OSDN Git Service

android-x86/external-webkit.git
13 years agoMerge "Fix text-shadow style when blur is 0px"
Teng-Hui Zhu [Mon, 8 Nov 2010 17:25:40 +0000 (09:25 -0800)]
Merge "Fix text-shadow style when blur is 0px"

13 years agoFix some compiler warnings
Steve Block [Wed, 3 Nov 2010 20:41:40 +0000 (21:41 +0100)]
Fix some compiler warnings

Noe that in FindCanvas.h we include
external/icu4c/common/unicode/umachine.h, rather than
external/webkit/WebCore/icu/unicode/umachine.h to avoid warnings about
redefining U_NAMESPACE_BEGIN

We should be using the header from external/icu4c because we don't
include any ICU headers from WebCore.

The fix works becuase external/icu4c/common is on the include path,
but WebCore/icu is not.

Change-Id: I5c0ff1c5815018ba97962595debca12a17bdd0cd

13 years agoMerge "Fix compilation error when ANDROID_LAYOUT is not defined."
Shimeng (Simon) Wang [Sat, 6 Nov 2010 01:06:20 +0000 (18:06 -0700)]
Merge "Fix compilation error when ANDROID_LAYOUT is not defined."

13 years agoFix text-shadow style when blur is 0px
Teng-Hui Zhu [Sat, 6 Nov 2010 01:04:27 +0000 (18:04 -0700)]
Fix text-shadow style when blur is 0px

When blur is 0px , the should at least show up without any blurry.
http://b/issue?id=2581273

Change-Id: I3db26647c1ed3f7dd29e7045ae6f5c5b81722da6

13 years agoFix compilation error when ANDROID_LAYOUT is not defined.
Shimeng (Simon) Wang [Fri, 5 Nov 2010 22:45:11 +0000 (15:45 -0700)]
Fix compilation error when ANDROID_LAYOUT is not defined.

Change-Id: I6bdb9832bd94017a5f4a894474c30f2cea89d39d

13 years agoFix the blank screen bug.
Nicolas Roard [Fri, 5 Nov 2010 04:14:41 +0000 (21:14 -0700)]
Fix the blank screen bug.

As the texture generator thread isn't an infinite loop, we could miss
a signaled new request.

Bug:3155078
Change-Id: Ifad1ec356b061eec24399cad4b572866f6d74f06

13 years agoSupport partial invalidation of tiles based on webkit's inval rect.
Derek Sollenberger [Fri, 5 Nov 2010 16:07:58 +0000 (12:07 -0400)]
Support partial invalidation of tiles based on webkit's inval rect.

This CL reverts the rollback of the original CL now that we have
fixed bug #3165953.

This reverts commit 144ccd9c8dba05ffaa0ae598f9b70032050fc20e.

Change-Id: Ibb5254f652ed95aa3b7b71bf24a2578c51e74a70

13 years agoMerge "Do not create GL specific objects if we are not using hw acceleration"
Derek Sollenberger [Fri, 5 Nov 2010 16:35:28 +0000 (16:35 +0000)]
Merge "Do not create GL specific objects if we are not using hw acceleration"

13 years agoDo not create GL specific objects if we are not using hw acceleration
Derek Sollenberger [Fri, 5 Nov 2010 13:16:18 +0000 (09:16 -0400)]
Do not create GL specific objects if we are not using hw acceleration

bug: 3166223
Change-Id: I5c18b76ac19119c3620636ce5d3b33892e18f523

13 years agoMerge "Code cleanup for FontCacheAndroid.cpp"
Russell Brenner [Fri, 5 Nov 2010 15:56:02 +0000 (15:56 +0000)]
Merge "Code cleanup for FontCacheAndroid.cpp"

13 years agoMerge "Fix browser crash."
Patrick Scott [Fri, 5 Nov 2010 15:12:15 +0000 (15:12 +0000)]
Merge "Fix browser crash."

13 years agoFix browser crash.
Patrick Scott [Fri, 5 Nov 2010 15:09:11 +0000 (11:09 -0400)]
Fix browser crash.

Also fixed indenting (3 spaces?).

Change-Id: If4a137d1af3ec8a597e8b929e5b4f10240f2f054

13 years agoMerge "Update method to take const std::string& argument"
Kristian Monsen [Fri, 5 Nov 2010 14:44:45 +0000 (14:44 +0000)]
Merge "Update method to take const std::string& argument"

13 years agoMerge "Adjust layers when parent layer scrolls"
Cary Clark [Fri, 5 Nov 2010 14:17:11 +0000 (14:17 +0000)]
Merge "Adjust layers when parent layer scrolls"

13 years agoSend the AutoFill Preview string java side.
Ben Murdoch [Tue, 19 Oct 2010 18:41:08 +0000 (19:41 +0100)]
Send the AutoFill Preview string java side.

When we have determined that the form can be autofilled, send
a preview string over to Java so that we can display it in the
drop down box.

Requires a change in frameworks/base:
https://android-git.corp.google.com/g/#change,77127

Change-Id: Ia0dd899d659c6e5710155f33749255058d7c3faf

13 years agoUpdate method to take const std::string& argument
Kristian Monsen [Fri, 5 Nov 2010 10:45:07 +0000 (10:45 +0000)]
Update method to take const std::string& argument

It's safe to pass this across threads since NewRunnableMethod
copies the argument.

