OSDN Git Service
Andrei Popescu [Mon, 29 Mar 2010 16:43:47 +0000 (17:43 +0100)]
Make sure WebKit's page cache does not cache pages with unload handlers.
Fix b:
2552004
Change-Id: Iac6737fbb9a1eca734ea7474f99ccf66ee916bd7
Steve Block [Fri, 26 Mar 2010 11:15:43 +0000 (11:15 +0000)]
Cherry pick from master.
Cherry-picks WebKit change 56105 to eliminate use of bzero in WebKit common code
Note that manual changes were required to JNIUtilityPrivate.cpp as this file has
changed upstreamed, but the merge will be trivial.
See http://trac.webkit.org/changeset/56105
Bug:
2535696
Change-Id: I464a4aa78b1625961724b6001caba1c315ed9f91
Andrei Popescu [Mon, 29 Mar 2010 11:08:36 +0000 (12:08 +0100)]
Cherry pick patch in https://bugs.webkit.org/show_bug.cgi?id=36665
which adds V8 bindings for page cache.
Fix b:
2533219
Change-Id: I57f067adbbef76b4f8ec6c50b9e85fb2b7fc619a
Leon Scroggins III [Thu, 25 Mar 2010 17:01:13 +0000 (13:01 -0400)]
Remove a warning.
Change-Id: Iffd11a306df9a5140b3406e825185713d17d273e
Cary Clark [Thu, 25 Mar 2010 14:18:35 +0000 (10:18 -0400)]
don't track layer node twice
If a cache node is both in a layer and contains a layer, don't
add it the second time.
Change-Id: I45cc302e234565bfbc87649ca503f515b1f487e8
http://b/
2543736
Dan Egnor [Wed, 24 Mar 2010 22:32:36 +0000 (15:32 -0700)]
Apparently this needs to happen again (at least, it did for my build).
Change-Id: I89dbec1be74da68afb2a03b75ad5bf87426d8889
Nicolas Roard [Wed, 24 Mar 2010 20:41:14 +0000 (13:41 -0700)]
Merge "Mark positioned elements as needing a relayout when checking for hit test. This is necessary when using composited fixed layers, as the layers position may have moved UI-side, while the webkit's related position of those layers may not have been updated yet."
Nicolas Roard [Wed, 24 Mar 2010 18:27:42 +0000 (18:27 +0000)]
Mark positioned elements as needing a relayout when checking for hit test.
This is necessary when using composited fixed layers, as the layers position
may have moved UI-side, while the webkit's related position of those layers
may not have been updated yet.
Fix Bug:
2457215
Change-Id: If4f0e9c3d1a4786b29dcc7c2dc3510ba090b6a36
Kristian Monsen [Wed, 24 Mar 2010 19:35:36 +0000 (12:35 -0700)]
Merge "compile fix for V8 with debug builds"
Kristian Monsen [Wed, 24 Mar 2010 19:26:22 +0000 (19:26 +0000)]
compile fix for V8 with debug builds
Change-Id: Icffc7b0d5f9a9b2da66dca37c0c8d084c11491b7
Steve Block [Wed, 24 Mar 2010 18:28:50 +0000 (11:28 -0700)]
Merge "Fixes XHR requests to use the username and password supplied from JavaScript"
Patrick Scott [Wed, 24 Mar 2010 18:18:50 +0000 (14:18 -0400)]
Fix the background of the placeholder.
Draw a light gray 1 pixel border instead of using an asset.
Bug:
2411524
Change-Id: Ibc95148e2bf13e0dcc58d747e3733cc6993dddf6
Steve Block [Wed, 24 Mar 2010 16:27:17 +0000 (16:27 +0000)]
Fixes XHR requests to use the username and password supplied from JavaScript
Currently, the browser does not use the credentials supplied from JavaScript.
If a request returns a 401 Unauthorized, the browser always prompts the user.
