OSDN Git Service
Kristian Monsen [Mon, 13 Dec 2010 15:59:46 +0000 (15:59 +0000)]
Clean the private cookies when we create the first or remove the last context
Bug
3243891
Change-Id: Ia67ae91bd5ca13c896e45bf4dcc79debd77051db
Kristian Monsen [Fri, 10 Dec 2010 17:29:31 +0000 (17:29 +0000)]
Make private browsing cache in-memory
This is better as private browsing data should not be written to
disk if it can be avoided.
Bug
3274225
Change-Id: I5c1832bbe26e58bc7e430ecdaa7cbd0a6ba1c384
Iain Merrick [Mon, 13 Dec 2010 14:22:43 +0000 (06:22 -0800)]
Merge "Add JNI hooks for JniUtil class."
Patrick Scott [Thu, 9 Dec 2010 23:12:20 +0000 (18:12 -0500)]
Fix hit testing inside layers.
* Prevent asking for a sync in GraphicsLayerAndroid if some property has not
changed.
* Remove scrolling logic from LayerAndroid and create a subclass for scrollable
layers.
* Report the scrolling limits to the layer in order to scroll iframes (not
turned on) and to avoid computing them each time the layer is scrolled.
* Change the foreground rect calculations to better match the non-overflow case.
* During hit testing, intersect the hitTestClip with the foreground and
background to prevent false positives in the layer.
* Prepare for iframe scrolling by adding code to trigger compositing for
iframes. This currently works great except for navigation so it disabled for
now.
Bug:
3258631
Change-Id: I0da2d8dbe25376c6aa4f485c9350048c82c6f563
Iain Merrick [Fri, 10 Dec 2010 15:50:24 +0000 (15:50 +0000)]
Add JNI hooks for JniUtil class.
This is needed because I moved the useChromiumHttpStack()
function into JniUtil (it was previously in CookieManager).
Corresponding Java change in I56b0ddca
BUG=
3270236
Change-Id: I78a92cb2a6cee4a93957f2638d96ea5b82a8620c
Ben Murdoch [Thu, 9 Dec 2010 19:54:29 +0000 (11:54 -0800)]
Merge "Update Audio MediaPlayer to cope with a live stream"
Cary Clark [Thu, 9 Dec 2010 19:29:51 +0000 (11:29 -0800)]
Merge "look for the parent anchor"
Cary Clark [Thu, 9 Dec 2010 19:00:42 +0000 (14:00 -0500)]
look for the parent anchor
The node hit may be the text inside an anchor, so chase
the parent containership to the first element before testing.
bug:
3263340
Change-Id: I2e2e4b1342024deef5539d7eb4dbd376a8b329f4
Derek Sollenberger [Thu, 9 Dec 2010 18:30:55 +0000 (10:30 -0800)]
Merge "Refactor recent changes to SharedTexture and TilesManager."
Derek Sollenberger [Thu, 9 Dec 2010 16:32:41 +0000 (11:32 -0500)]
Refactor recent changes to SharedTexture and TilesManager.
These changes are directly related to the comments found in this
CL --> https://android-git.corp.google.com/g/#change,83984
Change-Id: Ida0acb772f9bcaa4f4f5b394676d45b27f4149d2
Iain Merrick [Thu, 9 Dec 2010 18:12:45 +0000 (10:12 -0800)]
Merge "WebResponse headers should be case-insensitive."
Cary Clark [Thu, 9 Dec 2010 17:14:36 +0000 (09:14 -0800)]
Merge "call hideCursor directly, avoiding a newer cache"
Iain Merrick [Thu, 9 Dec 2010 16:59:21 +0000 (16:59 +0000)]
WebResponse headers should be case-insensitive.
The CTS test WebViewTest.testSetDownloadListener was failing due to
a disagreement over the capitalization of "Content-Disposition". Rather
than making our stack exactly match the CTS server, it looks like the
headers should actually be case-insensitive. (Note that we still preserve
the case of the headers exactly as reported by the server, so hopefully
this won't break anything else.)
BUG=
3242063
TEST=WebViewTest.testSetDownloadListener
Change-Id: I0931da98aab9080127e355e51d89ac3fde12e7c2
Cary Clark [Thu, 9 Dec 2010 16:47:01 +0000 (11:47 -0500)]
call hideCursor directly, avoiding a newer cache
The WebView interface hideCursor gets a newer cache if
one is available. This interface was also called when
the cursor animation completes, before the cursor is drawn.
