep 2011 22:31:01 +0000 (15:31 -0700)]
Merge "Fix low-res tiles rendering too early"

12 years agoTurn on the new Gpu Upload code path
Teng-Hui Zhu [Fri, 30 Sep 2011 21:47:05 +0000 (14:47 -0700)]
Turn on the new Gpu Upload code path

bug:5395019
Change-Id: Ifc16eb59fec62e5f4b95ad4e35f140969226506d

12 years agoFix low-res tiles rendering too early
Chris Craik [Fri, 30 Sep 2011 21:16:28 +0000 (14:16 -0700)]
Fix low-res tiles rendering too early

bug:5369978

Low-res tiles were being prepared before high res, so despite being lower
priority, they had a chance to start rendering when a new page was loaded before
any other tiles existed in the priority queue.

Change-Id: I479f01a00c2507bcabf13929eb099571d8886c4c

12 years agoMerge "The new Gpu upload code path."
Teng-Hui Zhu [Fri, 30 Sep 2011 19:59:01 +0000 (12:59 -0700)]
Merge "The new Gpu upload code path."

12 years agoMerge "Initialize to BUCKET_SIZE instead of 0"
Russell Brenner [Fri, 30 Sep 2011 18:00:21 +0000 (11:00 -0700)]
Merge "Initialize to BUCKET_SIZE instead of 0"

12 years agoInitialize to BUCKET_SIZE instead of 0
Russell Brenner [Thu, 29 Sep 2011 20:11:14 +0000 (13:11 -0700)]
Initialize to BUCKET_SIZE instead of 0

Some execution path appears to be leaving mBucketSizeX and/or Y set
to 0, causing a divide-by-zero exception. This should only be
feasible when mWidth and mHeight are also 0, but, to keep things
safe, we'll initialize to BUCKET_SIZE instead.

Bug: 5391435
Change-Id: I4e01f980731619e6a6fb70a6eb315c44dd677c7b

12 years agoThe new Gpu upload code path.
Teng-Hui Zhu [Thu, 29 Sep 2011 00:28:58 +0000 (17:28 -0700)]
The new Gpu upload code path.

This relied on the hacky Surface Texture change:
https://android-git.corp.google.com/g/#/c/138516/

bug:5347539
Change-Id: I78e0f677a9c201ba49e11ddb24fa095471cbf3c9

12 years agoMerge "Make the Gpu Upload path faster"
Teng-Hui Zhu [Fri, 30 Sep 2011 16:22:46 +0000 (09:22 -0700)]
Merge "Make the Gpu Upload path faster"

12 years agoMake the Gpu Upload path faster
Teng-Hui Zhu [Fri, 30 Sep 2011 00:51:13 +0000 (17:51 -0700)]
Make the Gpu Upload path faster

Basically, move unnecessary function calls out from the loop.
And move the clear related code into DEBUG only.
This is saving us some CPU overhead.

bug:5347539
Change-Id: I237162fd6a502e93ea32c9856f907c5a9ddb2e78

12 years agoMerge "Use state machine to track tile state"
Chris Craik [Thu, 29 Sep 2011 23:44:59 +0000 (16:44 -0700)]
Merge "Use state machine to track tile state"

12 years agoUse state machine to track tile state
Chris Craik [Tue, 27 Sep 2011 17:17:32 +0000 (10:17 -0700)]
Use state machine to track tile state

bug:5369978

Manage invalidations, painting, transferring, and swapping with a
state machine.

notes:
* readyFor shouldn't be needed if all of the events that would cause a readyfor
  fail mark the tile dirty (such as stealing, scale change)

* changing the scale of a page should discard all textures

* m_dirty should be more gracefully worked into the state machine

* a tile may drop dirtiness notifications if it's already painting, since upon
  completion m_dirty will be cleared - (issues for layers only)

