1953005 - Fix more tests that fail when New Tab is packaged as a built-in addon

Status: RESOLVED FIXED

(Firefox :: New Tab Page, task, P1)

Description

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

There are a few tests outside of browser/extensions/newtab that are failing, and upon investigation of this, I have some additional changes to introduce in advance of packaging newtab as a built-in addon.

Comment 1

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Make test_AboutNewTabRedirector.js more reliable in built-in addon case. r?#home-newtab-reviewers!

It looks like attempting to create an nsIChannel / nsILoadInfo right at
the onset of an xpcshell test isn't a great idea and will fail intermittently
with NS_ERROR_NOT_AVAILABLE. This updates the test to create the nsIChannel
and nsILoadInfo in the add_task function instead.

Comment 2

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Have AboutNewTabRedirector wait for resource-mapping init rather than addon init to resume channels. r?willdurand!,#home-newtab-reviewers!

Waiting until the addon database is ready and the addonReady promise resolves
can take far, far too long, and cause regressions in our Talos tests. What
we really care about waiting for is resource-mapping.js having finished its
job to map resource://newtab and chrome://newtab URIs to the right places.

So we just have resource-mapping.js directly inform AboutNewTabRedirector
when that has occurred.

Comment 3

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Exempt programmatically defined newtab resource and chrome URIs from browser_all_files_referenced.js. r?baku!

Comment 4

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Change strategy for chrome URI registration for newtab. r?baku!,#home-newtab-reviewers!

In order to make activity-stream.css accessible to the newtab page, we need to register
it at a chrome URI. We were originally doing this via ComponentManager.addBootstrappedManifestLocation,
which is a holdover from old bootstrapped addons. It turns out that this only works when
the path passed to it is either to an XPI or a "real" directory on the file system. When
packaging newtab as a built-in addon, when not running an out-of-band update, it's hosted
in the omni.ja JAR file, and that's not supported by addBootstrappedManifestLocation.

This patch does two things:

1. It gets rid of the chrome.manifest file, and
2. Has resource-mapping.js call into amIAddonManagerStartup's registerChrome instead.

Notably, however, contentaccessible=yes directives for chrome URIs were being ignored
except for testing environments. That was originally added in bug 1451503 when the
reftest addon was ported to a WebExtension. I've gone ahead and removed the NS_ENSURE
check for being in the automation case.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/f1a6b824064a
Make test_AboutNewTabRedirector.js more reliable in built-in addon case. r=home-newtab-reviewers,amy
https://hg.mozilla.org/integration/autoland/rev/ae06e85459f6
Have AboutNewTabRedirector wait for resource-mapping init rather than addon init to resume channels. r=willdurand,home-newtab-reviewers,amy
https://hg.mozilla.org/integration/autoland/rev/babb52cb2c3b
Exempt programmatically defined newtab resource and chrome URIs from browser_all_files_referenced.js. r=baku

Comment 6

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/f1a6b824064a
https://hg.mozilla.org/mozilla-central/rev/ae06e85459f6
https://hg.mozilla.org/mozilla-central/rev/babb52cb2c3b

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/707865adbdb7
Change strategy for chrome URI registration for newtab. r=home-newtab-reviewers,extension-reviewers,robwu,thecount

Comment 8

[Serban Stanca [:SerbanS]]

https://hg.mozilla.org/mozilla-central/rev/707865adbdb7

Comment 9

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Change how newtab xpcshell tests install built-in newtab addon. r?baku!

This allows the tests to have the newtab built-in addon be installed
and enabled properly when running in automation.

Comment 10

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Adjust browser_all_files_referenced, browser_parsable_css and allowed-dupes for built-in newtab addon case. r?baku!

Comment 11

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Attachment: Bug 1953005 - Make tp5n test work with built-in addons. r?baku!

Without this preference set in the test, none of the built-in addons
(including the newtab build-in addon) will be available during the
test.

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/59115c6c330a
Change how newtab xpcshell tests install built-in newtab addon. r=baku
https://hg.mozilla.org/integration/autoland/rev/0fa2c4c68344
Adjust browser_all_files_referenced, browser_parsable_css and allowed-dupes for built-in newtab addon case. r=baku
https://hg.mozilla.org/integration/autoland/rev/614f78590327
Make tp5n test work with built-in addons. r=baku,perftest-reviewers,sparky

Comment 13

[Serban Stanca [:SerbanS]]

https://hg.mozilla.org/mozilla-central/rev/59115c6c330a
https://hg.mozilla.org/mozilla-central/rev/0fa2c4c68344
https://hg.mozilla.org/mozilla-central/rev/614f78590327

Comment 14

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

With bug 1949511 landed and merged, this is done.

Comment 15

[Mike Conley (:mconley) (:⚙️) (PTO June 8 - 12)]

Mass moving from [hnt-trainhop] to [hnt-trainhop-project] for better JIRA book-keeping.