Change-Id: I620ebfe64ead7947ac3a6733507865938e04a8df

13 years agoRemove extras from GLWebViewState for now
Nicolas Roard [Thu, 4 Nov 2010 17:46:00 +0000 (10:46 -0700)]
Remove extras from GLWebViewState for now

Change-Id: Ic4d4866302645edb6c0a457ef6d0bd1d8f9cf484

13 years agoCode cleanup for FontCacheAndroid.cpp
Russell Brenner [Thu, 4 Nov 2010 21:02:05 +0000 (14:02 -0700)]
Code cleanup for FontCacheAndroid.cpp

Minor changes to allow check-webkit-style to complete without warning for
FontCacheAndroid.cpp.

Change-Id: Ifaa8634d53374c94251641d6b9daf5f8114af9e9

13 years agoMerge "Revert "Support partial invalidation of tiles based on webkit's inval rect.""
Derek Sollenberger [Thu, 4 Nov 2010 20:28:34 +0000 (13:28 -0700)]
Merge "Revert "Support partial invalidation of tiles based on webkit's inval rect.""

13 years agoRevert "Support partial invalidation of tiles based on webkit's inval rect."
Derek Sollenberger [Thu, 4 Nov 2010 20:22:28 +0000 (16:22 -0400)]
Revert "Support partial invalidation of tiles based on webkit's inval rect."

This reverts commit fc92ca2409a95b539274985812d88016b6b84b7e.

13 years agoAdjust layers when parent layer scrolls
Cary Clark [Mon, 1 Nov 2010 21:25:39 +0000 (17:25 -0400)]
Adjust layers when parent layer scrolls

When a page has layers, it causes the main body of the
page to become a layer as well. The scroll position
of the WebView must be applied to the layer to
translate the cursor rings correctly for hit-testing.

Today, applying the scroll position of the WebView
screws up overflow scrolling. Until this gets figured
out, skip using the scroll position in CacheFrame
adjust() and unadjust().

The position of the layer, and all of its parent layers,
should be considered when adjusting the rectangle
bounds contained by that layer. Before, only the
child layer's position was considered.

The clip used by scrollable divs may be initialized
but never set. If it is empty, ignore it, rather than
clipping out all content.

In CacheBuilder, remove a couple of unnecessary
lines -- the CacheInput initialization clears all
members.

In SelectText, reverse the order of the xfer mode
and paint objects to remove a Skia ref count assert.

In CachedInput, make the debug printout current.

Overall, added more debugging output (turned
off by default).

bug:3030370

Change-Id: Ic19c24b3bf33d081a1d0c1f8c06601dcb56ae881

13 years agoMerge "Sync up with chromium/FontLinux.cpp"
claireho [Thu, 4 Nov 2010 18:06:08 +0000 (11:06 -0700)]
Merge "Sync up with chromium/FontLinux.cpp"

13 years agoMap Chromium error codes to WebKit error codes.
Iain Merrick [Thu, 4 Nov 2010 17:25:14 +0000 (17:25 +0000)]
Map Chromium error codes to WebKit error codes.

On network errors, we were seeing strange "Data connectivity problem"
alert boxes. This is because we were using the HTTP status code instead
of a real error code; and when there's a connection error, the HTTP
status is uninitialised, so the "error code" was 0 (no error).

Bug: 3140283
Change-Id: I0566c29359a0ecb2998573d1a2a77306922c87be

13 years agoMerge "Fixed 2720133. Better lookup with @font-face local"
Russell Brenner [Thu, 4 Nov 2010 17:35:36 +0000 (10:35 -0700)]
Merge "Fixed 2720133. Better lookup with @font-face local"

13 years agoMerge "Support partial invalidation of tiles based on webkit's inval rect."
Derek Sollenberger [Thu, 4 Nov 2010 16:56:56 +0000 (09:56 -0700)]
Merge "Support partial invalidation of tiles based on webkit's inval rect."

13 years agoLet Chrome HTTP stack handle number of connections
Kristian Monsen [Thu, 4 Nov 2010 14:49:26 +0000 (14:49 +0000)]
Let Chrome HTTP stack handle number of connections

First of changes based on willchan's feedback.

Note that this code doesn't really work as intended anyway.

Change-Id: I91d43f1daa78248700738338cf3d24a12496ac19

13 years agoMerge "Hook chromium logging up in WebKit"
Kristian Monsen [Thu, 4 Nov 2010 14:13:31 +0000 (07:13 -0700)]
Merge "Hook chromium logging up in WebKit"

13 years agoMerge "Add support for file uploads in WebKit"
Kristian Monsen [Thu, 4 Nov 2010 14:13:23 +0000 (07:13 -0700)]
Merge "Add support for file uploads in WebKit"

13 years agoSupport partial invalidation of tiles based on webkit's inval rect.
Derek Sollenberger [Tue, 2 Nov 2010 15:15:08 +0000 (11:15 -0400)]
Support partial invalidation of tiles based on webkit's inval rect.