Change-Id: I5909fb5d208da2fb276e223c56bf143741a9a24c

12 years agoAdd debug informations to keep better track of layers allocation
Nicolas Roard [Thu, 29 Sep 2011 20:58:29 +0000 (13:58 -0700)]
Add debug informations to keep better track of layers allocation
Fix coherency issue in ClassTracker (it's accessed from two different
threads and we were not locking the datastructure)

bug:5366807

Change-Id: I7dbd4417a49f4b3af3e09d614cec6b4568b09f7a

12 years agoFix a memory leak with LayerAndroid
Nicolas Roard [Thu, 29 Sep 2011 02:08:00 +0000 (19:08 -0700)]
Fix a memory leak with LayerAndroid
(we had a cycle between LayerAndroid and PaintedSurface)

bug:5366807
Change-Id: I1b7341984b3cc6fd998b9ab54205bb963eb9223d

12 years agoMerge "Make bucket sizes in increments of 256"
Russell Brenner [Wed, 28 Sep 2011 20:22:18 +0000 (13:22 -0700)]
Merge "Make bucket sizes in increments of 256"

12 years agoMerge "Don't scale text selection handles"
John Reck [Wed, 28 Sep 2011 18:12:08 +0000 (11:12 -0700)]
Merge "Don't scale text selection handles"

12 years agoDon't scale text selection handles
John Reck [Wed, 28 Sep 2011 03:15:22 +0000 (20:15 -0700)]
Don't scale text selection handles

 Bug: 5367280
 Don't scale the old software handle and, more importantly,
 don't scale the touch target.

Change-Id: I9a731b2117b3f2fe3bd6ca35388da61c47724d91

12 years agoDynamically set the upper limit of layer tiles' texture number
Teng-Hui Zhu [Tue, 27 Sep 2011 22:38:35 +0000 (15:38 -0700)]
Dynamically set the upper limit of layer tiles' texture number

Instead of using a static number, now the upper limit of the layer tiles is
also dependent on the viewport size.
At the same time, get rid of the old code only useful without tiled layer.

bug:5347539

Change-Id: Ie720449ea15043d42a56f4c8af57820fa7f6e93c

12 years agoMake bucket sizes in increments of 256
Russell Brenner [Tue, 27 Sep 2011 22:48:24 +0000 (15:48 -0700)]
Make bucket sizes in increments of 256

As pages grow, limit the number of bucket refills by making them in
chunks of 256 (BUCKET_SIZE). This also fixes a divide-by-zero
error in the earlier update to limit the bucket brigade.

Bug: 5382243
Change-Id: I4ee6eb86c5934175b7cf0f7b69bcfb416d68b967

12 years agoMerge "Increase the texture number for layers"
Teng-Hui Zhu [Tue, 27 Sep 2011 18:35:22 +0000 (11:35 -0700)]
Merge "Increase the texture number for layers"

12 years agoMerge "Do not force "mobile" sites to use a viewport"
Ben Murdoch [Tue, 27 Sep 2011 17:22:51 +0000 (10:22 -0700)]
Merge "Do not force "mobile" sites to use a viewport"

12 years agoIncrease the texture number for layers
Teng-Hui Zhu [Tue, 27 Sep 2011 16:54:40 +0000 (09:54 -0700)]
Increase the texture number for layers

This will break the Gpu upload code path on phones which is turned off by
default, but it will work on the tablet.

bug:5347539
Change-Id: Idc9889a8681035addd85843191322039a69110e5

12 years agoMerge "Limit the number of buckets in a PictureSet"
Russell Brenner [Tue, 27 Sep 2011 16:32:59 +0000 (09:32 -0700)]
Merge "Limit the number of buckets in a PictureSet"

12 years agoMerge "Add temporary work-around to allow loading local data from Documents with...
Steve Block [Tue, 27 Sep 2011 09:11:58 +0000 (02:11 -0700)]
Merge "Add temporary work-around to allow loading local data from Documents with substituted data"

12 years agoLimit the number of buckets in a PictureSet
Russell Brenner [Thu, 22 Sep 2011 23:54:49 +0000 (16:54 -0700)]
Limit the number of buckets in a PictureSet

Added MAX_BUCKET_COUNT_X and _Y so that absurdly large pages, such as
those in LayoutTests, don't run out of memory due to millions of
buckets.

