ServiceWorker: Call SyncMatchingRegistration when document_url is changed

third_party/WebKit/LayoutTests/http/tests/serviceworker/resources/update-claim-worker.php

Index: third_party/WebKit/LayoutTests/http/tests/serviceworker/resources/update-claim-worker.php
diff --git a/third_party/WebKit/LayoutTests/http/tests/serviceworker/resources/update-claim-worker.php b/third_party/WebKit/LayoutTests/http/tests/serviceworker/resources/update-claim-worker.php
new file mode 100644
index 0000000000000000000000000000000000000000..9bc49ed7c49995aac448e6e8841006ba01b5535d
--- /dev/null
+++ b/third_party/WebKit/LayoutTests/http/tests/serviceworker/resources/update-claim-worker.php
@@ -0,0 +1,21 @@
+<?php
+header("Cache-Control: no-cache, must-revalidate");
+header("Pragma: no-cache");
+header('Content-Type:application/javascript');
+
+echo '// ', microtime();
+?>
+
+addEventListener('fetch', function(event) {
+    return;
+  });

[nhiroki, 2016/08/19 14:32:53]

Is this noop handler necessary?

[shimazu, 2016/08/22 02:38:10]

Yes, necessary. This is for the browser not to skip this script.

[nhiroki, 2016/08/22 06:35:49]

I may not still understand this.

'install' and 'activate' event handlers should be fired regardless of the
existence of 'fetch' event handler, right? Or, do you want to fire the no-op
'fetch' event handler for some reason?

[shimazu, 2016/08/22 09:29:40]

Yes, install and activate will be fired when registration.update() is called.
This fetch is triggered when requesting redirect.php, and I thought if sw was
skipped, initialization of swph would be also skipped. 

As far as now I check the code around SWPH::DidLookupRegistrationForMainResource
again, our current implementation seems to have no problem, but I think it might
be better to have fetch handler for simulating real use cases (and bypassing the
no-fetch optimization clearly).

WDYT?

[nhiroki, 2016/08/22 12:41:37]

I see, thank you for the clarification. Keeping a fetch handler for explicitly
avoiding the optimization sounds reasonable. I think future code readers may not
know this optimization, so can we have a comment why we have a no-op handler
here?

[nhiroki, 2016/08/23 01:31:46]

By the way, if this test depends on Chromium-specific behavior, it'd be better
to move it to http/tests/serviceworker/chromium.

[shimazu, 2016/08/23 01:53:52]

Added a comment.

This optimization has been already spec'ed at [1], so I think here is fine.

[1]: Green box at 15.
https://w3c.github.io/ServiceWorker/spec/service_worker/index.html#on-fetch-r...

+
+addEventListener('install', function(event) {
+    console.log('install; skipwaiting');

[nhiroki, 2016/08/19 14:32:53]

Please remove console logs.

[shimazu, 2016/08/22 02:38:10]

Oops, sorry. Done.

+    self.skipWaiting();

[nhiroki, 2016/08/19 14:32:54]

event.waitUntil(self.skipWaiting());

(I'm assuming this is a sw script :p)

+  });
+
+addEventListener('activate', function(event) {
+    console.log('activated; claim will be called');

[nhiroki, 2016/08/19 14:32:54]

ditto.

[shimazu, 2016/08/22 02:38:10]

Done.

+    self.clients.claim();

[nhiroki, 2016/08/19 14:32:54]

event.waitUntil(self.clients.claim());

[shimazu, 2016/08/22 02:38:10]

Done.

+  });