This CL causes the selective repainting of BaseTiles (and their
associated GL textures) using the information provided by Webkit.
If tiles need repainted they are marked with as dirty and the
thread painting and uploading the textures will only operate on
dirty tiles. This change resulted in some significant refactoring
most of which revolved around these changes...

(1) Removed PaintingInfo from the Texture object and instead track
the state of the tile in Tile object.
(2) Removed all pending TileSets for a TiledPage when the page
produces a new set. This ensures that the tiles currently visible
to the user are painted instead of ones that may already be offscreen.

Change-Id: I93845d8e6e7b066e6bab84bcde11be4a6940002f

13 years agoHook chromium logging up in WebKit
Kristian Monsen [Wed, 3 Nov 2010 10:21:33 +0000 (10:21 +0000)]
Hook chromium logging up in WebKit

Change-Id: Ide08371e0bc8f74ef80fe25d416623bd03ac6038

13 years agoAdd SSL layout tests at WebKit r70209.
Iain Merrick [Thu, 4 Nov 2010 12:10:40 +0000 (12:10 +0000)]
Add SSL layout tests at WebKit r70209.

Only 3 tests, all passing. See http://b/issue?id=3164771

Change-Id: I2f3b03969a97618f1d4ba9dd91d782e1814914f0

13 years agoAdd support for file uploads in WebKit
Kristian Monsen [Mon, 1 Nov 2010 14:28:16 +0000 (14:28 +0000)]
Add support for file uploads in WebKit

This also moves some of the android file handling to Java so it is not
duplicated in both places.

This CL needs https://android-git.corp.google.com/g/#change,77400
in frameworks/base

Change-Id: I90c1726e6c323a9de3fd64f2e6feef4b64171053

13 years agoSync up with chromium/FontLinux.cpp
claireho [Wed, 3 Nov 2010 22:44:38 +0000 (15:44 -0700)]
Sync up with chromium/FontLinux.cpp

Adopted the code style change for TextRunWalker class from
chromium/FontLinux.cpp to android/FontAndroid.cpp.

Change-Id: I7f49efcaf5685181980a4ffa3b654d1577e3f9cd

13 years agoFixed 2720133. Better lookup with @font-face local
Russell Brenner [Wed, 3 Nov 2010 18:06:53 +0000 (11:06 -0700)]
Fixed 2720133. Better lookup with @font-face local

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.

Change-Id: I8007caae96f26cd58ea1cf5bf2da8170e3ce6d9d
Bug: 2720133

13 years agoMerge "Bug 3158404 : Performance improvement for browser complex script handle."
claireho [Wed, 3 Nov 2010 19:52:05 +0000 (12:52 -0700)]
Merge "Bug 3158404 : Performance improvement for browser complex script handle."

13 years agoBug 3158404 : Performance improvement for browser complex script handle.
claireho [Tue, 2 Nov 2010 21:26:59 +0000 (14:26 -0700)]
Bug 3158404 : Performance improvement for browser complex script handle.

1. Reduce the number of calls of normalization function.
   The purpose of NFC normalization is to compose base letter and non-spacing
   combining marks to be accented character. i.e. Convert A(U+0041) and
   acute(U+0301) to A with Acute(U+00C1).
   NFC includes three parts - decomposition, canonical sort and composition
   for entire string. It is very expensive. TextRunWalker class would like
   to convert '\t', '\n' and non-break space(\U+00A0) to space.  NFC cannot
   achieve this purpose. So, it is not necessary to call normalization for
   Font::treatAsSpace().
   This improvement saves the NFC nomalization calls for complex script text
   has '\t', '\n' or 0xA0 and does not have non-spacing combining marks.
2. FontAndroid.cpp scans text for space and punctuation in 2 separate
   loops. Combine these 2 loops to one pass.

Change-Id: I6b1ff20dd1072cf1f03066ce1c87ad39d0040779

13 years agoMerge Chromium at r63472 : WebKit fixes.
Iain Merrick [Tue, 2 Nov 2010 12:51:09 +0000 (12:51 +0000)]
Merge Chromium at r63472 : WebKit fixes.

Chromium now uses protocol buffers so we need libprotobuf.
See http://src.chromium.org/viewvc/chrome?view=rev&revision=63221

HttpCache takes an extra constructor argument.
See http://src.chromium.org/viewvc/chrome?view=rev&revision=61789

Change-Id: I5e78b0de9a8ff86040b13212455eb37a4ed0e39e

13 years agoMerge "Unref the picture instead of deleting it"
Cary Clark [Wed, 3 Nov 2010 14:46:49 +0000 (07:46 -0700)]
Merge "Unref the picture instead of deleting it"

13 years agoUnref the picture instead of deleting it
Cary Clark [Wed, 3 Nov 2010 14:21:19 +0000 (10:21 -0400)]
Unref the picture instead of deleting it

This fixes a couple of bugs related to layers
and hardware acceleration.

The picture used by select text is now accessed
asynchronously from the UI thread. The UI thread
now unrefs it rather than deleting it outright.