Renamed checkDimensions() to setDimensions() and made inval an
optional parameter so that it can be used more generally within
PictureSet to adjust settings.

Bug: 5300146
Change-Id: Ic6c2dc97085ef38d5782aa005b71a6797d165e24

12 years agoMerge "Egl sync code path should be only activated when we are in GpuUpload mode"
Teng-Hui Zhu [Mon, 26 Sep 2011 22:37:44 +0000 (15:37 -0700)]
Merge "Egl sync code path should be only activated when we are in GpuUpload mode"

12 years agoMerge "Add shared images for layers"
Nicolas Roard [Mon, 26 Sep 2011 22:06:45 +0000 (15:06 -0700)]
Merge "Add shared images for layers"

12 years agoAdd temporary work-around to allow loading local data from Documents with substituted...
Steve Block [Mon, 26 Sep 2011 12:12:35 +0000 (13:12 +0100)]
Add temporary work-around to allow loading local data from Documents with substituted data

This is a temporary workaround for a WebCore bug which is awaiting resolution.
See https://bugs.webkit.org/show_bug.cgi?id=68711 and this bug for details.

This change adds a new property to Frame to track when its Document pointer is
up-to-date. This is is used only when setting up the securty context for a
Document with substituted data and allows us to skip the buggy check in
Document::loader().

Bug: 5188895
Change-Id: I5d7d2cca83fa0c6db084d505f5b48207046a9cd0

12 years agoEgl sync code path should be only activated when we are in GpuUpload mode
Teng-Hui Zhu [Mon, 26 Sep 2011 20:33:30 +0000 (13:33 -0700)]
Egl sync code path should be only activated when we are in GpuUpload mode

bug:5347539
Change-Id: I12b8b45792d5aa684295b897be308181a7d44631

12 years agoAdd shared images for layers
Nicolas Roard [Wed, 31 Aug 2011 23:02:00 +0000 (16:02 -0700)]
Add shared images for layers

bug:5242595 bug:5218173

Change-Id: I37d395e85441671312aac3e236cc8276019aa990

12 years agoReconcile with ics-factoryrom-release
The Android Open Source Project [Mon, 26 Sep 2011 19:11:27 +0000 (12:11 -0700)]
Reconcile with ics-factoryrom-release

Change-Id: I796bddb9801561a822514af709f557b017cd4cea

12 years agoDo not force "mobile" sites to use a viewport
Ben Murdoch [Mon, 26 Sep 2011 13:01:21 +0000 (14:01 +0100)]
Do not force "mobile" sites to use a viewport

Remove some old code for detecting mobile sites and forcing
them to use a mobil viewport. This broke some websites, e.g.
Wikipedia.

Bug: 5371421
Change-Id: Icb93a1345f24d77d23b783fc965558a11135037a

12 years agomerge in ics-factoryrom-release history after reset to ics-factoryrom
The Android Automerger [Mon, 26 Sep 2011 01:15:15 +0000 (18:15 -0700)]
merge in ics-factoryrom-release history after reset to ics-factoryrom

12 years agoam 8e0c7b9e: (-s ours) Cherry-pick master fix for ANRs - Do not merge
Bart Sears [Sun, 25 Sep 2011 23:10:37 +0000 (16:10 -0700)]
am 8e0c7b9e: (-s ours) Cherry-pick master fix for ANRs - Do not merge

* commit '8e0c7b9ea6359c938b4a1994a6e890eaff473a4b':
  Cherry-pick master fix for ANRs - Do not merge

12 years agoCherry-pick master fix for ANRs - Do not merge
Bart Sears [Sun, 25 Sep 2011 20:59:29 +0000 (13:59 -0700)]
Cherry-pick master fix for ANRs - Do not merge

cherry-pick CL: 941349353627b11c3b9a4deeee6cd7ae831836c0

Avoid infinite prepare loop if bad scale provided

bug:5362098

Note: we shouldn't be getting bad scales, now that the scale corruption issue
has been reverted. Added logging for these to wrap transfer queue as well to
detect fp corruption regression.

