Make WebSocket available again in service workers

Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588
Committed: https://crrev.com/dd3b58a1f0073215d58bcb8dc216b89e546bc9f7
Cr-Commit-Position: refs/heads/master@{#437406}

[tyoshino (SeeGerritForStatus), 2016-12-08 07:06:35]

Description was changed from

==========
Make WebSocket available again in Service Workers

BUG=671588
==========

to

==========
Make WebSocket available again in Service Workers

BUG=671588
==========

[tyoshino (SeeGerritForStatus), 2016-12-08 07:07:28]

The CQ bit was checked by tyoshino@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-08 07:07:45]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[falken, 2016-12-08 07:29:25]

falken@chromium.org changed reviewers:
+ falken@chromium.org

[falken, 2016-12-08 07:29:26]

Thanks for the fix. Could you add to the CL description why
"(document()->frame()->interfaceProvider())" was incorrect?

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:34:
};
this file should also be indented in service worker style.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html
(right):

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:5:
<script src="/resources/get-host-info.js?pipe=sub"></script>
i don't think you need this

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:24:
registration.active.postMessage({port: channel.port2}, [channel.port2]);
You don't need ports. Just registration.active.postMessage(). And inside the
worker, you do event.source.postMessage().

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:28:
return service_worker_unregister(t, SCOPE);
you can just call add_completion_callback(() => { r.unregister(); }) after
registration so we cleanup even if the test fails.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:30:
}, 'Verify WebSockets can be created in a Service Worker');
this should be indented in service worker test style:

promise_test(t => {
     const ....;
     return unregistre_and_register(...)
       .then(r => {
           registratino = r;
         })
       .then(() => {
            .....
         });
  }, 'Verify');

[commit-bot: I haz the power, 2016-12-08 07:45:34]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-12-08 07:45:34]

Dry run: Try jobs failed on following builders:
  android_n5x_swarming_rel on master.tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/android_n5x_...)

[tyoshino (SeeGerritForStatus), 2016-12-08 08:28:15]

The CQ bit was checked by tyoshino@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-08 08:28:31]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[tyoshino (SeeGerritForStatus), 2016-12-08 08:55:01]

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:34:
};
On 2016/12/08 07:29:26, falken wrote:
> this file should also be indented in service worker style.

Done.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html
(right):

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:5:
<script src="/resources/get-host-info.js?pipe=sub"></script>
On 2016/12/08 07:29:26, falken wrote:
> i don't think you need this

Done.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:24:
registration.active.postMessage({port: channel.port2}, [channel.port2]);
On 2016/12/08 07:29:26, falken wrote:
> You don't need ports. Just registration.active.postMessage(). And inside the
> worker, you do event.source.postMessage().

Done.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:28:
return service_worker_unregister(t, SCOPE);
On 2016/12/08 07:29:26, falken wrote:
> you can just call add_completion_callback(() => { r.unregister(); }) after
> registration so we cleanup even if the test fails.

Done.

https://codereview.chromium.org/2564493002/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:30:
}, 'Verify WebSockets can be created in a Service Worker');
On 2016/12/08 07:29:26, falken wrote:
> this should be indented in service worker test style:
> 
> promise_test(t => {
>      const ....;
>      return unregistre_and_register(...)
>        .then(r => {
>            registratino = r;
>          })
>        .then(() => {
>             .....
>          });
>   }, 'Verify');

Done.

[tyoshino (SeeGerritForStatus), 2016-12-08 08:57:28]

We need to change LocalFrame to make it distinguishable whether the retrieved
interface provider is empty impl or not.

[tyoshino (SeeGerritForStatus), 2016-12-08 13:03:52]

The CQ bit was checked by tyoshino@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-08 13:04:05]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[tyoshino (SeeGerritForStatus), 2016-12-08 13:04:08]

tyoshino@chromium.org changed reviewers:
+ yhirano@chromium.org

[falken, 2016-12-08 13:12:30]

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:34:
};
sorry I gave a bad review about simple.js formatting. the original formatting
was correct since none of these functions are function arguments. so basically
this whole file should be two-space indent, not four. i.e.:

function a(b) {
  // two space indent
  console.log('hi');
}

onmessage = function(a) {
  // two space indent
  console.log('hi');
};

is actually correct. but we would do:

.then(r => {
    add_completion_callback(...)
  });

as in websocket-in-service-worker.html since the function is an argument to
then().