This violates http://www.w3.org/TR/XMLHttpRequest/#the-send-method
Bug:
2533522
Change-Id: Ic232003a3f321167810fc7263d6ecb654eeeb5eb
Mike Reed [Wed, 24 Mar 2010 17:45:41 +0000 (10:45 -0700)]
Merge "drawRect is defined (in header) to always stroke with 1.0"
Mike Reed [Wed, 24 Mar 2010 17:33:09 +0000 (13:33 -0400)]
drawRect is defined (in header) to always stroke with 1.0
Change-Id: I95bef5b769c9b2637424ae4f22ccd19ed6e4ff9a
Nicolas Roard [Wed, 24 Mar 2010 17:22:54 +0000 (10:22 -0700)]
Merge "Renders fixed layers with the root canvas matrix. Fix some positioning issues."
Nicolas Roard [Fri, 19 Mar 2010 17:10:50 +0000 (17:10 +0000)]
Renders fixed layers with the root canvas matrix. Fix some positioning issues.
Bug:
2526966 Bug:
1818168
The current rendering code exposes some issues with the fact that we have
fixed layers in the layers hierarchy -- parents transformations are also applied
to the fixed layers, which is not what we want (fixed layers should be applied
on the original canvas, with the original transform -- e.g. toolbar present or not --
but no more).
One previously discussed solution was to move the fixed layers to their own hierarchy;
but doing so would mean to also redo all the z-index management that we already have
in the current system. The simplest way is therefore to use the original matrix (the
canvas' matrix) when we have a fixed layer. The way we do this is by inserting a new
LayerAndroid before the LayerAndroid root, setting the matrix of that new root to be
the canvas' matrix. The drawing is then unaffected, but we can ask skia to draw using
the root's matrix.
The second issue solved in the CL is some positioning troubles; layers may have
different dimensions than their render view, and the previous code was considering that
the views were always drawn at the origin in the layer. By removing the parents layers
transforms, this is not the case anymore, and we therefore need to take the render view
offset into account. Finally there is some additional debug code in LayerAndroid.
Change-Id: Id353ad3dfd9808252643f0e4f0140dde67480719
Grace Kloba [Wed, 24 Mar 2010 16:10:46 +0000 (09:10 -0700)]
Merge "Add a special case to handle "wap" style html."
Patrick Scott [Wed, 24 Mar 2010 13:50:27 +0000 (09:50 -0400)]
Do not redirect data to our placeholder.
Bug:
2539312
Change-Id: I952b854cc65c5481e9fed04d2169b0f01536d5a4
Grace Kloba [Tue, 23 Mar 2010 23:14:34 +0000 (16:14 -0700)]
Add a special case to handle "wap" style html.
If we get here, it is html mimetype, not wap mimetype.
Treat the "wap" style like "mobile" style by using
device-width as viewport width.
Fix http://b/issue?id=
2538320
Patrick Scott [Tue, 23 Mar 2010 20:36:20 +0000 (13:36 -0700)]
Merge "Add on-demand plugin support."
Patrick Scott [Tue, 23 Mar 2010 20:30:18 +0000 (13:30 -0700)]
Merge "Revert "Revive my change to not add anchor navigations to the history in some cases.""
Patrick Scott [Tue, 23 Mar 2010 20:27:00 +0000 (16:27 -0400)]
Revert "Revive my change to not add anchor navigations to the history in some cases."
This reverts commit
a6e5d463b2fbb6ad4e154719d6846231a32764a7.
Patrick Scott [Mon, 22 Mar 2010 14:09:16 +0000 (10:09 -0400)]
Add on-demand plugin support.
The Settings object now has an on-demand flag for plugins (this was to avoid
more edits to webkit code). If plugins are on-demand and a plugin is installed
that can handle the content, insert a placeholder widget. If the user clicks on
the placeholder, the plugin will be enabled.
The widget currently does not clip the context correctly. It only clips based on
the widget frame. This is due to a bug (already filed) where the scroll offset
is producing bad clip rectangles.
Requires a framework change.
Bug:
2411524
Change-Id: If3931da8da2339a2385ae78b609c49fa069892ab
Derek Sollenberger [Tue, 23 Mar 2010 19:25:14 +0000 (12:25 -0700)]
Merge "Allowing native code access to the application context as long as there is at least one valid instance of the webviewcore class."