Getting a newer cache at this time can cause the data
in m_ring to be invalid.
Pass the root cache stored as part of the ring data instead,
so that the ring is drawn consistently.
bug:
3262275
Change-Id: Id205a10f96329e9d5ac794a8537f25a05918cafd
Ben Murdoch [Thu, 9 Dec 2010 15:46:20 +0000 (15:46 +0000)]
Update Audio MediaPlayer to cope with a live stream
The Android framework gives us a duration of 0s for a live stream
so in that case tell WebCore we have an infinite duration.
Change-Id: I5caea8401406785eba61424cdb02abf9bfab0912
Teng-Hui Zhu [Thu, 9 Dec 2010 00:15:33 +0000 (16:15 -0800)]
When the tile get painted, it should be cleared first.
This will clear the residue for the areas not in viewport, either
the area belongs to the tiles inside or outside the viewport.
In another word, area in the viewport will be redraw anyway,
but area not in the viewport will get an invalidate only, that
is why we need to clear the content when invalidating.
bug:
3253934
Change-Id: I29b88c20cb5e68d6c0f917e7fa9fb22b395d6321
Nicolas Roard [Wed, 8 Dec 2010 22:47:40 +0000 (14:47 -0800)]
Add a DeleteTextureOperation class and use it in SharedTexture
Change-Id: If0fd058911d84263b4f6257bf19cf6b2132c029a
Nicolas Roard [Wed, 8 Dec 2010 19:33:44 +0000 (11:33 -0800)]
Merge "TexturesGenerator refactoring"
Nicolas Roard [Wed, 8 Dec 2010 18:46:15 +0000 (10:46 -0800)]
TexturesGenerator refactoring
Introduce a generic QueuedOperation class to process various operations
on the Textures generation thread. Right now this is just a refactoring,
but we need this for the layers (to queue up layers painting as well as
texture deletion operations)
Change-Id: Ib8c8ceda1965bbd178a6c720cebf1d2eb0312d8a
Jeff Brown [Wed, 8 Dec 2010 01:39:40 +0000 (17:39 -0800)]
Merge "Add app-switch keycode."
Teng-Hui Zhu [Tue, 7 Dec 2010 19:37:35 +0000 (11:37 -0800)]
GLWebViewState should get the baseLayer set during init
Because WebView::setBaseLayer can happen all before the first call
to drawGL, where the m_glWebViewState got allocated,
GLWebViewState::setBaseLayer is not called properly.
The fix is to call the GLWebViewState::setBaseLayer when
m_glWebViewState first allocated.
bug:
3171575
Change-Id: I58769aa4a883f442dc2b582ea70c5b7f6c5a04f2
Teng-Hui Zhu [Tue, 7 Dec 2010 01:00:39 +0000 (17:00 -0800)]
Get rid of some Debug spew which is out of data
For bug :
3171575
Change-Id: I8736ba7e174a7ea8e4228bfbdacb8ca443828f99
Jeff Brown [Tue, 7 Dec 2010 20:31:20 +0000 (12:31 -0800)]
Add app-switch keycode.
Change-Id: I536fbb4545953098626a7b8d804a5a22746895b1
Patrick Scott [Tue, 7 Dec 2010 18:17:33 +0000 (10:17 -0800)]
Merge "Change scrollBy to scrollTo in order to support fling."
Leon Scroggins [Mon, 6 Dec 2010 21:24:29 +0000 (16:24 -0500)]
Fix for <select> elements.
Bug:
3230016
Allow webkit to handle the click on a <select> element if
it is a RenderMenuList.
Implement PopupMenu class, using PopupMenuClient to
interact with the <select> element.
Change-Id: I9611c23304fc2fc3eb01ecbd7a46fa02cd52df9a
Patrick Scott [Mon, 6 Dec 2010 22:43:36 +0000 (17:43 -0500)]
Change scrollBy to scrollTo in order to support fling.
Requires a change in frameworks/base.
Report the scrollable rectangle when asking for a scrollable layer. This
rectangle contains the current offset and the max offset for scrolling. These
are content coordinates relative to 0, 0.
Change-Id: Ia27dd23d6232057a648892d866109770f8e5bb55
Iain Merrick [Mon, 6 Dec 2010 10:43:52 +0000 (10:43 +0000)]
Fix CookieManager.hasCookies().
