Added <link rel=preconnect> crossorigin attribute

Added <link rel=preconnect> crossorigin attribute

In order to support preconnect for both anonymous and non-anonymous connection pools we need to add support for the crossorigin attribute in <link rel=preconnect>.
This CL implements the core functionality of this change on the Blink side.

BUG=468005
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=195184

[Yoav Weiss, 2015-05-11 12:49:04]

yoav@yoav.ws changed reviewers:
+ mkwst@chromium.org

[Yoav Weiss, 2015-05-11 12:49:05]

Hey Mike,

Can you take a look?

[Mike West, 2015-05-11 12:53:36]

LGTM % nit.

https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
File Source/core/html/CrossOriginAttribute.h (right):

https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
Source/core/html/CrossOriginAttribute.h:19: return
CrossOriginAttributeAnonymous;
Is this the behavior we current exhibit? That is, if I have
`crossorigin="ljkahsfgkjbsdnfk;j"`, what happens?

https://codereview.chromium.org/1138743002/diff/1/Source/platform/network/Net...
File Source/platform/network/NetworkHints.cpp (right):

https://codereview.chromium.org/1138743002/diff/1/Source/platform/network/Net...
Source/platform/network/NetworkHints.cpp:43: // TODO(yoav): Should send the
crossOrigin data to the Chrome side.
Why not define the new method in this patch and give it a default implementation
that forwards to the existing implementation? Then you'll have one less thing to
do in the patch where you remove the old implementation.

https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
File public/platform/WebCrossOriginAttribute.h (right):

https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
public/platform/WebCrossOriginAttribute.h:13:
CrossOriginAttributeUseCredentials,
You'll also need "no-cors".

[sof, 2015-05-11 12:57:31]

https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
File Source/core/html/CrossOriginAttribute.h (right):

https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
Source/core/html/CrossOriginAttribute.h:19: return
CrossOriginAttributeAnonymous;
On 2015/05/11 12:53:35, Mike West (traveling. slow.) wrote:
> Is this the behavior we current exhibit? That is, if I have
> `crossorigin="ljkahsfgkjbsdnfk;j"`, what happens?

The invalid value default for the crossorigin attribute is "anonymous", 

 https://html.spec.whatwg.org/#cors-settings-attribute

so this looks consistent with that.

[Yoav Weiss, 2015-05-11 12:59:33]

On 2015/05/11 12:53:36, Mike West (traveling. slow.) wrote:
> LGTM % nit.
> 
>
https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
> File Source/core/html/CrossOriginAttribute.h (right):
> 
>
https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
> Source/core/html/CrossOriginAttribute.h:19: return
> CrossOriginAttributeAnonymous;
> Is this the behavior we current exhibit? That is, if I have
> `crossorigin="ljkahsfgkjbsdnfk;j"`, what happens?

According to the spec, this should be "anonymous". I added tests that make sure
this is the case.
Looking at the layout tests, that seems to be the case for images and scripts as
well.

> 
>
https://codereview.chromium.org/1138743002/diff/1/Source/platform/network/Net...
> File Source/platform/network/NetworkHints.cpp (right):
> 
>
https://codereview.chromium.org/1138743002/diff/1/Source/platform/network/Net...
> Source/platform/network/NetworkHints.cpp:43: // TODO(yoav): Should send the
> crossOrigin data to the Chrome side.
> Why not define the new method in this patch and give it a default
implementation
> that forwards to the existing implementation? Then you'll have one less thing
to
> do in the patch where you remove the old implementation.

I'll do that.

> 
>
https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
> File public/platform/WebCrossOriginAttribute.h (right):
> 
>
https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
> public/platform/WebCrossOriginAttribute.h:13:
> CrossOriginAttributeUseCredentials,
> You'll also need "no-cors".

What's the difference between "no-cors" and "unset"? I guess I misunderstood you
earlier...

[Mike West, 2015-05-11 13:15:28]

On 2015/05/11 at 12:59:33, yoav wrote:
> >
https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
> > File Source/core/html/CrossOriginAttribute.h (right):
> > 
> >
https://codereview.chromium.org/1138743002/diff/1/Source/core/html/CrossOrigi...
> > Source/core/html/CrossOriginAttribute.h:19: return
> > CrossOriginAttributeAnonymous;
> > Is this the behavior we current exhibit? That is, if I have
> > `crossorigin="ljkahsfgkjbsdnfk;j"`, what happens?
> 
> According to the spec, this should be "anonymous". I added tests that make
sure this is the case.
> Looking at the layout tests, that seems to be the case for images and scripts
as well.

Great! Weird, but great!

> >
https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
> > File public/platform/WebCrossOriginAttribute.h (right):
> > 
> >
https://codereview.chromium.org/1138743002/diff/1/public/platform/WebCrossOri...
> > public/platform/WebCrossOriginAttribute.h:13:
> > CrossOriginAttributeUseCredentials,
> > You'll also need "no-cors".
> 
> What's the difference between "no-cors" and "unset"? I guess I misunderstood
you earlier...

You may be right. 'no-cors' might only be explicitly necessary in `fetch()`.

[Yoav Weiss, 2015-05-11 14:58:24]

The CQ bit was checked by yoav@yoav.ws

[Yoav Weiss, 2015-05-11 14:58:25]

The patchset sent to the CQ was uploaded after l-g-t-m from mkwst@chromium.org
Link to the patchset: https://codereview.chromium.org/1138743002/#ps20001
(title: "Added interface with crossOrigin and fixed link issue")

[commit-bot: I haz the power, 2015-05-11 14:58:45]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1138743002/20001

[commit-bot: I haz the power, 2015-05-11 16:45:19]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=195184

[Yoav Weiss, 2015-05-11 20:56:41]

A revert of this CL (patchset #2 id:20001) has been created in
https://codereview.chromium.org/1131513004/ by yoav@yoav.ws.

The reason for reverting is: The patch seems to be causing preconnect related
crashes:
https://build.chromium.org/p/chromium.webkit/builders/Linux%20ChromiumOS%20Te...

Reverting until I get a clearer picture as to why that is the case..