The layer position may not be up to date when
the cursor rings are drawn, so update them if
the cursor ring is in a layer.

bug:3112657
bug:3030370
Change-Id: I99d07175ec30dcba895a26cf1d44bc0364430793

13 years agoMerge "Add volume mute key."
Jeff Brown [Wed, 3 Nov 2010 00:38:28 +0000 (17:38 -0700)]
Merge "Add volume mute key."

13 years agoFix for deadlock
Nicolas Roard [Wed, 3 Nov 2010 00:07:14 +0000 (17:07 -0700)]
Fix for deadlock

Bug:3155066

Change-Id: I3e249050c779d10db75f97b177e09144a52f70b1

13 years agoSet the page background color in BaseLayerAndroid
Nicolas Roard [Tue, 2 Nov 2010 19:52:48 +0000 (12:52 -0700)]
Set the page background color in BaseLayerAndroid
This allows us to show the correct background even if the tiles aren't yet painted.

Change-Id: Ia3ef97770aadb5f64490e44f20110ea21578ccd0

13 years agoMerge "add function to find color on picture's edge"
Cary Clark [Tue, 2 Nov 2010 13:36:37 +0000 (06:36 -0700)]
Merge "add function to find color on picture's edge"

13 years agoUpdate the layers only
Nicolas Roard [Sat, 30 Oct 2010 01:39:15 +0000 (18:39 -0700)]
Update the layers only

C++ counterpart to https://android-git.corp.google.com/g/#change,77500

Bug:2975990

Change-Id: I3b8fd27f991d6776059a15eef36d0c9a7b44f9bb

13 years agoAdd volume mute key.
Jeff Brown [Mon, 1 Nov 2010 22:24:52 +0000 (15:24 -0700)]
Add volume mute key.

Also fixed incorrect keycode constant for numpad parens.

Bug: 2912307
Change-Id: I65d02ca1c131638ea70f333bf06982747f34223c

13 years agoadd function to find color on picture's edge
Cary Clark [Thu, 28 Oct 2010 21:29:48 +0000 (17:29 -0400)]
add function to find color on picture's edge

bug:3065048
Change-Id: I18ef06a4b9ac3bef077f08efd738958bd5abaafe

13 years agoMerge "Disable -Wsign-promo when building WebKit to avoid excessive warnings from...
Steve Block [Mon, 1 Nov 2010 16:45:35 +0000 (09:45 -0700)]
Merge "Disable -Wsign-promo when building WebKit to avoid excessive warnings from Chromium headers"

13 years agoDisable -Wsign-promo when building WebKit to avoid excessive warnings from Chromium...
Steve Block [Mon, 1 Nov 2010 16:15:34 +0000 (16:15 +0000)]
Disable -Wsign-promo when building WebKit to avoid excessive warnings from Chromium headers

If the root of the problem is fixed upstream, this warning can be
re-enabled.

Bug: 3143773
Change-Id: Ic419e353e3d0536d3a3c86f9e8e4615d30d9fb25

13 years agoFix WebCoreStringToJString() to not delete the local reference to the new jstring
Steve Block [Mon, 1 Nov 2010 13:16:30 +0000 (13:16 +0000)]
Fix WebCoreStringToJString() to not delete the local reference to the new jstring

The JNI spec allows for local objects to be GC'ed before the JNI frame
returns, so this is dangerous.

Also rename the method to WtfStringToJstring() to match the existing
jstringToWtfString() and update call sites to JNIEnv::NewString() to
make use of it.

Change-Id: I84aa25ba1aca2aee346bf6e3d83a58afc3a95219

13 years agoMerge "Enhance auto hyphenation."
Shimeng (Simon) Wang [Fri, 29 Oct 2010 18:44:01 +0000 (11:44 -0700)]
Merge "Enhance auto hyphenation."

13 years agoEnhance auto hyphenation.
Shimeng (Simon) Wang [Fri, 29 Oct 2010 00:14:27 +0000 (17:14 -0700)]
Enhance auto hyphenation.

1. Bypass leading white spaces, since webkit does pass them down.
2. Return better hyphenation point, since previously it's more
aggressive and causes display issue in Google books.

issue: 2672163
Change-Id: I8ae47f7c553f533f752d6f7c697cf2fffd421e5b

13 years agoMerge changes I28fb1392,I6b881465
Steve Block [Fri, 29 Oct 2010 15:51:22 +0000 (08:51 -0700)]
Merge changes I28fb1392,I6b881465

* changes:
  CookieManager.getCookie() should return null rather than the empty string
  Hook up remaining CookieManager methods

13 years agoNotify WebKit of redirects in the Chrome HTTP stack.
Iain Merrick [Fri, 29 Oct 2010 11:42:54 +0000 (12:42 +0100)]
Notify WebKit of redirects in the Chrome HTTP stack.

In this CL we just defer the redirect until WebKit has had a chance
to check it, and potentially cancel it.

There's a comment in the old code asking what to do if we're given
a bad request. Looks like this should never happen; it's a DCHECK in
Chrome so I've turned it into an ASSERT here.

We only follow redirects if WebKit does not modify the URL (this is
the same behaviour as Chrome).