Derek Sollenberger [Tue, 23 Mar 2010 18:07:05 +0000 (14:07 -0400)]
Allowing native code access to the application context as long as
there is at least one valid instance of the webviewcore class.
This change allows plugins access to the application context when
the plugin package is loaded, prior to any instances of the plugin
being created.
Change-Id: I74a58cf76cc3c2d0ef7bca19346f13f58f616830
Kristian Monsen [Tue, 23 Mar 2010 16:37:55 +0000 (09:37 -0700)]
Merge "Only replaced flash content if youtube app is installed. Fix for bug
2521517."
Kristian Monsen [Tue, 23 Mar 2010 10:17:19 +0000 (10:17 +0000)]
Only replaced flash content if youtube app is installed. Fix for bug
2521517.
Change-Id: Id1b2401728d03410a10bac6674450f5123b4785e
Daisuke Miyakawa [Tue, 23 Mar 2010 02:07:30 +0000 (11:07 +0900)]
Use EmojiFoctory::GetShiftJisConverterName() instead of relying on static string.
This change should be with the change Iccbef288.
Bug:
2327318
Change-Id: Ie67d6c1ef60d8e57784e903e3e00a73d3bec0f82
Nicolas Roard [Mon, 22 Mar 2010 22:18:04 +0000 (15:18 -0700)]
Merge "Only enable composited fixed elements for mobile sites."
Nicolas Roard [Fri, 19 Mar 2010 11:51:19 +0000 (11:51 +0000)]
Only enable composited fixed elements for mobile sites.
We can consider a website as being a 'mobile' site if:
- the viewport width is either undefined (-1) or equal to device-width (0)
- no scaling is allowed
Change-Id: Id88257278312077a170eb10d9f666c46bad135b6
Ben Murdoch [Fri, 12 Mar 2010 20:46:03 +0000 (20:46 +0000)]
Add a new API on the resource loader to pause the current load. We use this when the plugin cannot accept any more data to prevent WebKit
buffering the data that the plugin can't take. This can result in a crash if the buffer gets too big (when watching an HD movie clip in
flash, for example).
Requires a frameworks/base change.
Change-Id: Ibb63d38df1014ff70fa4d65275613cbf3a217c1a
Steve Block [Mon, 22 Mar 2010 10:44:14 +0000 (03:44 -0700)]
Merge "Implements layoutTestController.setAppCacheMaximumSize"
Mike Reed [Fri, 19 Mar 2010 17:26:35 +0000 (13:26 -0400)]
use picture recording flag to speed up clipPath (and avoid asserts in very large edgelists)
override (and no-op) clipPath() in IsEmptyCanvas, same reasons as above.
Change-Id: I70b1565a11c1e314fdde507c1b2118b9aba651e6
Nicolas Roard [Fri, 19 Mar 2010 17:10:50 +0000 (17:10 +0000)]
Add back a missing ifdef
Change-Id: Ib0f0cf689e95b7333045df53ca645d34a0db21a4
Steve Block [Fri, 19 Mar 2010 15:31:42 +0000 (15:31 +0000)]
Implements layoutTestController.setAppCacheMaximumSize
Change-Id: I2cf3d3135fbd21158525866af458eb89f902fb66
Brian Carlstrom [Fri, 19 Mar 2010 00:46:38 +0000 (17:46 -0700)]
Merge "Fix comment from "WebView.LOAD_*" to "WebSettings.LOAD_*""
Brian Carlstrom [Fri, 19 Mar 2010 00:06:20 +0000 (17:06 -0700)]
Fix comment from "WebView.LOAD_*" to "WebSettings.LOAD_*"
While improving the javadoc for
android.webkit.BrowserFrame.startLoadingResource, I noticed that the
Java class metioned in the C++ comments was wrong.
Change-Id: Ibf6234735fc6b92b444f6cf3b4c6bb5c33675525
Shimeng (Simon) Wang [Thu, 18 Mar 2010 18:17:50 +0000 (11:17 -0700)]
Merge "Fill in appropriate referer URL plugin initiated resource loading."
