Verify service worker behavior with initial about:blank iframe docume… #6304

wanderview · 2017-06-21T16:27:24Z

…nts.

…nts. DO NOT MERGE

w3c-bots · 2017-06-21T16:38:47Z

Firefox (nightly)

Testing web-platform-tests at revision dfbfa83
Using browser at version BuildID 20170612100241; SourceStamp 27cad9749cddf68e11fdd4e5d73dad84a8f8cf23
Starting 10 test iterations
All results were stable

All results

1 test ran

/service-workers/service-worker/about-blank-replacement.https.html

Subtest	Results	Messages
	OK
`Initial about:blank is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`
`Initial about:blank modified by parent is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`
`Popup initial about:blank is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`

w3c-bots · 2017-06-21T16:38:48Z

View the complete job log.

Sauce (safari)

Testing web-platform-tests at revision dfbfa83
Using browser at version 10.0
Starting 10 test iterations
All results were stable

All results

1 test ran

/service-workers/service-worker/about-blank-replacement.https.html

Subtest	Results	Messages
	OK
`Service Worker: about:blank replacement handling`	FAIL	`SyntaxError: Unexpected keyword 'function'`

w3c-bots · 2017-06-21T16:38:48Z

View the complete job log.

Chrome (unstable)

Testing web-platform-tests at revision dfbfa83
Using browser at version 61.0.3135.4 dev
Starting 10 test iterations
All results were stable

All results

1 test ran

/service-workers/service-worker/about-blank-replacement.https.html

Subtest	Results	Messages
	OK
`Initial about:blank is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`
`Initial about:blank modified by parent is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`
`Popup initial about:blank is controlled, exposed to clients.matchAll(), and matches final Client.`	FAIL	`assert_false: result: failure: could not find about:blank client expected false got true`

w3c-bots · 2017-06-21T16:38:49Z

View the complete job log.

Sauce (MicrosoftEdge)

Testing web-platform-tests at revision dfbfa83
Using browser at version 14.14393
Starting 10 test iterations
All results were stable

All results

1 test ran

/service-workers/service-worker/about-blank-replacement.https.html

Subtest	Results	Messages
	OK
`Service Worker: about:blank replacement handling`	FAIL	`Expected ';'`

wanderview · 2017-11-17T21:22:18Z

@jungkees @jakearchibald @mattto Do you have any objections to me landing this test in gecko and upstreaming it? I believe it matches our current understanding about initial about:blank handling. The test does not depend on any reserved or resultingClientId behavior.

jakearchibald · 2017-11-20T11:10:32Z

service-workers/service-worker/about-blank-replacement.https.html

+// Helper routine that creates an iframe that internally has some kind
+// of nested window.  The nested window could be another iframe or
+// it could be a popup window.
+function createFrameWithNestedWindow(url) {


createIframeWithNestedWindow? Just to be specific that it's an iframe.

Nit: I think I'd have made this two functions. One to add the iframe (which I think we already have), and another to await the message. As it stands it feels like this function is doing two somewhat unrelated things, but maybe it's just me.

The issue with that is with_frame() waits for the URL passed to it to load in the frame. But what we really want is to get the load event of the nested frame. Its unclear to me if the load events of the first frame and the nested frame are guaranteed. Just combining the operations together seemed safest and most likely to work in all browsers.

Fair enough!

jakearchibald · 2017-11-20T11:32:37Z

service-workers/service-worker/resources/about-blank-replacement-frame.html

+  parent.postMessage({ type: 'NESTED_LOADED' }, '*');
+}
+</script>
+<iframe src="?nested=true" id="nested" onload="nestedLoaded()"></iframe>


Won't this become recursive if the page isn't service worker controlled? As in, this iframe is pointing to the same URL, so it'll include itself etc etc.

Might be safer to point it at a blank page?