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html
(right):

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:19:
navigator.serviceWorker.onmessage = t.step_func(msg => {
generally you don't need step_func in promise chain promise_tests since an error
in the promise chain bubbles up to result in a rejected promise returned by the
test. you could confirm manually.

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:26:
}, 'Verify WebSockets can be created in a Service Worker');
nit: indent line 26 an additional two spaces

[tyoshino (SeeGerritForStatus), 2016-12-08 13:50:04]

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:34:
};
On 2016/12/08 13:12:29, falken wrote:
> sorry I gave a bad review about simple.js formatting. the original formatting
> was correct since none of these functions are function arguments. so basically
> this whole file should be two-space indent, not four. i.e.:
> 
> function a(b) {
>   // two space indent
>   console.log('hi');
> }
> 
> onmessage = function(a) {
>   // two space indent
>   console.log('hi');
> };
> 
> is actually correct. but we would do:
> 
> .then(r => {
>     add_completion_callback(...)
>   });
> 
> as in websocket-in-service-worker.html since the function is an argument to
> then().

Oops. No. It's my bad, sorry. I was hurried and misunderstood that SW
directories are still using 4sp indent. You were talking about lines right after
line wrapping for e.g. { and (.

Fixed!

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html
(right):

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:19:
navigator.serviceWorker.onmessage = t.step_func(msg => {
On 2016/12/08 13:12:30, falken wrote:
> generally you don't need step_func in promise chain promise_tests since an
error
> in the promise chain bubbles up to result in a rejected promise returned by
the
> test. you could confirm manually.

Yes, but this line is setting a handler function to onmessage. Not inside
.then() or the Promise construction function where any error will bubble up to
the Promise and get propagated to itself or the promise returned by the .then().
AssertionErrors wouldn't get handled correctly by the testharness unless we
insert step_func() here.

https://codereview.chromium.org/2564493002/diff/80001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/websocket-in-service-worker.html:26:
}, 'Verify WebSockets can be created in a Service Worker');
On 2016/12/08 13:12:30, falken wrote:
> nit: indent line 26 an additional two spaces

Done.

[tyoshino (SeeGerritForStatus), 2016-12-08 14:03:21]

Sorry, I've uploaded the patch with the fixes now. PTAL at ps7.

In ps6, I've finally fixed the issue while keeping the Android third-party
blocking working. I said:

> We need to change LocalFrame to make it distinguishable whether the retrieved
interface provider is empty impl or not.

But this would need more code change. Instead, I just added a comparison with
InterfaceProvider::getEmptyInterfaceProvider() so that we use
document()->frame()->interfaceProvider() only when it's not the empty provider.

InterfaceProvider::getEmptyInterfaceProvider() returns a static instance of
InterfaceProvider with no-op impl. A LocalFrame::interfaceProvider() returns the
empty provider when a nullptr was provided to LocalFrame::create(), and
otherwise it returns the InterfaceProvider provided to LocalFrame::create().

[tyoshino (SeeGerritForStatus), 2016-12-08 14:04:15]

The CQ bit was checked by tyoshino@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-08 14:04:33]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[tyoshino (SeeGerritForStatus), 2016-12-08 14:14:57]

Description was changed from

==========
Make WebSocket available again in Service Workers

BUG=671588
==========

to

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

BUG=671588
==========

[tyoshino (SeeGerritForStatus), 2016-12-08 14:15:19]

Description was changed from

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

BUG=671588
==========

to

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

BUG=671588
==========

[tyoshino (SeeGerritForStatus), 2016-12-08 14:15:52]

I've updated the CL description to explain all the background and what change
this patch made.