Change-Id: I0c8b8cd61c501527a29dda5aca521a7df4a3ccef

13 years agoCookieManager.getCookie() should return null rather than the empty string
Steve Block [Fri, 29 Oct 2010 12:24:17 +0000 (13:24 +0100)]
CookieManager.getCookie() should return null rather than the empty string

This is to match the behaviour of the Android Java HTTP stack.

Bug: 3116410
Change-Id: I28fb1392c36754a7c2fd78bc76281db8318a1056

13 years agoHook up remaining CookieManager methods
Steve Block [Thu, 28 Oct 2010 14:35:47 +0000 (15:35 +0100)]
Hook up remaining CookieManager methods

hasCookies(), removeExpiredCookie(), removeSessionCookie() and setCookie()

Requires a change to frameworks/base ...
https://android-git.corp.google.com/g/76898

Bug: 3116410
Change-Id: I6b881465f78e090ed29b215785c4018119ccc5d8

13 years agoMerge "Allow Gmail attachments to be viewed."
Iain Merrick [Fri, 29 Oct 2010 10:22:24 +0000 (03:22 -0700)]
Merge "Allow Gmail attachments to be viewed."

13 years agoClear chromium side autofill profiles when java syncs a null profile.
Ben Murdoch [Tue, 26 Oct 2010 12:32:48 +0000 (13:32 +0100)]
Clear chromium side autofill profiles when java syncs a null profile.

This implements the native code needed to support the deletion
of a profile through the editor UI.

Change-Id: I7966551324d709612bff7cafeaf8a2cf93c218f0

13 years agoMerge "allow contained draws to identify a link"
Cary Clark [Thu, 28 Oct 2010 18:07:18 +0000 (11:07 -0700)]
Merge "allow contained draws to identify a link"

13 years agoallow contained draws to identify a link
Cary Clark [Thu, 28 Oct 2010 17:01:20 +0000 (13:01 -0400)]
allow contained draws to identify a link

The first cut of rewriting hiding and clipping links if
they are obscured by a subsequent draw looks for the
contained text to find the original link in the picture.

Some links contain only bitmaps, and others, like text
fields, may contain nothing at all. To identify these,
look for bitmaps, and to cover the text field case, look for
a background erase. These draws can be used to identify
where the link is in the picture.

Because bitmaps and background erases can also obscure
links, distinguish between draws that are contained by
the link from those that overlap it.

bug:3120589
Change-Id: Ied8fe2378f27253a162f5f9636f5ade12ce6013c

13 years agoAllow Gmail attachments to be viewed.
Iain Merrick [Thu, 28 Oct 2010 12:49:52 +0000 (13:49 +0100)]
Allow Gmail attachments to be viewed.

When displaying attachments, WebKit calls downloadFile() on an
Android-specific URL. This was crashing because it expected to
have a URLRequest available; however, there's a different code
path in WebRequest.cpp for Android URLs that doesn't create a
URLRequest.

This CL hoists the implementation of downloadFile() into
WebUrlLoaderClient, and gets the necessary headers from the
WebResponse object in didReceiveResponse (which should always
be called before downloadFile).

Gmail appends the MIME type to the URL after a ? separator,
and we have to extract this to display the file correctly.

Tests:
- Can display .txt attachments in Gmail (now fixed)
- Can download .apk files (checking for regressions)

Change-Id: I966220977972da7a2de29e663d7781b1de0e6fd4

13 years agoMerge "CookieManager.getCookie() should include HTTP-only cookies"
Steve Block [Thu, 28 Oct 2010 14:48:45 +0000 (07:48 -0700)]
Merge "CookieManager.getCookie() should include HTTP-only cookies"

13 years agoMerge "Cleaning up BackedDoubleBufferedTexture."
Derek Sollenberger [Thu, 28 Oct 2010 13:17:21 +0000 (06:17 -0700)]
Merge "Cleaning up BackedDoubleBufferedTexture."

13 years agoUse Tasks for WebRequest -> WebUrlLoaderClient callbacks.
Iain Merrick [Thu, 28 Oct 2010 09:39:47 +0000 (10:39 +0100)]
Use Tasks for WebRequest -> WebUrlLoaderClient callbacks.

We were using some hand-rolled boilerplate for inter-thread communication,
with all data packed into the same LoaderData struct. This CL uses the
Chromium utility function NewRunnableMethod to pack up the parameters, so
the callback methods don't have to be static and can have different
parameters. Also using smart pointers wherever possible.

Testing: manually verified that both synchronous and asynchronous requests
work correctly with no leaks. Will also do stress-testing to check for
race conditions.

Change-Id: I934a2ee795138f8eee43803a94bb7494ee73031d

13 years agoCookieManager.getCookie() should include HTTP-only cookies
Steve Block [Thu, 28 Oct 2010 12:48:16 +0000 (13:48 +0100)]
CookieManager.getCookie() should include HTTP-only cookies

This matches the behaviour of the Android Java HTTP stack

Change-Id: I280e524a849dc93bae54795543b1440c3be98c27

13 years agoMake sure we delete local references to jclass objects
Steve Block [Wed, 27 Oct 2010 14:47:41 +0000 (15:47 +0100)]
Make sure we delete local references to jclass objects

Change-Id: I1b398bd0a84ce366eb2e686e5f14335d7aa5ed31

13 years agoFix for split text bug.
Nicolas Roard [Wed, 27 Oct 2010 21:24:10 +0000 (14:24 -0700)]
Fix for split text bug.