The thing I struggle with there is that I need both the outer frame and the nested frame to be controlled under the same scope. Doing a query parameter was an easy way to accomplish that without creating a lot of extra sub-directories in the test. I suppose I could also use a server side to .py to return a blank document in that case.

jakearchibald · 2017-11-20T11:35:56Z

service-workers/service-worker/resources/about-blank-replacement-worker.js

+// inherit the controller from its parent.
+async function getClientByURL(url) {
+  let list = await clients.matchAll();
+  for (client of list) {


const client of list

fwiw, this could be:

return list.find(client => client.url === url);

jakearchibald · 2017-11-20T11:43:56Z

service-workers/service-worker/resources/about-blank-replacement-worker.js

+
+  // Delay the load of the nested frame so we can examine the Client
+  // representing the window while it has the initial about:blank document.
+  evt.respondWith(delayedResponse());


Does resolveDelayed need to be a global here? As far as I can tell, it doesn't need to be scoped beyond this fetch callback.

Could it be written as an async function, to make it read more sequentially? Eg:

evt.respondWith(async function() { const client = await getClientByURL('about:blank'); if (!client) return new Response('failure: could not find about:blank client'); if (url.searchParams.get('ping')) { const loc = await pingPong(client); if (loc !== 'about:blank') { return new Response(`failure: got location {$loc}, expected about:blank`); } } return new Response(client.id); }());

Yea, the async function is much cleaner. Thanks!

jakearchibald · 2017-11-20T12:20:56Z

service-workers/service-worker/about-blank-replacement.https.html

+  // Execute a test where the nested frame is simple loaded normally.
+  await doAsyncTest(t, 'resources/about-blank-replacement-frame.html');
+}, 'Initial about:blank is controlled, exposed to clients.matchAll(), and ' +
+   'matches final Client.');


I think this would fail in Edge, since it reports the iframe's URL as its src. Test: https://iframe-global-retain.glitch.me/typical.html.

My worry about these tests is we really want the clients API to reflect what the browser does, more than what the standards say they should do. My worry is that browsers monkey-patch this for the clients API, but don't fix their underlying implementation.

Maybe it's worth a note at the top of this test?

I agree clients should reflect what the browser is doing, but if the browser doesn't create initial about:blank documents at all then maybe they should just mark this test as expected fail until they fix that.

I'll add a comment, though.

jakearchibald · 2017-11-20T12:28:27Z

service-workers/service-worker/resources/about-blank-replacement-ping-frame.html

+  parent.postMessage({ type: 'NESTED_LOADED' }, '*');
+}
+</script>
+<iframe src="?nested=true&ping=true" id="nested" onload="nestedLoaded()"></iframe>


Same recursion issue here. Also & to &.

jakearchibald · 2017-11-20T12:30:15Z

service-workers/service-worker/resources/about-blank-replacement-popup-frame.html

+  parent.postMessage({ type: 'NESTED_LOADED' }, '*');
+}
+
+let popup = window.open('?nested=true');


Same recursion case.

wanderview · 2017-11-22T20:31:01Z

I'm also going to add a test case for w3c/ServiceWorker#1232. At a minimum we should verify we do not end up with a controlled iframe when there is no service worker matching the scope.

wanderview · 2017-11-22T22:04:19Z

The last commit addresses review feedback, but does not add the new test case yet.

…:blank can end up uncontrolled based on its load URL.

wanderview · 2017-11-22T22:24:08Z

@jakearchibald, do you mind reviewing the additional test case I added? I did not verify whether or not the global is shared. I only verified that there is an initial about:blank and that the final load is not controlled. I think this gives us flexibility in w3c/ServiceWorker#1232. I also think its the minimum requirement for this to make any sense.

w3c-bots · 2017-11-22T22:28:32Z

Build PASSED

Started: 2017-11-27 22:20:43
Finished: 2017-11-27 22:26:19

View more information about this build on:

jakearchibald

LGTM!

jakearchibald · 2017-11-23T12:03:31Z

service-workers/service-worker/about-blank-replacement.https.html

+// Helper routine that creates an iframe that internally has some kind
+// of nested window.  The nested window could be another iframe or
+// it could be a popup window.
+function createFrameWithNestedWindow(url) {


Fair enough!

wanderview · 2017-11-23T13:58:50Z

Thanks! I realized I forgot to do the & change you requested.

