Bleeding-edge WebKit

[ariya]

ariya.hi...@gmail.com commented:

With QtWebKit being separated from Qt, it makes sense to include a copy of WebKit code into the source tree. That way, it is possible to build PhantomJS with a much more up-to-date/recent version of WebKit.

Disclaimer:
This issue was migrated on 2013-03-15 from the project's former issue tracker on Google Code, Issue #31.
🌟   52 people had starred this issue at the time of migration.

[curiousdannii]

curiousdannii@gmail.com commented:

Maybe instead of including it directly in git, add an action to the makefile to download the source.

[ariya]

ariya.hi...@gmail.com commented:

The advantage of including it directly is to simplify the build, this is because several pregenerated files can be included (same as how WebKit in included in Qt source tree).

Fetching the source at build time, applying patches (if any), and building WebKit there would require more dependencies (https://trac.webkit.org/wiki/QtWebKit#BuildInstructions).

Maybe we can do a hybrid solution: the plain WebKit sources are fetched from the official repository (particular stable revision) and the generated files are included in PhantomJS repository in case they are optionally needed at build time.

[CodeMonkeyKevin]

dco...@gmail.com commented:

Yes, this will also allow us to include patches from chrome webkit build (like --disable-web-security)

[ariya]

goun...@neptis.info commented:

[ariya]

ariya.hi...@gmail.com commented:

Postpone after 1.2 since there are enough big changes in 1.2 already.

 
Metadata Updates

• Milestone updated: FutureRelease (was: Release1.2)

[mscottford]

sc...@corgibytes.com commented:

One thing too keep in mind is the size of the WebKit source tree. It's huge! That may make it unpleasant for people cloning for the first time. But your milage may vary depending on bandwidth and how many commits from the WebKit tree you keep.

[ariya]

ariya.hi...@gmail.com commented:

WebKit source tree is huge because it contains tons of layout tests. We only need to most absolute minimal source files needed to build the library.

OTOH, I expect this is to be a snapshot import, probably only once a while. There is no need to merge every single revision. Thus, the growing of the repository size can be controlled.

[ariya]

ariya.hi...@gmail.com commented:

Actually, I was playing with the idea to include Qt as well. Of course, only the minimal configuration (i.e. no GraphicsView). This way, it's a matter of building everytime at once.

It would be useful for those who want to build from source, or put it in a cloud platform.

The challenge here is to find the most minimalistic sources possible so that we don't grow astronomically.

[ariya]

ariya.hi...@gmail.com commented:

For including Qt, see the new issue 226.

[ariya]

ariya.hi...@gmail.com commented:

 

[ariya]

ariya.hi...@gmail.com commented:

Related:

issue 448: Investigate the move to Qt 5

[aaronmgriffin]

aaronmgr...@gmail.com commented:

The following patches can be used to build against the system-wide Qt and QtWebKit. This builds fine against QtWebKit 2.2.1 and works fine.

Note that patch 0001 simply removed the src/qt dir, but it is too big to attach here

[ariya]

ariya.hi...@gmail.com commented:

There is no plan to switch to use system-wide Qt and QtWebKit, that's just reverting the effort put into importing Qt!

See issue 226 for details and the mailing-list for the discussion.

[thomasbachem]

m...@thomasbachem.com commented:

Several QtWebKit font bugs are really a show-stopper for me right now, but seem to be solved in most recent QtWebKit versions (http://trac.webkit.org/changeset/113968), so it's not just about bleeding-edge feature support but also bleeding-edge bug fixes :).

[leeight]

leei...@gmail.com commented:

Now can we compile the phantomjs with the bleeding-edge qt-webkit?

[ariya]

n...@boardcast.us commented:

Wondering if this issue is still intended for a future release, we'd love to see up to date Qt support.

[ariya]

ariya.hi...@gmail.com commented:

Yes, it is still intended for a future release.

[ariya]

juan...@gmail.com commented:

Hi Ariya,

First, thanks for the awesome work on PhantomJS and Esprima.

I'm testing something new that only uses ES5 features, like .bind. Can you give me an idea if this will be included in the next release and if you have an ETA (before the end of the year or first quarter 2013).

Cheers,

Juan Pinzon

[ariya]

ariya.hi...@gmail.com commented:

Juan, the answer for your question is "I don't know".

[ariya]

ariya.hi...@gmail.com commented:

Some more info: http://blogs.kde.org/2012/11/14/introducing-qtwebkit-23.

[darwin]

antonin....@gmail.com commented:

Hi Ariya, thanks for keeping an eye on this issue and your heroic work on PhantomJS.

What does that blog post mean? I'm not very familiar with Qt project.

I started a small toy project, it depends on PhantomJS and latest webkit API "DOM Mutation Observers":
https://github.com/darwin/terraform

Missing mutation observers in current PhantomJS is the only thing which holds me back. I look forward to see updated webkit in PhantomJS. Thanks.

[ariya]

dgrego...@bericotechnologies.com commented:

Is there anything we, the community, can do to help see a more recent version of Webkit in PhantomJS more quickly? I'm personally getting bit by lack of Function.prototype.bind, and Issue #522 is waiting on this issue for its resolution as well.

[ariya]

ariya.hi...@gmail.com commented:

It's not going to happen for 1.8, see https://groups.google.com/d/topic/phantomjs/rl-jNxUMB9k/discussion.

[yanivtal]

yanivtal...@gmail.com commented:

Getting SVG crashes on Mac OSX. This would really help! Thanks for your great work Ariya.

[ariya]

ariya.hi...@gmail.com commented:

Some latest development information: https://groups.google.com/d/topic/phantomjs/2Zb1Ryc5y8E/discussion.

[ariya]

Another update: https://groups.google.com/d/msg/phantomjs/nXGCwSdsDk8/HX8p-yIT5sMJ. There is a possibility we just want to get this by moving to Qt 5 all together.

[JamesMGreene]

For the record:
As far as I've been able to discern, the WebKit included in Qt 5.0.2 is from revision 136242, which was last changed on 2012-11-30 (just a few days before the Qt 5 release). That seems plenty current.

I can't figure out what the corresponding 5xx.xx version number would be, though...?

[domenic]

@JamesMGreene does it have Function.prototype.bind?

[JamesMGreene]

Update: Looks like the WebKit version is 537.21.

@domenic: Har har, yes, it appears to include Function.prototype.bind. 😉

Results from my PowerGREP over the downloaded source ZIP:

qt-everywhere-opensource-src-5.0.2\qtwebkit\Source\JavaScriptCore\runtime\FunctionPrototype.cpp:
      42  static EncodedJSValue JSC_HOST_CALL functionProtoFuncBind(ExecState*);
      66      JSFunction* bindFunction = JSFunction::create(exec, globalObject, 1, exec->propertyNames().bind.string(), functionProtoFuncBind);
     171  EncodedJSValue JSC_HOST_CALL functionProtoFuncBind(ExecState* exec)

[Havelock-Vetinari]

Is there a chance for addressing this issue #10375 [phantomjs pretends to be a touch browser]?

[JamesMGreene]

Cross-reference, as this other is likely required first: #10448

[ariya]

Master branch is now running fresher QtWebKit, see also #10448.

[coderzbx]

set option:load_image as True, resolve my problem