Bug:3134026
Change-Id: I0ccaf27b8ede58603aeb2284c56527d4fe0a589b
(cherry picked from commit ca3a75fc3f9712c15e4a84ebcfd48699963896bb)

13 years agoMerge "Fix a missing namespace in WebCache"
Steve Block [Wed, 27 Oct 2010 20:33:41 +0000 (13:33 -0700)]
Merge "Fix a missing namespace in WebCache"

13 years agoFix a missing namespace in WebCache
Steve Block [Wed, 27 Oct 2010 20:24:20 +0000 (21:24 +0100)]
Fix a missing namespace in WebCache

Change-Id: I6d1bf42b3f5a81b24613ca4c62ba4e02400a63b0

13 years agoMerge "Hook up CookieManager.getCookie() for Chromium HTTP stack"
Steve Block [Wed, 27 Oct 2010 19:29:18 +0000 (12:29 -0700)]
Merge "Hook up CookieManager.getCookie() for Chromium HTTP stack"

13 years agoHook up CookieManager.getCookie() for Chromium HTTP stack
Steve Block [Wed, 27 Oct 2010 15:59:53 +0000 (16:59 +0100)]
Hook up CookieManager.getCookie() for Chromium HTTP stack

Requires a change to frameworks/base ...
https://android-git.corp.google.com/g/76605

Bug: 3116410
Change-Id: I1b66c1ac9f9aaf5388ad1da92f54b1f16d4a3626

13 years agoMerge "Hook up CookieManager.acceptCookie() and setAcceptCookie() for the Chromium...
Steve Block [Wed, 27 Oct 2010 18:47:48 +0000 (11:47 -0700)]
Merge "Hook up CookieManager.acceptCookie() and setAcceptCookie() for the Chromium HTTP stack"

13 years agoHook up CookieManager.acceptCookie() and setAcceptCookie() for the Chromium HTTP...
Steve Block [Mon, 25 Oct 2010 11:20:52 +0000 (12:20 +0100)]
Hook up CookieManager.acceptCookie() and setAcceptCookie() for the Chromium HTTP stack

We also update PlatformBridge::cookiesEnabled() to query the Chromium
HTTP stack directly. This avoids calling
CookieClient::cookiesEnabled(), which calls the Java
CookieManager::acceptCookie() which in turns calls back to native
code.

Also requires a change to frameworks/base ...
https://android-git.corp.google.com/g/76065

Bug: 3116410
Change-Id: Id853463f3bcef76b220e8c44dd2b30c0d6752624

13 years agoMerge "Fix some problems in WebCoreFrameBridge when using JNIEnv::NewString()"
Steve Block [Wed, 27 Oct 2010 18:44:05 +0000 (11:44 -0700)]
Merge "Fix some problems in WebCoreFrameBridge when using JNIEnv::NewString()"

13 years agoMerge "Update playback state after full screen video return."
Shimeng (Simon) Wang [Wed, 27 Oct 2010 17:39:33 +0000 (10:39 -0700)]
Merge "Update playback state after full screen video return."

13 years agoUpdate playback state after full screen video return.
Shimeng (Simon) Wang [Tue, 26 Oct 2010 21:10:27 +0000 (14:10 -0700)]
Update playback state after full screen video return.

This change will correctly update the play/pause button when
a full screen video is returned in the middle.

issue:3102273
Change-Id: Icf5e9194393fa30dd6d9195045df21480842aa6b

13 years agoMerge "Fix for simulator build"
Kristian Monsen [Wed, 27 Oct 2010 17:05:22 +0000 (10:05 -0700)]
Merge "Fix for simulator build"

13 years agoFix for simulator build
Kristian Monsen [Wed, 27 Oct 2010 16:47:28 +0000 (17:47 +0100)]
Fix for simulator build

Added include in non-upstreamed section

Change-Id: I16556b1e46b43c509de04078487f13af2a9a3b72

13 years agoFix some problems in WebCoreFrameBridge when using JNIEnv::NewString()
Steve Block [Wed, 27 Oct 2010 15:47:51 +0000 (16:47 +0100)]
Fix some problems in WebCoreFrameBridge when using JNIEnv::NewString()

- Remove superfluous casts to unsigned short*
- Avoid converting from Unicode to UTF8 and back

Change-Id: Ib7cbc610f1b27ecd52082bf412afa7c7b3fec4b2

13 years agorewrite select text and others for layers
Cary Clark [Tue, 26 Oct 2010 14:56:53 +0000 (10:56 -0400)]
rewrite select text and others for layers

Layers contain pictures, and draw them offset from the top of
the page. Several readers of pictures need to account for this
displacement when computing what part of the picture intersects
a tap on the screen.

The tap may not correspond to the first layer that intersects
it, so all layers must be checked to find the best match. The
root layer usually draws everywhere, so for a match to correspond
to the root, the match must additionally intersect text.

