1736103 - Fix browser_extension_correction.js to work when download improvements are enabled

Status: RESOLVED FIXED

(Firefox :: Downloads Panel, task, P1)

Description

[Molly Howell (she/her) (mostly inactive)]

browser_extension_correction.js in the exthandler tests expects to have the download modal open when a download begins, and much of its logic is based on that assumption. But turning on the browser.download.improvements_to_download_panel pref will disable the modal, so we need to alter the test to work under both scenarios.

Comment 1

[Molly Howell (she/her) (mostly inactive)]

Attachment: Bug 1736103 - Fix the download file extension correction test to not depend on the download modal when it is disabled. r=gijs,mtigley

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/3da967f1e8fa
Fix the download file extension correction test to not depend on the download modal when it is disabled. r=Gijs

Comment 3

[:Gijs (he/him)]

Attachment: Bug 1736103 - fix download file extension correction test some more for the no-download-modal case, r?sfoster

On try and on my machine, after D128648 the test still times out because
for the test_broken_saved_handlerinfo_and_useless_mimetypes test, first
awaiting the task that gets passed means that the download has already
finished by the time we create the download finished promise.

This patch moves creating that promise before the task is started.
It also broadens the conditions in which we await this promise, because
in the case without file extensions and without a dialog, we were otherwise
no longer awaiting anything, which meant that the next test's check for
a finished download would resolve with the previous test's download, causing
confusion/breakage.

Comment 4

[Atila Butkovits]

https://hg.mozilla.org/mozilla-central/rev/3da967f1e8fa

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1350a6c88f94
fix download file extension correction test some more for the no-download-modal case, r=sfoster

Comment 6

[Iulian Moraru]

https://hg.mozilla.org/mozilla-central/rev/1350a6c88f94

Comment 7

[Andra Esanu (needinfo me)]

(In reply to Atila Butkovits from comment #4)

https://hg.mozilla.org/mozilla-central/rev/3da967f1e8fa

== Change summary for alert #32085 (as of Mon, 25 Oct 2021 12:27:54 GMT) ==

Improvements:

Ratio	Test	Platform	Options	Absolute values (old vs new)
5%	build times	windows2012-64	plain taskcluster-c4.4xlarge	2,841.75 -> 2,713.12
3%	build times	windows2012-64	plain taskcluster-c4.4xlarge	2,841.94 -> 2,754.53

For up to date results, see: https://treeherder.mozilla.org/perfherder/alerts?id=32085

Comment 8

[:Gijs (he/him)]

This was a test-only change, so it can't have been responsible for build times improvements.

Comment 9

[Andra Esanu (needinfo me)]

Thanks for noticing, I managed to retrigger some data points and the graph shows the right improvement now.