Shimeng (Simon) Wang [Thu, 18 Mar 2010 17:13:46 +0000 (10:13 -0700)]
Fill in appropriate referer URL plugin initiated resource loading.
Didn't think of better way to code in the common code.
Bug:
2512527
Change-Id: Ie656ce1c399f238187045be41ece1fc2feefab90
Leon Scroggins [Thu, 18 Mar 2010 17:20:41 +0000 (10:20 -0700)]
Merge "Pass selection data along with message to open the keyboard."
Nicolas Roard [Thu, 18 Mar 2010 16:57:34 +0000 (09:57 -0700)]
Merge "Implement margin for fixed elements. Last missing piece to fix Bug:
2252505"
Leon Scroggins [Thu, 18 Mar 2010 16:43:35 +0000 (12:43 -0400)]
Pass selection data along with message to open the keyboard.
Fixes http://b/issue?id=
2431351
Requires a change to frameworks/base
Change-Id: I7f703746bdd00aaae2607a6a41ac4916f43d37b1
Nicolas Roard [Thu, 18 Mar 2010 16:35:27 +0000 (16:35 +0000)]
Implement margin for fixed elements.
Last missing piece to fix Bug:
2252505
Billmonk.com uses a centered fixed element with negative margins,
and we were not accounting for those...
Accounting for the margins move the fixed layer at the same place
than webkit think it is, and therefore clicks do work.
Change-Id: I6a736d06273df7d2bbc597c17ce042aa8b75b4e2
Steve Block [Thu, 18 Mar 2010 15:55:23 +0000 (08:55 -0700)]
Merge "Adds Android-specific results for an appcache layout test"
Nicolas Roard [Thu, 18 Mar 2010 15:35:01 +0000 (08:35 -0700)]
Merge "Fix click issues when using fixed elements. This CL also fix the positioning of a fixed layer when no position is defined, and also only use the screen to position only the fixed elements, not other types of positioned elements."
Steve Block [Thu, 18 Mar 2010 15:12:02 +0000 (15:12 +0000)]
Adds Android-specific results for an appcache layout test
Currently this test fails only because of textual differences in the console
message produced on Android.
Change-Id: I6f2e75b1bad9f3e8d7636d5f2b59b2d4b22fd779
Nicolas Roard [Wed, 17 Mar 2010 21:15:55 +0000 (21:15 +0000)]
Fix click issues when using fixed elements. This CL also fix the positioning
of a fixed layer when no position is defined, and also only use the screen
to position only the fixed elements, not other types of positioned elements.
Bug:
2521087
The click issues were due to not returning the fixed element when looking
for a parent stackingContext in RenderLayer::stackingContext().
This resulted in incorrect coordinates for the layers children of a
fixed layer, that we then had to recompute in
RenderLayer::convertToLayerCoords(), but this in turns was invaliding
hit test detection...
Fixed elements are now positioned relative to the screen instead of the
virtual viewport (ANDROID_FIXED_ELEMENTS); but this was applying
indiscriminantly to all positioned elements, absolute elements included.
The CL modify RenderBox::containingBlockWidthForPositioned() and
RenderBox::containingBlockHeightForPositioned() to only do this for
fixed elements.
Finally, fixed layers were wrongly positioned if the positions were not
fully set (e.g. only setting top:0 but no left or right). The change to
LayerAndroid::updateFixedLayersPositions() fixes this.