To pass the CTS test, this method needs to check the persistent
database directly, ignoring any cookies that are only stored in
memory.
Depends on change Id498d281 in external/chromium, which adds the
GetCookieCount() method.
Change-Id: Ibeceaaa9942b5f4f1160fc2eeeca8b2322169a90
Iain Merrick [Mon, 6 Dec 2010 14:49:38 +0000 (06:49 -0800)]
Merge "CookieManager::flushCookieStore() is now synchronous."
Ben Murdoch [Thu, 2 Dec 2010 15:38:31 +0000 (15:38 +0000)]
Refactor AutoFill so it runs lazily.
This CL refactors the AutoFill code so that we only search for forms
when we need to, i.e. the user has focused a form field. This helps
page load times. We also re-search the document for forms if the dom
version number has changed. These changes fix a bug where a crash
occurs if the form is focused before the page has finished loading.
We also move the callsite of formFieldFocused so that it only occurs
when the IME is requested which is a more appropriate place called less
frequently than it's current callsite in EditorClientAndroid.
Bug:
3243567
Change-Id: I04fe13944d3f8248197d71a0e58c21f01c4abdd0
Patrick Scott [Fri, 3 Dec 2010 22:00:49 +0000 (17:00 -0500)]
Fix gmail crash with HW acceleration.
Bug:
3252105
Change-Id: I78801b603d46c4cbe1b8d26b47f8748bb5e3df2c
Shimeng (Simon) Wang [Wed, 1 Dec 2010 23:58:58 +0000 (15:58 -0800)]
Avoid reflow if textWrapWidth's diff is small.
Also don't try re-scroll if anchor is at left top corner.
issue:
3225233
Change-Id: I65fdeb09abb58c0543efe8e8505f3b58f05730a2
Russell Brenner [Fri, 3 Dec 2010 17:15:22 +0000 (09:15 -0800)]
Merge "Fix
3185363 by reverting
fd0bf63b4: Better lookup with @font-face local"
Patrick Scott [Fri, 3 Dec 2010 15:41:08 +0000 (07:41 -0800)]
Merge "Attempt to scroll layers everytime the user drags."
Patrick Scott [Fri, 3 Dec 2010 15:00:15 +0000 (10:00 -0500)]
Attempt to scroll layers everytime the user drags.
This is part of a frameworks/base change.
Use the composite root bounds for the inval region when updating the UI picture.
This fixes several drawing issues with outdated pictures.
Bug:
3022562
Change-Id: Ic28cfbbe381fdd59368a6d83bb73b01df912b530
Ben Murdoch [Fri, 3 Dec 2010 15:25:26 +0000 (07:25 -0800)]
Merge changes If294e929,I1baea2ee,Ie2bf417b
* changes:
Merge Chromium at r66597: Merge in AutoFill changes in r66310
Merge Chromium at r66597: Merge in AutoFill changes from r66237
Merge Chromium at r66597: AutoFill: no more queryId
Ben Murdoch [Fri, 26 Nov 2010 12:34:37 +0000 (12:34 +0000)]
Merge Chromium at r66597: Merge in AutoFill changes in r66310
See http://src.chromium.org/viewvc/chrome?view=rev&revision=66310
Change-Id: If294e929ee4d27fd6804939986be7650ff7224ad
Ben Murdoch [Fri, 26 Nov 2010 15:14:56 +0000 (15:14 +0000)]
Merge Chromium at r66597: Merge in AutoFill changes from r66237
See http://src.chromium.org/viewvc/chrome?view=rev&revision=66237
Change-Id: I1baea2eea0e6cf0def36ce3f69e8fb53a17eb2ba
Ben Murdoch [Fri, 26 Nov 2010 11:40:11 +0000 (11:40 +0000)]
Merge Chromium at r66597: AutoFill: no more queryId
See http://src.chromium.org/viewvc/chrome?view=rev&revision=66237
Change-Id: Ie2bf417bf20686ca782f6c0e4d2527e666c88ad4
Cary Clark [Fri, 3 Dec 2010 15:07:31 +0000 (07:07 -0800)]
Merge "keep rects separate across canvas push"
Cary Clark [Fri, 3 Dec 2010 14:33:42 +0000 (09:33 -0500)]
keep rects separate across canvas push
The nav cache identifies visible elements by accumulating
the drawing commands used in an interesting part of the picture,
then eliminating the portions that are too large or overdrawn.