Change-Id: I5e6d2afc1d483452140fab5390395c9581db86ca

12 years agomerge in ics-factoryrom-release history after reset to ics-factoryrom
The Android Automerger [Sun, 25 Sep 2011 13:28:13 +0000 (06:28 -0700)]
merge in ics-factoryrom-release history after reset to ics-factoryrom

12 years agoam fab9df83: Rename Lohit_Hindi to Lohit-Devanagari
Russell Brenner [Sat, 24 Sep 2011 00:40:32 +0000 (17:40 -0700)]
am fab9df83: Rename Lohit_Hindi to Lohit-Devanagari

* commit 'fab9df838471ec0516f4ce0360c6bc2f7248f53f':
  Rename Lohit_Hindi to Lohit-Devanagari

12 years agoMerge "Crash if stack is corrupted"
Chris Craik [Fri, 23 Sep 2011 23:33:32 +0000 (16:33 -0700)]
Merge "Crash if stack is corrupted"

12 years agoCrash if stack is corrupted
Chris Craik [Fri, 23 Sep 2011 23:21:40 +0000 (16:21 -0700)]
Crash if stack is corrupted

bug:5362098
For tracking down very rare, very wrong scale values

Change-Id: I00a5dd8d75c47c4c13c903e587b3983ea23496d7

12 years agoMerge "Enable the CPU upload path for Browser."
Teng-Hui Zhu [Fri, 23 Sep 2011 22:58:09 +0000 (15:58 -0700)]
Merge "Enable the CPU upload path for Browser."

12 years agoRename Lohit_Hindi to Lohit-Devanagari
Russell Brenner [Fri, 23 Sep 2011 21:53:35 +0000 (14:53 -0700)]
Rename Lohit_Hindi to Lohit-Devanagari

Bug: 5364436
Change-Id: I0ce921e330ab892d8334b3b8bf4539e6a175a78c

12 years agoMerge "Avoid infinite prepare loop if bad scale provided"
Chris Craik [Fri, 23 Sep 2011 21:54:03 +0000 (14:54 -0700)]
Merge "Avoid infinite prepare loop if bad scale provided"

12 years agoEnable the CPU upload path for Browser.
Teng-Hui Zhu [Fri, 23 Sep 2011 20:34:05 +0000 (13:34 -0700)]
Enable the CPU upload path for Browser.

The default for WebView should be still be GPU upload.
That means Gmail will be default to using GPU upload.

bug:5347539

The browser change is in
https://android-git.corp.google.com/g/#/c/137606/

Change-Id: Ic7d42331511b24cf8a58f2f5fb64aaffc18cae5f

12 years agoAvoid infinite prepare loop if bad scale provided
Chris Craik [Fri, 23 Sep 2011 20:55:59 +0000 (13:55 -0700)]
Avoid infinite prepare loop if bad scale provided

bug:5362098

Note: we shouldn't be getting bad scales, now that the scale corruption issue
has been reverted. Added logging for these to wrap transfer queue as well to
detect fp corruption regression.

Change-Id: I5e6d2afc1d483452140fab5390395c9581db86ca

12 years agoMerge "If we have a bad framework inval rect, request entire screen redraw"
Chris Craik [Fri, 23 Sep 2011 20:54:16 +0000 (13:54 -0700)]
Merge "If we have a bad framework inval rect, request entire screen redraw"

12 years agoIf we have a bad framework inval rect, request entire screen redraw
Chris Craik [Fri, 23 Sep 2011 20:35:27 +0000 (13:35 -0700)]
If we have a bad framework inval rect, request entire screen redraw

bug:5365139

This problem is occurring because we convert webkit invals to framework ones,
but webkit invals don't account for stolen/discarded tiles/textures.

This is just a temporary fix, see also bug:5321078 which describes the root of
the problem.

Change-Id: I6d823315dba8d669171c9c7c9ed3ca20fbed07ae