Change-Id: I07a179dd631a2bc1a313e33ffcf69ef388ecb7ca
Cary Clark [Thu, 18 Mar 2010 15:01:14 +0000 (11:01 -0400)]
window may not be initialized when plugin is first called
Change-Id: I37ba49b9cbd5170613f0d8e694aa79b49585a8ee
http://b/
2524032
Cary Clark [Wed, 17 Mar 2010 19:55:24 +0000 (15:55 -0400)]
get bounds of focused node
companion fix in framework/base
Change-Id: I7d9191ea9c95dafac7f7e91a70d02c7d055c6967
http://b/
2521087
Steve Block [Wed, 17 Mar 2010 19:22:09 +0000 (12:22 -0700)]
Merge "Fixes the Mac build"
Steve Block [Wed, 17 Mar 2010 18:23:16 +0000 (18:23 +0000)]
Fixes the Mac build
- JavaInstanceJSC.cpp - Adds ANDROID guards around some temporary logging
- Geolocation.cpp - Adds ANDROID guards around changes waiting to be upstreamed
- WebChromeClient.h / EmptyClients.h - Pulls the Mac part of a WebKit change for which only the Android part was cherry-picked in https://android-git.corp.google.com/g/#change,41747
Change-Id: Ide97cb6b05e78528fe4b9430466837dedaadc4cf
Patrick Scott [Wed, 17 Mar 2010 18:39:25 +0000 (11:39 -0700)]
Merge "Compare the client rather than the handle as the handle may not be set yet."
Patrick Scott [Wed, 17 Mar 2010 18:33:58 +0000 (14:33 -0400)]
Compare the client rather than the handle as the handle may not be set yet.
Bug:
2256744
Change-Id: I45baf86d56192c8752cb34f33795d2fa015ebe26
Steve Block [Wed, 17 Mar 2010 18:09:57 +0000 (11:09 -0700)]
Merge "Disable workers"
Steve Block [Wed, 17 Mar 2010 14:37:19 +0000 (14:37 +0000)]
Disable workers
This is because V8 on Android does not have the required locking.
Also disables channel messaging, which is used only with workers.
Bug:
2522239
Change-Id: I6cb91b4048c7e1a0351e422561625397a2e98986
Shimeng (Simon) Wang [Wed, 17 Mar 2010 18:05:54 +0000 (11:05 -0700)]
Merge "Fix a webkit bug."
Steve Block [Wed, 17 Mar 2010 17:49:56 +0000 (10:49 -0700)]
Merge changes I23ce1e5a,Ieee214f4
* changes:
Add missing WORKERS guards to V8 bindings
Cherry-pick WebKit change 56112 to add WORKERS guards to V8 bindings
Shimeng (Simon) Wang [Tue, 16 Mar 2010 18:32:31 +0000 (11:32 -0700)]
Fix a webkit bug.
Change-Id: Ib8ae67e76c03027fcece73f2d36fc0ea41164e40
Steve Block [Wed, 17 Mar 2010 17:47:17 +0000 (10:47 -0700)]
Merge changes If72cb478,Ie6365073
* changes:
Cherry-pick WebKit change 55084 to add SHARED_WORKERS guard in V8 globalObjectPrototypeIsDOMWindow
Cherry-pick WebKit change 54972 to merge DOMWindow and WorkerContext object wrapping in V8 bindings
Nicolas Roard [Wed, 17 Mar 2010 17:11:14 +0000 (10:11 -0700)]
Merge "Refactor how we set up the layers hierarchy when using fixed composited layers, and fix the z-index position."
Kristian Monsen [Wed, 17 Mar 2010 15:49:12 +0000 (08:49 -0700)]
Merge "Fix for Bug
2486292, transparent backgrounds not transparent in Froyo."
Grace Kloba [Wed, 17 Mar 2010 15:40:44 +0000 (08:40 -0700)]
Merge "Pass mouse events to the subframe as the other ports by calling subframe handler."
Steve Block [Wed, 17 Mar 2010 15:02:30 +0000 (15:02 +0000)]
Add missing WORKERS guards to V8 bindings
The motivation for this change is to allow us to disable workers on Android,
because V8 on Android does not have the required locking.
This change will not be upstreamed because the code has been refactored in
tip-of-tree WebKit.
See http://trac.webkit.org/changeset/55096
Change-Id: I23ce1e5aecc2897da240b5229585b8e6617d8ebf
Steve Block [Wed, 17 Mar 2010 15:01:37 +0000 (15:01 +0000)]
Cherry-pick WebKit change 56112 to add WORKERS guards to V8 bindings
The motivation for this change is to allow us to disable workers on Android,
because V8 on Android does not have the required locking.