Consecutive draws of the same type are combined, so that, for
instance, glyphs assimilate to form a string, or bitmaps form
a nine-patch. Draws of the same type shouldn't be combined when
they represent different objects.
Break apart draws of the same object if there is a push/pop
pair surrounding the previous draw of that object type.
bug:
3241830
bug:
3241992
Change-Id: I4eb65a0ae0d923d4950cda88e16cb4bbacfde663
Patrick Scott [Thu, 4 Nov 2010 14:11:34 +0000 (10:11 -0400)]
Rewrite scrollable layers using multiple layers.
Remove PlatformBridge::updateLayers and instead call layersDraw from
ChromeClientAndroid during a requested sync.
Instead of LayerAndroid having knowledge of scrolling pictures, use multiple
layers (background, mask, foreground).
Update navigation to use the new structure. Always record nodes in absolute,
unscrolled coordinates. Do not track layers unless the node has a composited
layer and a last child. Check for composited parents when adding to tracked
layers as well.
TODO:
* developer.android.com doesn't completely work with navigation due to some
nodes being clipped out.
* BaseLayerAndroid needs to be able to operate with a matrix other than
identity.
* nested fixed position elements do not draw correctly.
* nested overflow layers haven't been tested.
Change-Id: I0e2bd37612341e8884d68153ab36194cb2dc1eeb
Iain Merrick [Thu, 2 Dec 2010 11:24:53 +0000 (11:24 +0000)]
CookieManager::flushCookieStore() is now synchronous.
This ensures that all data is correctly stored before onPause() returns.
There's no API change on the Java side (although I should update the
comment to explain it's now synchronous; I'll do that in another CL).
Depends on a new completion callback added to CookieMonster.
See change Ieb3e787b in external/chromium.
Change-Id: I008f35d101a9223fe705396273bc4cd563a5f77e
Steve Block [Fri, 3 Dec 2010 10:04:19 +0000 (10:04 +0000)]
Remove some temporary logging in the JNI bridge
This was added to try to track down a crashing bug but is no longer
required.
Bug:
2435360
Change-Id: I5e88a842e42cd89e96c7f2eea54e96a440f890ca
Russell Brenner [Thu, 2 Dec 2010 19:35:32 +0000 (11:35 -0800)]
Fix
3185363 by reverting
fd0bf63b4: Better lookup with @font-face local
This reopens bug
2720133: @font-face does not work if a local font
is specified before a remote font in the src list.
Change-Id: Ibfe93549b3ea7fd88cffd6ee75674a03f5e94d43
Russell Brenner [Wed, 24 Nov 2010 18:16:29 +0000 (10:16 -0800)]
Merge WebKit at r72274: Update WebKit revision number.
Change-Id: I38e2e15fe1061f51a5fc299f45763b0e49300b4c
Ben Murdoch [Tue, 9 Nov 2010 18:17:50 +0000 (18:17 +0000)]
Merge WebKit at r72274: Implement PlatformBridge memory management methods for WebKit change 71639
Get device class specific memory constraints for V8.
V8 can be more efficient in terms of memory management but to do
so needs to know about what resources are available to it.
This varies device to device on Android so add implement the functions
that the the PlatformBridge needs to query the relevant values from
Java.
MemoryUsage.cpp will compile for JSC, but needs tuning to report
correct JSC stats. It's currently a non-issue, as this routine is only
used with V8, at present.
Requires a frameworks/base change:
https://android-git.corp.google.com/g/#change,78690
Bug:
3075565
Requires a previously upstreamed webkit change:
http://trac.webkit.org/changeset/71639
Change-Id: I25ff54b5c8e25f5de5380e0d43e56551587adc5d
Russell Brenner [Tue, 23 Nov 2010 01:32:22 +0000 (17:32 -0800)]
Merge WebKit at r72274: Moved definition of EXPORT
The definition of EXPORT was removed from WebCorePrefix.h, citing that it's only needed by WebKit/android. Added definition to WebKit/android.
http://trac.webkit.org/changeset/71578
Change-Id: I3fe6909d5e5c668299ca25dc39ef0e444620fd33
Russell Brenner [Tue, 23 Nov 2010 00:58:26 +0000 (16:58 -0800)]
Merge WebKit at r72274: Fix CodeGeneratorV8.pm
Some of the script-generation code puts include statements in source files when
they need to be in headers.
