1187766 - Test loading plugins scenarios with fetch interception

Status: RESOLVED FIXED

(Core :: DOM: Service Workers, defect)

Comment 1

[Noemí Freire (:noemi)]

blocking v1, feel free to change it as needed. Thanks!

Comment 2

[Dimi Lee [:dimi]]

Hi Ben,
I write a web page with following:

<object data="./test.swf"></object>
<embed src="./test.swf"></embed>
<APPLET ALT="Applet" CODE="./test.class" CODEBASE="applets"></APPLET>

But SW won't intercept these resources when load this page, and I try finding if there is spec talking about what kind of resources SW should intercept but I cannot find it. So I am not sure if current behavior is correct.
Could you provide some suggestion ? Thanks for your help!

Comment 3

[Josh Matthews [:jdm]]

According to http://slightlyoff.github.io/ServiceWorker/spec/service_worker/index.html#on-fetch-request-algorithm we should not intercept "potential client requests" (https://fetch.spec.whatwg.org/#potential-client-request) so <object> and <embed> are behaving correctly (also see bug 1168676). I do not see any such special case for <applet>, however.

Comment 4

[Ben Kelly [:bkelly, not reviewing]]

I'll ask Google what they do in chrome this week.  They did mention they do not intercept network requests for "plugins" at the meeting last week.  Not sure if that includes applets.

Comment 5

[Ben Kelly [:bkelly, not reviewing]]

For now, can you write tests that verify object, embed, and applet are NOT intercepted?  I think its good to have those to protect against accidentally intercepting them in the future.

Thanks!

Comment 6

[Dimi Lee [:dimi]]

Attachment: Patch v1

I found <applet> doesn't go through the path[1] this bug trying to check unless i miss something. When loading <applet> it will create new channel through DTD case[2] but with the flow i don't understand very well.

So in this patch I only check <object> & <embed>.

[1]http://hg.mozilla.org/mozilla-central/file/0b901209064c/dom/base/nsObjectLoadingContent.cpp#l2482
[2]http://hg.mozilla.org/mozilla-central/file/0b901209064c/parser/htmlparser/nsExpatDriver.cpp#l806

Comment 7

[Ben Kelly [:bkelly, not reviewing]]

Comment on attachment 8640435 [details] [diff] [review]
Patch v1

Review of attachment 8640435 [details] [diff] [review]:
-----------------------------------------------------------------

Overall looks good, but I'm not sure about the test completion condition.  r=me with that fixed.  Thanks!

::: dom/workers/test/serviceworkers/fetch/plugin/plugins.html
@@ +25,5 @@
> +      ok(false, "<object> should not be intercepted");
> +    } else if (e.data.context === "embed") {
> +      ok(false, "<embed> should not be intercepted");
> +    } else if (e.data.context === "fetch") {
> +      navigator.serviceWorker.removeEventListener("message", onMessage);

Won't the plugins.html fetch event trigger this before the object and embed elements can be added?  Can you check the URL as well to make sure its completing on the foo.txt fetch?

Comment 8

[Dimi Lee [:dimi]]

After applying patch to latest code, request.context become undefined.
Will check the root cause.

Comment 9

[Dimi Lee [:dimi]]

(In reply to Dimi Lee[:dimi][:dlee] from comment #8)
> After applying patch to latest code, request.context become undefined.
> Will check the root cause.

Ah, I found this is because of Bug 1188062

Comment 10

[Dimi Lee [:dimi]]

Encounter try server error :
LoadObject called while not bound to an active document: 'Not Reached'

Comment 11

[Dimi Lee [:dimi]]

Attachment: Patch v2

Address Ben's comment, add check for test completion

Comment 12

[Dimi Lee [:dimi]]

Try result :
https://treeherder.mozilla.org/#/jobs?repo=try&revision=178d94dd10f3

Comment 13

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/8693948df8f0

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/8693948df8f0