Layers may create offscreen bitmaps when drawing to correctly
alpha blend the results to the screen, but this causes the items
in the bitmap to draw to an unexpected location when the picture
is treated as a spatial database. To get around this, call the
SkCanvas::save() from the overridden saveLayer() to push and
pop the canvas layer state without creating an offscreen.

WebCore/platform/graphics/android/LayerAndroid.cpp
WebCore/platform/graphics/android/LayerAndroid.h
- In find(), iterate through all children, instead of stopping
  on the first match.
- Check to see if the child actually draws at the desired location,
  and if it draws text there as well.
- Specify a slop factor to allow for inaccuracies in touch.
- Check the root for text before checking the children.

WebKit/android/nav/CachedFrame.cpp
WebKit/android/nav/CachedFrame.h
- Modify the (x,y) co-ordinate by the layer's offset, when
  finding the picture corresponding to a point.

WebKit/android/nav/CachedLayer.cpp
WebKit/android/nav/CachedLayer.h
- More plumbing to adjust the point if the picture is contained
  in an offset layer.

WebKit/android/nav/CachedRoot.cpp
WebKit/android/nav/CachedRoot.h
- Correct the (x,y) locations by the layer offset.
- Add some debugging (disabled by default)

WebKit/android/nav/ParsedCanvas.h
- One stop shopping that calls save() from saveLayer().
- Reset the bounder to null to balance its ref count.

WebKit/android/nav/SelectText.cpp
WebKit/android/nav/SelectText.h
- Rearrange the way pictures are tracked. Record the picture
  corresponding to the input location when the selection starts,
  requiring that the picture remain unchanged as the selection
  extends.
- Only draw adornments for when the corresponding picture is
  drawn. This fixes a Gmail specific problem, where the layers
  come and go as the page scrolls.
- Always use the supplied visible bounds instead of computing
  it from the canvas.
- Correct location points by layer offsets.
- Add to the picture ref count so it can't be deleted during
  selection.

WebKit/android/nav/WebView.cpp
- Simplify visibleRect code.
- Simplify all SelectText interfaces.

bug:3114609
Change-Id: I43dc3252fc86c4b6500edcd650126b2559f530e3

13 years agoCleaning up BackedDoubleBufferedTexture.
Derek Sollenberger [Wed, 27 Oct 2010 14:56:34 +0000 (10:56 -0400)]
Cleaning up BackedDoubleBufferedTexture.

The cleanup consisted of removing unused fields, ensuring mutexes
are only used for cross-thread interactions, and adding comments.

Change-Id: Ie008cd8f803299d418439793f21feb90d8325a8e

13 years agoUpdate Chromium HTTP stack to get database and cache paths from CookieSyncManager
Steve Block [Wed, 27 Oct 2010 10:29:59 +0000 (11:29 +0100)]
Update Chromium HTTP stack to get database and cache paths from CookieSyncManager

Currently these paths are obtained from the BrowserFrame via JNI.
However, the paths are not synced to the BrowserFrame until the
WebCore thread has started up. This means that if the
WebRequestContext is created on the IO thread, the paths may not be
available.

This change moves the paths from BrowserFrame to CookieSyncManager to
avoid the need to wait for the WebCore thread.

Note that the new methods may be called on either the UI or WebCore
threads, so are synchronised.

Requires a change to frameworks/base ...
https://android-git.corp.google.com/g/76580

Change-Id: I61f88772d13901d66892118c79db3cac412e6bd1

13 years agoFix WebRequestContext::get() to be threadsafe
Steve Block [Wed, 27 Oct 2010 08:58:13 +0000 (09:58 +0100)]
Fix WebRequestContext::get() to be threadsafe

This allows the method to be called from both the WebCore thread and
the UI thread, as required to hook up CookieManager.

Change-Id: I895d14909b14f0baba77fb2cb75926d9103ee01f

13 years agoFixes style in WebRequestContext
Steve Block [Wed, 27 Oct 2010 08:49:32 +0000 (09:49 +0100)]
Fixes style in WebRequestContext

Previously this file used Chromium style, as it extends a Chromium
class. However, I think it's best to switch to WebKit style to prevent
Chromium style from leaking into the rest of WebKit.

Also switches static member methods to file-scope static functions
where possible.

Finally, return WebRequestContext* rather than URLRequestContext* to
allow us to call methods in the derived class. this will be required
to hook up CookieManager.

Change-Id: Ifdca62230a3728e338904ef1b9b392640af06a92

13 years agoMerge "Cleanup the getAvailableTexture() function."
Derek Sollenberger [Wed, 27 Oct 2010 12:54:47 +0000 (05:54 -0700)]
Merge "Cleanup the getAvailableTexture() function."

13 years agoFix build break.
Iain Merrick [Wed, 27 Oct 2010 11:23:51 +0000 (12:23 +0100)]
Fix build break.

Missing edit in I50da7599, ended up on the wrong CL.

Change-Id: I4d4c6713fbfe30cdc8cdf0e733c4725b33a09673

13 years agoOnly AV formats should be handled by MediaPlayer (C++ side)
Iain Merrick [Wed, 27 Oct 2010 09:59:48 +0000 (10:59 +0100)]
Only AV formats should be handled by MediaPlayer (C++ side)

Fixing JNI bindings. See Java-side change for details (Id39b770b).