See http://trac.webkit.org/changeset/56112
Change-Id: Ieee214f4765af3188bbdc51c56f14883f6a6779c
Steve Block [Wed, 17 Mar 2010 14:36:17 +0000 (14:36 +0000)]
Cherry-pick WebKit change 55084 to add SHARED_WORKERS guard in V8 globalObjectPrototypeIsDOMWindow
The motivation for this change is to allow us to disable workers on Android,
because V8 on Android does not have the required locking.
See http://trac.webkit.org/changeset/55084
Change-Id: If72cb478fb7dad6be44eaa2c1ca19e9790a9dda0
Steve Block [Wed, 17 Mar 2010 15:25:43 +0000 (15:25 +0000)]
Cherry-pick WebKit change 54972 to merge DOMWindow and WorkerContext object wrapping in V8 bindings
The motivation for this change is to disable workers on Android, because V8 on
Android does not have the required locking.
The current version of WebKit in Android (54731) uses workers features without
ENABLE(WORKERS) guards in the V8 bindings. In particular,
V8DOMWrapper::instantiateV8ObjectInWorkerContext is used from generated code
without guards.
Fixing the code generator would be difficult and would be wasted effort as
V8DOMWrapper::instantiateV8ObjectInWorkerContext no longer exists in tip-of-tree
WebKit. Instead, we cherry-pick the WebKit change which removes this method.
See http://trac.webkit.org/changeset/54972
Change-Id: Ie6365073d7a4d92aa7c1553d87a7e1c1ed514118
Nicolas Roard [Tue, 16 Mar 2010 15:35:12 +0000 (15:35 +0000)]
Refactor how we set up the layers hierarchy when using fixed
composited layers, and fix the z-index position.
Bug:
2497910 Bug:
2450006
We add a new COMPOSITED_FIXED_ELEMENTS define to isolate the
changes in the webkit common code.
We previously had a problem where the hierarchy of GraphicsLayer
(i.e. the backed surfaces associated to the composited RenderLayer)
was not reflecting that layers were children of a fixed layer.
The workaround we currently have is not fully satisfactory, due to the way
we draw layers on screen (in some cases layers were wrongly translated,
see Bug:
2497910).
Instead, modifying the webkit common code simplify things a lot, and makes
the patch more likely to be upstreamed to webkit, as it's now a
reasonably well-delimited feature (use composited layers for fixed elements).
What we do now is to consider fixed elements as a stacking context, which
makes all layers children of such elements children too in the
GraphicsLayer hierarchy, and modifying the offset of those children
accordingly (in RenderLayer.cpp).
In addition, we fixes the z-index bugs we had
by signaling that there is a fixed element to its siblings, and turning
the siblings as composited layers as well (so that the ordering works
fully UI-side).
Change-Id: I735c6c14d955ef54653f0053187d3495bef1f332
Kristian Monsen [Mon, 15 Mar 2010 15:41:48 +0000 (15:41 +0000)]
Fix for Bug
2486292, transparent backgrounds not transparent in Froyo.
Recreating the transparent and color of old window when creating a new one, and setting transparent if all the colors are 0.
Updated patch to fix comments from Grace.
Change-Id: I934401630fd7041e7e4ae1cbeeaa198271025c87
Grace Kloba [Wed, 17 Mar 2010 04:10:18 +0000 (21:10 -0700)]
Pass mouse events to the subframe as the other ports
by calling subframe handler.
Most of the time, nav cache found the iframe and calls
the correct eventhandler. But when nav cache fails,
we use the top frame and we need to pass the mouse
events to the subframe.
http://b/issue?id=
2521087
Nicolas Roard [Wed, 17 Mar 2010 03:34:40 +0000 (20:34 -0700)]
Merge "Compute the position of the fixed elements to be relative to the screen and not the virtual viewport."
Nicolas Roard [Wed, 17 Mar 2010 02:06:20 +0000 (02:06 +0000)]
Compute the position of the fixed elements to be relative to the
screen and not the virtual viewport.