Also, do we squash commits before merging here? Or I might land it in gecko and upstream.

mfalken

Nice tests! lgtm.

mfalken · 2017-11-24T01:08:07Z

service-workers/service-worker/about-blank-replacement.https.html

+        return;
+      }
+      window.removeEventListener('message', onMsg);
+      if(evt.data.result && evt.data.result.startsWith('failure:')) {


space after if for consistency

mfalken · 2017-11-24T01:12:54Z

service-workers/service-worker/about-blank-replacement.https.html

+}
+
+promise_test(async function(t) {
+  // Execute a test where the nested frame is simple loaded normally.


simply loaded normally?

Verify service worker behavior with initial about:blank iframe docume…

aa98fbd

…nts. DO NOT MERGE

wpt-pr-bot added service-workers wg-webplatform labels Jun 21, 2017

wpt-pr-bot requested review from mkruisselbrink, ehsan and mfalken June 21, 2017 16:27

wanderview mentioned this pull request Jun 21, 2017

consider Client behavior for windows where initial about:blank is replaced with a loaded document w3c/ServiceWorker#1091

Closed

Add a window.open() test case.

aec1af3

wpt-pr-bot added the wg-sw label Nov 17, 2017

jakearchibald requested changes Nov 20, 2017

View reviewed changes

Address review feedback from @jakearchibald.

1265a0a

Add a test case that verifies a frame with a controlled initial about…

fb4d639

…:blank can end up uncontrolled based on its load URL.

jakearchibald approved these changes Nov 23, 2017

View reviewed changes

mfalken approved these changes Nov 24, 2017

View reviewed changes

Fix more review feedback

e5d1b7e

wanderview merged commit 49e48ef into web-platform-tests:master Nov 27, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Verify service worker behavior with initial about:blank iframe docume… #6304

Verify service worker behavior with initial about:blank iframe docume… #6304

wanderview commented Jun 21, 2017 •

edited by wpt-pr-bot

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

wanderview commented Nov 17, 2017

jakearchibald Nov 20, 2017

wanderview Nov 22, 2017

jakearchibald Nov 23, 2017

jakearchibald Nov 20, 2017

wanderview Nov 22, 2017

jakearchibald Nov 20, 2017

jakearchibald Nov 20, 2017

jakearchibald Nov 20, 2017

wanderview Nov 22, 2017

jakearchibald Nov 20, 2017

wanderview Nov 22, 2017

jakearchibald Nov 20, 2017

jakearchibald Nov 20, 2017

wanderview commented Nov 22, 2017

wanderview commented Nov 22, 2017

wanderview commented Nov 22, 2017

w3c-bots commented Nov 22, 2017 •

edited

Loading

jakearchibald left a comment

jakearchibald Nov 23, 2017

wanderview commented Nov 23, 2017

mfalken left a comment

mfalken Nov 24, 2017

mfalken Nov 24, 2017

Verify service worker behavior with initial about:blank iframe docume… #6304

Verify service worker behavior with initial about:blank iframe docume… #6304

Conversation

wanderview commented Jun 21, 2017 • edited by wpt-pr-bot Loading

w3c-bots commented Jun 21, 2017 • edited Loading

Firefox (nightly)

All results

w3c-bots commented Jun 21, 2017 • edited Loading

Sauce (safari)

All results

w3c-bots commented Jun 21, 2017 • edited Loading

Chrome (unstable)

All results

w3c-bots commented Jun 21, 2017 • edited Loading

Sauce (MicrosoftEdge)

All results

wanderview commented Nov 17, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wanderview commented Nov 22, 2017

wanderview commented Nov 22, 2017

wanderview commented Nov 22, 2017

w3c-bots commented Nov 22, 2017 • edited Loading

Build PASSED

jakearchibald left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wanderview commented Nov 23, 2017

mfalken left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wanderview commented Jun 21, 2017 •

edited by wpt-pr-bot

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Jun 21, 2017 •

edited

Loading

w3c-bots commented Nov 22, 2017 •

edited

Loading