[tyoshino (SeeGerritForStatus), 2016-12-08 14:59:32]

OK. All bots are green now for ps6.

[commit-bot: I haz the power, 2016-12-08 15:45:31]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-12-08 15:45:31]

Dry run: This issue passed the CQ dry run.

[falken, 2016-12-08 15:48:24]

lgtm

https://codereview.chromium.org/2564493002/diff/120001/third_party/WebKit/Lay...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/120001/third_party/WebKit/Lay...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:26:
reportFailure('Closed before receiving reply');
nit: won't make a difference but could return; here

[tyoshino (SeeGerritForStatus), 2016-12-08 16:07:05]

https://codereview.chromium.org/2564493002/diff/120001/third_party/WebKit/Lay...
File
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js
(right):

https://codereview.chromium.org/2564493002/diff/120001/third_party/WebKit/Lay...
third_party/WebKit/LayoutTests/http/tests/serviceworker/websocket/resources/simple.js:26:
reportFailure('Closed before receiving reply');
On 2016/12/08 15:48:24, falken wrote:
> nit: won't make a difference but could return; here

Ooo. Good catch. Fixed.

[tyoshino (SeeGerritForStatus), 2016-12-08 16:13:05]

Description was changed from

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

BUG=671588
==========

to

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588
==========

[tyoshino (SeeGerritForStatus), 2016-12-08 16:13:16]

The CQ bit was checked by tyoshino@chromium.org

[tyoshino (SeeGerritForStatus), 2016-12-08 16:13:16]

The patchset sent to the CQ was uploaded after l-g-t-m from falken@chromium.org
Link to the patchset: https://codereview.chromium.org/2564493002/#ps140001
(title: "Addressed #26")

[commit-bot: I haz the power, 2016-12-08 16:13:47]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-12-08 17:47:00]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-12-08 17:47:02]

Try jobs failed on following builders:
  android_n5x_swarming_rel on master.tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/android_n5x_...)

[falken, 2016-12-09 00:38:05]

The CQ bit was checked by falken@chromium.org

[commit-bot: I haz the power, 2016-12-09 00:38:39]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-12-09 01:21:32]

CQ is committing da patch.

Bot data: {"patchset_id": 140001, "attempt_start_ts": 1481243885883320,
"parent_rev": "88c08792d4d33b9fd6b1de293fff8cbea3cae224", "commit_rev":
"e844f364e85b3c3f787b18d0d1bdc8ffe970e4d2"}

[commit-bot: I haz the power, 2016-12-09 01:21:57]

Description was changed from

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588
==========

to

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588
==========

[commit-bot: I haz the power, 2016-12-09 01:21:58]

Committed patchset #8 (id:140001)

[commit-bot: I haz the power, 2016-12-09 01:26:09]

Description was changed from

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588
==========

to

==========
Make WebSocket available again in service workers

The patch https://codereview.chromium.org/2284473002/ was basically a
refactoring change, but it also included a change that it started using
the InterfaceProvider instance of the LocalFrame of the associated
Document if any. However, in a service worker, the LocalFrame returns
the empty provider obtained by calling
InterfaceProvider::getEmptyInterfaceProvider(). This is because the
WebEmbeddedWorkerImpl used for service workers doesn't override
the WebFrameClient::interfaceProvider() method which returns nullptr by
default while one on the WebSharedWorkerImpl and RenderFrameImpl
returns an effective InterfaceProvider instance.

This patch fixes the issue by changing the code to use the
InterfaceProvider of the LocalFrame only when it's not the empty
provider.

R=falken@chromium.org
TBR=yhirano@chromium.org
BUG=671588

Committed: https://crrev.com/dd3b58a1f0073215d58bcb8dc216b89e546bc9f7
Cr-Commit-Position: refs/heads/master@{#437406}
==========

[commit-bot: I haz the power, 2016-12-09 01:26:11]

Patchset 8 (id:??) landed as
https://crrev.com/dd3b58a1f0073215d58bcb8dc216b89e546bc9f7
Cr-Commit-Position: refs/heads/master@{#437406}