Cherry-picked from webkit:
http://trac.webkit.org/changeset/72948
Change-Id: I07d425bcb97abcdda25b5ff2ea7af887a925f489
Russell Brenner [Fri, 19 Nov 2010 22:12:47 +0000 (14:12 -0800)]
Merge WebKit at r72274: Add new flag to ImageSource constructor with asssociated storage member
Patch for webkit change:
http://trac.webkit.org/changeset/72130
Change-Id: I5718b8c1ee5b2a9c8afa31de24e882f30d62972f
Russell Brenner [Fri, 19 Nov 2010 21:57:02 +0000 (13:57 -0800)]
Merge WebKit at r72274: Add orientation stub
Orientation is hardwired to horizontal. Needs proper implementation.
http://trac.webkit.org/changeset/71970
Change-Id: If838dd9689e0e54b282baaa82515b2460e50578f
Russell Brenner [Tue, 23 Nov 2010 01:59:09 +0000 (17:59 -0800)]
Merge WebKit at r72274: Fix Makefiles
Add ResourceLoadScheduler.cpp to build
http://trac.webkit.org/changeset/71562
Change-Id: Ia2fabc508bbcd65023824424951484f644066b1d
Russell Brenner [Fri, 19 Nov 2010 17:57:40 +0000 (09:57 -0800)]
Merge WebKit at r72272: Fix conflicts.
WebCore/Android.v8bindings.mk
http://trac.webkit.org/changeset/71802
WebCore/bindings/v8/V8Proxy.cpp
http://trac.webkit.org/changeset/71735
WebCore/css/CSSParser.cpp
http://trac.webkit.org/changeset/71767
http://trac.webkit.org/changeset/72116 (several APIs purged)
WebCore/loader/ResourceLoader.cpp
http://trac.webkit.org/changeset/71884
WebCore/platform/graphics/MediaPlayer.cpp
http://trac.webkit.org/changeset/72117
WebCore/platform/image-decoders/ImageDecoder.cpp
http://trac.webkit.org/changeset/72130
WebCore/rendering/RenderBlockLineLayout.cpp
http://trac.webkit.org/changeset/72235
WebCore/rendering/RenderView.cpp
http://trac.webkit.org/changeset/71851
Change-Id: If0a3466754b9755c1f4be97bf589f88c48eece0a
Russell Brenner [Fri, 19 Nov 2010 01:33:13 +0000 (17:33 -0800)]
Merge WebKit at r72274: Initial merge by git.
Change-Id: Ie51f0b4a16da82942bd516dce59cfb79ebbe25fb
Ben Murdoch [Thu, 2 Dec 2010 17:03:03 +0000 (09:03 -0800)]
Merge "Fix crash when syncing autofill settings."
Jean-Baptiste Queru [Thu, 2 Dec 2010 15:59:57 +0000 (07:59 -0800)]
am
6a2e8695: am
5d1bf1f6: Merge from open-source master
* commit '
6a2e86958d2c1827daa0639e5e9ed9515252f3be':
Fix compilation error:
Jean-Baptiste Queru [Thu, 2 Dec 2010 15:53:20 +0000 (07:53 -0800)]
am
5d1bf1f6: Merge from open-source master
* commit '
5d1bf1f6b21848ec1086d59111306d2d4a4b5fc3':
Fix compilation error:
Jean-Baptiste Queru [Thu, 2 Dec 2010 15:49:06 +0000 (07:49 -0800)]
Merge from open-source master
Change-Id: I22fd2b911ff05cd0d01fe151b18d64b4ac98e547
Ben Murdoch [Thu, 2 Dec 2010 15:10:50 +0000 (15:10 +0000)]
Fix crash when syncing autofill settings.
Since we moved the autofill initiation code out of the constructor
and into it's own method, we may crash if there is an autofill profile
set but the feature is turned off (i.e. the user had it on at one point
and had a profile set up but turned it off without deleting the
profile). That's because in this case we execute setAutoFillProfile()
in WebSettings without having first called the new init method. It makes
sense to not sync any profiles though when the feature is not enabled,
so add that check.