This fixes Bug:
2515587 and Bug:
2457215
Change-Id: Ib987787cf9bf74b760b857dbbb397f85af334dce
Grace Kloba [Sat, 13 Mar 2010 02:45:47 +0000 (18:45 -0800)]
To enable per file log, only need to define PLUGIN_DEBUG_LOCAL
Derek Sollenberger [Tue, 16 Mar 2010 20:29:40 +0000 (16:29 -0400)]
Ensure a plugin is not shown if it is completely obscured by other
html content.
Change-Id: Iffcc30d47428708842ac5dddee41cc611d879c1c
Ben Murdoch [Tue, 16 Mar 2010 18:41:46 +0000 (11:41 -0700)]
Merge "Revert "Fix V8 crashes with plugins. The extra call to _NPN_ReleaseObject inside forgetV8ObjectForNPObject is causing a double deletion.""
Grace Kloba [Tue, 16 Mar 2010 18:40:52 +0000 (11:40 -0700)]
Merge "Fix the V8 NPObject reference count problem."
Ben Murdoch [Tue, 16 Mar 2010 18:37:48 +0000 (18:37 +0000)]
Revert "Fix V8 crashes with plugins. The extra call to _NPN_ReleaseObject inside forgetV8ObjectForNPObject is causing a double deletion."
Right problem, wrong fix.
This reverts commit
9e0d526d0170e26b1dc2d0da8347cca28b5f3507.
Grace Kloba [Tue, 16 Mar 2010 18:19:34 +0000 (11:19 -0700)]
Fix the V8 NPObject reference count problem.
To match JSC version, after getNPObject() is called,
the object should have ref count 2. One will be released
in ~PluginView(). Another one will be released by
the plugin.
This should fix the crash of Flash using V8.
Fix http://b/issue?id=
2511963
Ben Murdoch [Tue, 16 Mar 2010 15:58:16 +0000 (08:58 -0700)]
Merge "Fix V8 crashes with plugins. The extra call to _NPN_ReleaseObject inside forgetV8ObjectForNPObject is causing a double deletion."
Cary Clark [Tue, 16 Mar 2010 12:58:14 +0000 (08:58 -0400)]
set up the loader so the plugin will receive the first block of data
this makes our port follow the pattern of the gtk port
fixes http://b/
2484987
Change-Id: I7119a4b84c12f5ed69a9f260148536a3b76016d6
Ben Murdoch [Tue, 16 Mar 2010 11:28:32 +0000 (11:28 +0000)]
Fix V8 crashes with plugins. The extra call to _NPN_ReleaseObject inside forgetV8ObjectForNPObject is causing a double deletion.
In ScriptController::clearScriptObjects, we call Unregister (which without this change causes Release to get called in forgetV8ObjectForNpObject).
If the ref count gets to 0, we free the object in Release. Then when Unregister returns, we call Release again in the ScriptController with the
same just-free'd pointer. If we're unlucky, then the Deallocate and free gets called again. We're also trying to access members such as the reference
count and deallocation function through a dead pointer.
Also, in the case where the Flash does not clear it's window object, we call Deallocate directly inside ScriptController::clearScriptObjects. This causes
Unregister to get called (as the object is still alive) which calls Release through forgetV8ObjectForNPObject, which results in a recursive call to Deallocate,
which frees the memory. Then the stack unwinds and we happily try to free again from the first call to Deallocate.
Fix these issues by removing the extra call to NPN_ReleaseObject in forgetV8ObjectForNpObject.
Change-Id: I7f6d21cd4ff38e29bd1a178e4816d023560b6b16
Bart Sears [Mon, 15 Mar 2010 22:50:21 +0000 (15:50 -0700)]
Putting Grace's change back in now that we have the new apk from Adobe.
Revert "Temporary revert the new android_npapi. Waiting for"
This reverts commit
31ba9c5b345211c7c89d08a762c09582a3975738.
Cary Clark [Mon, 15 Mar 2010 20:50:26 +0000 (13:50 -0700)]
Merge "read old focus before it's thrown away"
Cary Clark [Mon, 15 Mar 2010 20:34:21 +0000 (16:34 -0400)]
read old focus before it's thrown away
Change-Id: Iab6a64aec5f146e8b0286d1249d353567b49ea49
http://b/
2515977
Derek Sollenberger [Mon, 15 Mar 2010 18:22:08 +0000 (14:22 -0400)]
If the plugin is in an IFrame we need to reposition the plugins surface
every time the IFrame changes position within the document.