Change-Id: I50da7599475476ebaaf9609f44e13fbdf212c90e

13 years agoCleanup the getAvailableTexture() function.
Derek Sollenberger [Tue, 26 Oct 2010 17:42:04 +0000 (13:42 -0400)]
Cleanup the getAvailableTexture() function.

We now check to see if the tile owns a texture and immediately
return if that is the case.  I also updated the documentation.

Change-Id: I88a6f2e38ef33fd37f5926c793c9e775e646bef2

13 years agoRemove locks in BaseTile. Also fix the bug where the browser hangs when closing a...
Nicolas Roard [Fri, 22 Oct 2010 00:02:16 +0000 (17:02 -0700)]
Remove locks in BaseTile. Also fix the bug where the browser hangs when closing a tab.

Bug:3112610
Change-Id: I4f4854a8a0688fc24b512e7671a2ad0708121075

13 years agoFix random crashes when HW acceleration is turned on.
Nicolas Roard [Tue, 19 Oct 2010 23:34:10 +0000 (16:34 -0700)]
Fix random crashes when HW acceleration is turned on.

Bug:3107362

Change-Id: I354a07369056e696deed7458a4f4e14d54b7f6c8

13 years agoMerge "Fix potential crash in Nav Cache code introduced by WebKit merge"
Ben Murdoch [Tue, 26 Oct 2010 18:21:04 +0000 (11:21 -0700)]
Merge "Fix potential crash in Nav Cache code introduced by WebKit merge"

13 years agoMerge "libwebcore cleanspec."
Steve Block [Tue, 26 Oct 2010 17:55:48 +0000 (10:55 -0700)]
Merge "libwebcore cleanspec."

13 years agolibwebcore cleanspec.
Ben Murdoch [Tue, 26 Oct 2010 17:51:19 +0000 (18:51 +0100)]
libwebcore cleanspec.

Reports of build errors in master, fixed by a clean build.

Change-Id: Idfee8a890cc69f4b69b1d72738525c43770e65b8

13 years agoFix potential crash in Nav Cache code introduced by WebKit merge
Ben Murdoch [Tue, 26 Oct 2010 17:27:56 +0000 (18:27 +0100)]
Fix potential crash in Nav Cache code introduced by WebKit merge

The refactoring in the nav cache code caused by the recent merge
to webkit r70209 can crash. Fix that.

Bug: 3132882
Change-Id: I98b605ae2fdf24cf0f25adc919f2dbba581abef0

13 years agoAdd DEBUG_COUNT stuff
Nicolas Roard [Mon, 25 Oct 2010 08:24:41 +0000 (01:24 -0700)]
Add DEBUG_COUNT stuff

Change-Id: I99a0d2bc23da6299228ba5adea7cf161e01d8fbf
(cherry picked from commit 8f12379d1df1bc5f2e85bce687d89ba20d4a5e9e)

13 years agoUpdate selection after keypresses.
Leon Scroggins [Fri, 22 Oct 2010 22:16:01 +0000 (18:16 -0400)]
Update selection after keypresses.

Bug:3054784
Change-Id: I5430969dc532feea10a960eb4117b3a3108ed702

13 years agoFix a layering violation in Document.
Ben Murdoch [Mon, 25 Oct 2010 18:10:19 +0000 (19:10 +0100)]
Fix a layering violation in Document.

PlatformBridge already has the necessary functionality.

Change-Id: I67a5d68890d3976da69cff0c20f7b48931366b1d

13 years agoMerge WebKit at r70209: Update WebKit revision
Ben Murdoch [Fri, 22 Oct 2010 12:02:34 +0000 (13:02 +0100)]
Merge WebKit at r70209: Update WebKit revision

Change-Id: I9a95c19b76015bfb43393e0570bd35f155dc7809

13 years agoMerge Webkit at r70209: Add Android expected results for new layout tests
Ben Murdoch [Mon, 25 Oct 2010 15:02:11 +0000 (16:02 +0100)]
Merge Webkit at r70209: Add Android expected results for new layout tests

fast/dom/DeviceOrientation/create-event-onorientationchange.html
See http://trac.webkit.org/changeset/69984

Change-Id: I345106587049a3e59040aa2c449432ac6d7fe5b1

13 years agoMerge WebKit at r70209: Cherry pick build fix
Ben Murdoch [Mon, 25 Oct 2010 17:27:58 +0000 (18:27 +0100)]
Merge WebKit at r70209: Cherry pick build fix

See https://bugs.webkit.org/show_bug.cgi?id=48243 for details.

Change-Id: Idf1bf291f2370e611ebb2595f820598f68925336

13 years agoMerge Webkit at r70209: defaultLanguage has been renamed.
Ben Murdoch [Fri, 22 Oct 2010 19:08:15 +0000 (20:08 +0100)]
Merge Webkit at r70209: defaultLanguage has been renamed.

It is now known as defaultPlatformLanguage.
See http://trac.webkit.org/changeset/70095

Change-Id: I523e93e78317a2e365429a8027967aa46d0370fa