Change-Id: I9d99e96347770e5b08c982513f2370cb770cbe94
Steve Block [Thu, 2 Dec 2010 12:54:32 +0000 (04:54 -0800)]
Merge "Fix style in WebCore/platform/network/android"
Kristian Monsen [Wed, 1 Dec 2010 16:04:54 +0000 (16:04 +0000)]
Enable SPDY, and renames initChromiumLogging
initChromiumLogging is now renamed to the more generic initChromium,
the files are renamed to ChromiumInit.*
Change-Id: Ic9639d637f25e664b51d076b9b0ee58c2867c1b5
Jean-Baptiste Queru [Wed, 1 Dec 2010 23:26:30 +0000 (15:26 -0800)]
Merge "Fix compilation error:"
Rob Tsuk [Wed, 1 Dec 2010 22:02:29 +0000 (14:02 -0800)]
am
1fc015f0: (-s ours) am
4741f411: Do not merge.
* commit '
1fc015f0ce86a2bed612d2eec53b01cac6b8a793':
Do not merge.
Rob Tsuk [Wed, 1 Dec 2010 21:59:11 +0000 (13:59 -0800)]
Steve Block [Tue, 30 Nov 2010 14:17:48 +0000 (14:17 +0000)]
Fix style in WebCore/platform/network/android
This matches what was upstreamed in
http://trac.webkit.org/changeset/73062
Change-Id: I2cf2064ab51105fb16039f8af285a3797abe49a9
Steve Block [Mon, 29 Nov 2010 17:18:33 +0000 (17:18 +0000)]
Use one WebRequestContext per WebView
This change switches from using a pair of WebRequestContexts (one for
private browsing, one for regular browsing) to using one
WebRequestContext per WebView. This is required to allow us to set the
userAgent on each WebView.
The WebRequestContext is now owned by the WebView.
Bug:
3113804
Change-Id: Iba2b1490e7ce4ff65c08a04a310963fa2c7e4f83
Huahui Wu [Wed, 1 Dec 2010 19:48:03 +0000 (11:48 -0800)]
Merge "b/
2864818 setting the appropriate error code for certificate error."
Rob Tsuk [Tue, 30 Nov 2010 22:32:49 +0000 (14:32 -0800)]
Do not merge.
Re-implement a security fix against our older webkit code base.
Change-Id: Iaefa1d060bb71b717558a427e326742f8fa5c4fd
modified: WebCore/rendering/RenderBlock.cpp
modified: WebCore/rendering/RenderBlock.cpp
Huahui Wu [Wed, 1 Dec 2010 19:24:06 +0000 (11:24 -0800)]
b/
2864818 setting the appropriate error code for certificate error.
Change-Id: I628b6fdd82089207c5a25ad816d3e6475d85911a
Steve Block [Wed, 1 Dec 2010 18:49:56 +0000 (10:49 -0800)]
Merge "Do more AutoFill steup lazily"
Steve Block [Wed, 1 Dec 2010 18:18:33 +0000 (18:18 +0000)]
Do more AutoFill steup lazily
Currently, much of the AutoFill setup is done in the WebAutoFill
constructor, which is called when the Page is created. This change moves
this setup to be done lazily when searchDocument() is called, which is
the entry point into the feature.
Change-Id: I0d437ed53391b5a7d8d16ec1000dc783c710e564
Steve Block [Wed, 1 Dec 2010 17:26:40 +0000 (09:26 -0800)]
Merge "Fix a bug when cleaning up private browsing"
Kristian Monsen [Wed, 1 Dec 2010 17:22:48 +0000 (09:22 -0800)]
Merge "Youtube urls can now start with /v/ or /e/"
Kristian Monsen [Wed, 1 Dec 2010 12:19:46 +0000 (12:19 +0000)]
Youtube urls can now start with /v/ or /e/
Fix for bug
3242961
Note that the youtube player crashes when you launch it,
but that also happens with old style urls.
Tested with the testcase in the bug.
Change-Id: I0cb59f8404fe47aabec34c0a21e614f553ea0c32
Steve Block [Tue, 30 Nov 2010 17:57:14 +0000 (17:57 +0000)]
Fix a bug when cleaning up private browsing
Delete the singleton WebCache and WebCookieJar objects when we
clear up private browsing, so that they are correctly recreated when
next required.
Note that this patch causes the WebCache to leak the HostResolver
object, as was the case before the refactoring in b/
3113804. See
b/
3243797.
