1435781 - fix fetch-event.https.html to not timeout by using a content-type for reflected post response

Status: RESOLVED FIXED

(Core :: DOM: Service Workers, enhancement, P2)

Description

[Ben Kelly [:bkelly, not reviewing]]

We started timing out fetch-event.https.html in bug 1435337.  I want to fix this so we don't lose the test coverage.

The issue is that the test now reflects a POST request body back to the browser, but without any content-type.  This means the browser has to try to sniff the type of response and firefox ends up opening a download box.

This test is not really about sniffing, but whether the body was sent correctly to the server or not.  Therefore I'm going to fix this by just setting a text/plain content-type on the response.

Comment 1

[Ben Kelly [:bkelly, not reviewing]]

Attachment: Make fetch-event.https.html POST tests set a content-type when reflecting the upload body back in the response. r=bz

Comment 2

[Ben Kelly [:bkelly, not reviewing]]

I cleaned up the WPT .ini expectations quite a bit.  Lots of old stuff in there.  So lets run a try build to make sure I didn't break it on some platform:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f1c78e5d7b212a214b530cc830e3a163780df601

Comment 3

[Ben Kelly [:bkelly, not reviewing]]

Comment on attachment 8948451 [details] [diff] [review]
Make fetch-event.https.html POST tests set a content-type when reflecting the upload body back in the response. r=bz

Boris, this WPT test is trying to verify that the POST upload body is sent to the server correctly if the service worker gets a FetchEvent, but does not call FetchEvent.respondWith().  Its doing this by reflecting the body back from the server .py as a response body.  The body ends up as the content of an iframe load.

We were timing out the test because the server .py does not set a content-type on the response.  This forces us to try to sniff the content type.  For whatever reason, this sniffing resulted in FF opening a download dialog instead of loading the content in the iframe.

The simplest fix here is to just set a content-type of text/plain.

While fixing the .ini expectations I also removed an old pref we no longer need.  The test used to contain a check for a case that is not permitted by default browser configurations.  That test case has been removed, so we don't need the pref any more.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8948451 [details] [diff] [review]
Make fetch-event.https.html POST tests set a content-type when reflecting the upload body back in the response. r=bz

r=me

Fwiw, I expect the issue was that we sniffed the URL filename and decided to open it as a Python file.  That comes before the text-or-binary sniff.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/370e267e1605
Make fetch-event.https.html POST tests set a content-type when reflecting the upload body back in the response. r=bz

Comment 6

[Narcis Beleuzu [:NarcisB]]

https://hg.mozilla.org/mozilla-central/rev/370e267e1605

Comment 7

[Web Platform Test Sync Bot (Matrix: #interop:mozilla.org)]

Created web-platform-tests PR https://github.com/w3c/web-platform-tests/pull/9608 for changes under testing/web-platform/tests

Comment 8

[Web Platform Test Sync Bot (Matrix: #interop:mozilla.org)]

Upstream PR merged

Comment 9

[Anny Gakhokidze [:annyG]]

Attachment: Bug 1435781 - Part 1: Prevent content sniffing of the response in original echo-content.py,

Review commit: https://reviewboard.mozilla.org/r/246750/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/246750/

Comment 10

[Ben Kelly [:bkelly, not reviewing]]

Comment on attachment 8980591 [details]
Bug 1435781 - Part 1: Prevent content sniffing of the response in original echo-content.py,

https://reviewboard.mozilla.org/r/246750/#review252892

Wrong bug number?  This bug was closed a while ago...