Change-Id: I03090bda439c3a3cac95dce42b3ea37901dc42f8
Derek Sollenberger [Mon, 15 Mar 2010 17:32:31 +0000 (10:32 -0700)]
Merge "Clean up full-screen mode plugins now that they handle touch in their java view."
Steve Block [Mon, 15 Mar 2010 17:31:10 +0000 (10:31 -0700)]
Merge "Adds some logging to JObjectWrapper"
Steve Block [Mon, 15 Mar 2010 16:15:47 +0000 (16:15 +0000)]
Adds some logging to JObjectWrapper
This is an attempt to track down the cause of a crashing bug in the WebCore
bridge, which is only occasionally triggered by the monkey.
Bug:
2435360
Change-Id: I3171eb1978de285fcf258f908aa0a08d95b72968
Cary Clark [Mon, 15 Mar 2010 12:40:48 +0000 (05:40 -0700)]
Merge "adjust cache layer position by offsets"
Derek Sollenberger [Tue, 9 Mar 2010 22:40:44 +0000 (17:40 -0500)]
Clean up full-screen mode plugins now that they handle touch
in their java view.
There is a matching commit in frameworks/base
Change-Id: Ibe76eb22c2b5ccc7fb80e4807edcbbc4d25e4793
Grace Kloba [Sat, 13 Mar 2010 02:45:47 +0000 (18:45 -0800)]
Temporary revert the new android_npapi. Waiting for
the new apk from Adobe.
Grace Kloba [Fri, 12 Mar 2010 22:13:39 +0000 (14:13 -0800)]
Merge "Use the bounds we layout the page for screenRect. This should fix the problem where links in google.com news panel show up in the subwindow."
Grace Kloba [Fri, 12 Mar 2010 21:34:26 +0000 (13:34 -0800)]
Add requestCenterFitZoom to android_npapi so that
Flash can trigger the smart-zoom
Fix http://b/issue?id=
2510670
Grace Kloba [Fri, 12 Mar 2010 21:54:21 +0000 (13:54 -0800)]
Use the bounds we layout the page for screenRect.
This should fix the problem where links in google.com
news panel show up in the subwindow.
Fix http://b/issue?id=
2509597
Cary Clark [Fri, 12 Mar 2010 20:54:48 +0000 (15:54 -0500)]
iterate through the children of the label to find its text
Change-Id: Ib4d713fd5797d9ee5c0a3080085a2db438c2bee0
http://b/
2502091
Cary Clark [Fri, 12 Mar 2010 19:31:50 +0000 (14:31 -0500)]
adjust cache layer position by offsets
Nodes in layers need to know where they are relative to the layers
so that the cursor rings are correctly positioned as the layer
moves (in document coordinates). In addition to tracking the
global offset to make the coordinates relative to any parent
frames, the node is offset relative to the body.
Sometimes layers have zero height or width; in this case, don't
treat them as clips.
Change-Id: Id2811c31a4a0674d316aadda210570ec93311013
http://b/
2503096
Leon Scroggins [Fri, 12 Mar 2010 16:09:10 +0000 (08:09 -0800)]
Merge "Outset the rectangle containing the match after checking the clip."
Leon Scroggins [Thu, 11 Mar 2010 21:53:18 +0000 (16:53 -0500)]
Outset the rectangle containing the match after checking the clip.
When considering whether a match is contained by the current clip,
do not outset it first. Check to see if it is contained by the
clip, and then add the outset so the drawn rectangle is larger.
Fix for http://b/issue?id=
2507893
Change-Id: I900d92432d412396e8c5b9e9e341085656ed0a2f
Cary Clark [Thu, 11 Mar 2010 18:33:39 +0000 (10:33 -0800)]
Merge "fix nav algorithm when node is clipped"
Ben Murdoch [Thu, 11 Mar 2010 17:39:27 +0000 (09:39 -0800)]
Merge "Prepare touch key modifiers for upstreaming."