Bug:
3113804
Change-Id: Ic0e2a45b5fd2b30097880e57eac2f6cadbd5dc85
Ben Murdoch [Wed, 1 Dec 2010 14:38:53 +0000 (14:38 +0000)]
WtfStringToJString should be able to return a valid string on empty input
It is sometimes desirable to return a valid but 0-length string created
from a 0-length WTF string but currently the WtfStringToJString function
will always return a null jstring on an empty input. Add a parameter to
specify that it should return an empty jstring on empty WTF string and
default it to false.
Change-Id: Ia449f1f9ae9b37c3da7d0272424e4fd4878ceb31
Iain Merrick [Wed, 1 Dec 2010 14:46:37 +0000 (06:46 -0800)]
Merge "Add FlushCookieStore() method to CookieManager (C++ side)."
Cary Clark [Wed, 1 Dec 2010 13:33:11 +0000 (05:33 -0800)]
Merge "attach the last extra to the next picture"
Cary Clark [Tue, 30 Nov 2010 20:51:10 +0000 (15:51 -0500)]
pass the x and y to retrieve anchor data
During a long press, the original pointer to the
node and frame may change. Pass the location instead
to attempt to find the anchor.
companion change in frameworks/base
bug:
3240869
Change-Id: Idf5eb67e2cb544dfa312e41f9224813c6ca80563
Cary Clark [Wed, 1 Dec 2010 12:38:01 +0000 (07:38 -0500)]
attach the last extra to the next picture
When a new picture is generated for the content,
use the last cursor ring / text selection etc.
for the interval between the picture message and
the view inval. This reduces flickering.
bug:
3230094
Change-Id: Ie099a8db395bfa707790ffe3c1f7f5398cc9afa4
Iain Merrick [Tue, 30 Nov 2010 11:42:19 +0000 (11:42 +0000)]
Add FlushCookieStore() method to CookieManager (C++ side).
The browser app will use this to sync cookies to flash when it is
sent to the background. Corresponding Java code is in I04e96aff.
Depends on external/chromium changes in Ie600c588.
Bug:
3231371
Change-Id: Ia9b56f3cdb5acd1fca23102b68b2f83a94ba6e69
Cary Clark [Wed, 1 Dec 2010 11:35:08 +0000 (03:35 -0800)]
am
e5097951: (-s ours) am
5793dcc1: Merge "Do not merge: fix dns prefetch, multiply range" into gingerbread
* commit '
e5097951a2019ae742b9c348b21e613631f59cdc':
Do not merge: fix dns prefetch, multiply range
Cary Clark [Tue, 30 Nov 2010 19:46:28 +0000 (11:46 -0800)]
am
5793dcc1: Merge "Do not merge: fix dns prefetch, multiply range" into gingerbread
* commit '
5793dcc19c1dd5f2f3daacd05646ae237f050e58':
Do not merge: fix dns prefetch, multiply range
Cary Clark [Tue, 30 Nov 2010 18:24:15 +0000 (10:24 -0800)]
Merge "hit test complete node even if partially visible"
Steve Block [Tue, 30 Nov 2010 15:04:16 +0000 (15:04 +0000)]
Remove a superfluous TODO in ResourceLoaderAndroid.cpp
Change-Id: I290228dce1ce3e0be56666775f591c6c2b7dab32
Cary Clark [Tue, 30 Nov 2010 13:08:37 +0000 (05:08 -0800)]
Merge "Do not merge: fix dns prefetch, multiply range" into gingerbread
Steve Block [Tue, 30 Nov 2010 12:20:07 +0000 (04:20 -0800)]
Merge "Provide a URLRequestContext directly to AutoFill, rather than a factory function"
Steve Block [Tue, 30 Nov 2010 11:32:14 +0000 (11:32 +0000)]
Provide a URLRequestContext directly to AutoFill, rather than a factory function
In b/
3044989, AutoFill was modified to take a factory function for the
URLRequestContext (which was then created when needed), rather than
taking the context directly. This was to avoid a race condition where
the context is created before the required storage paths have been
synced to the WebCore thread through the BrowserFrame.
Since https://android-git.corp.google.com/g/76579 and
https://android-git.corp.google.com/g/80056, we read the required
paths directly from JniUtils, so the race condition no longer exists. We
can therefore provide the context directly to AutoFill.
Note that while the context may be created before the Settings have been
synced to the WebCore thread, the necessary properties will be updated
as part of this sync before the context is used.
This change will also simplify things when we use one context per
WebView.
Also requires a change in external/chromium ...
https://android-git.corp.google.com/g/82146
Change-Id: Idfc34d435bae21654a02e5898aef0e84da98218a
Bart Sears [Tue, 30 Nov 2010 01:15:52 +0000 (17:15 -0800)]
am
a5b8875c: (-s ours) am
1baa4609: Cherry-pick WebKit security fix (webkit.org r64706) Do not merge
* commit '
a5b8875c0b61e9dc9eb76e16a4cc1f741709ee74':
Cherry-pick WebKit security fix (webkit.org r64706) Do not merge
Bart Sears [Tue, 30 Nov 2010 01:12:45 +0000 (17:12 -0800)]
am
1baa4609: Cherry-pick WebKit security fix (webkit.org r64706) Do not merge
* commit '
1baa4609252ea42065c4399b9081b2087c8d70a2':
Cherry-pick WebKit security fix (webkit.org r64706) Do not merge
Cary Clark [Mon, 29 Nov 2010 21:28:22 +0000 (16:28 -0500)]
hit test complete node even if partially visible
An older test incorrectly discards nodes if they are partially
off the screen. The test was attempting to ignore nodes that
were fully invisible, so change the test to do that.
bug:
3198707
Change-Id: I3cb13de3c86b43d3388849161bdfc44f84d93b3d
Nicolas Roard [Mon, 29 Nov 2010 18:57:05 +0000 (10:57 -0800)]
Merge "Refactoring the zoom"
Ben Murdoch [Mon, 29 Nov 2010 15:38:49 +0000 (07:38 -0800)]
Merge "AutoFill no longer needs a unique profile id native side."
Steve Block [Mon, 29 Nov 2010 15:19:42 +0000 (07:19 -0800)]
Merge changes I006cddf2,Ib304a845,I4622749d
* changes:
Moves implementation of ResourceLoaderAndroid to ResourceLoaderAndroid.cpp
Add a static WebRequestContext::acceptLanguage() method
Make WebCookieJar::get() threadsafe
Steve Block [Mon, 29 Nov 2010 14:55:36 +0000 (14:55 +0000)]
Moves implementation of ResourceLoaderAndroid to ResourceLoaderAndroid.cpp
This means that the implmentation of this class for both HTTP stacks is
in the same place.
Change-Id: I006cddf27c245a88327643314bb7564a2486ff38
Steve Block [Mon, 29 Nov 2010 14:09:58 +0000 (14:09 +0000)]
Add a static WebRequestContext::acceptLanguage() method
This removes the need for PlatformBridge::computeDefaultLanguage() to
get an instance of WebRequestContext. This is part of the refactoring to
use a WebRequestContext per WebView.
Bug:
3113804
Change-Id: Ib304a8457a82d09d80fe4d6298f00f688ec53b7c
Cary Clark [Thu, 11 Nov 2010 20:26:06 +0000 (15:26 -0500)]
scroll the selection when dragging near the edge
Return the bounds of a layer.
support for a companion change in frameworks/base
bug:
3186988
Change-Id: Ie8f69dc8c46c6329fd0b093ecdba4715f8cde2a9
Ben Murdoch [Mon, 29 Nov 2010 13:25:42 +0000 (13:25 +0000)]
AutoFill no longer needs a unique profile id native side.
AutoFill profiles now use GUIDs that they generate themselves
rather than injecting a unique int ID so we no longer need to
reach into the java class to grab the ID.
Change-Id: I26492aeddd130831e941ec79ef1f5e36271903c8
Steve Block [Mon, 29 Nov 2010 12:46:55 +0000 (12:46 +0000)]
Make WebCookieJar::get() threadsafe
This should have been part of
https://android-git.corp.google.com/g/78476 which factored
WebCookieJar out of WebRequestContext.
Also, now that WebRequestContext::get() is called nly on the WebCore
thread, it does not need to be threadsafe.
Bug:
3113804
Change-Id: I4622749d1650be0b9ca0d37541361955a367110f
Steve Block [Mon, 29 Nov 2010 11:51:10 +0000 (03:51 -0800)]
Merge "Factor out WebCache from WebRequestContext"
Steve Block [Fri, 26 Nov 2010 16:11:52 +0000 (16:11 +0000)]
Factor out WebCache from WebRequestContext
This is the second step in allowing us to have a WebRequestContext per
WebView, which is required to allow the useragent string to be
configured per WebView.
Bug:
3113804
Change-Id: Idcd252194562d1ef91d9d6e2baf8c65b8ff088d4