1075492 - ./mach webapprt-test-chrome no longer works on OSX due to v2 signature changes

Status: RESOLVED FIXED

(Testing :: General, defect)

Description

[Stephen A Pohl [:spohl]]

As stated in summary.

Comment 1

[Stephen A Pohl [:spohl]]

Attachment: Patch

I must say that I'm shocked that NSBundle's pathForAuxiliaryExecutable ever worked, but this is due to the fact that OSX will treat any folder under Contents/MacOS as an application bundle. In this case, it was able to successfully retrieve the webapprt folder even though it's not actually an application bundle.

Now that the webapprt folder is in it's correct place (Contents/Resources), NSBundle's pathForResource will do the right thing.

Comment 2

[Stephen A Pohl [:spohl]]

Attachment: Output before any v2 signing changes.

Comment 3

[Stephen A Pohl [:spohl]]

Attachment: Output after v2 signing changes and this patch applied.

Comment 4

[Stephen A Pohl [:spohl]]

Although the patch is seemingly doing the right thing, I'm seeing something odd between a run before any of the v2 signing stuff landed (local build compiled off of m-c cset 20c7e70e1b1a) and after the v2 signing changes with the patch here applied: We seem to crash pretty quickly in the build before the v2 signing stuff (see attachment 8498210 [details]). However, with the v2 signing stuff landed and the patch here applied, we no longer crash but end up with two test failures (browser_mozpay.js and browser_webperm.js). I would have expected the two runs to be virtually identical.

Marco, does this seem right to you? Were there webapprt fixes that landed after the v2 signing patches?

Comment 5

[Stephen A Pohl [:spohl]]

(In reply to Stephen Pohl [:spohl] from comment #4)
> However, with the v2 signing stuff landed and the patch here applied, we
> no longer crash but end up with two test failures (browser_mozpay.js and
> browser_webperm.js).

... see attachment 8498211 [details] for this run.

Comment 6

[Marco Castelluccio [:marco]]

During the last few weeks, there have been three regressions (four, if you include the regression from the v2 signing changes).
1) All tests started to time out (this is what you're seeing now in attachment 8498210 [details], probably the harness caused the crash).
2) browser_mozpay.js started to (intermittently?) fail
3) browser_webperm.js started to consistently fail (bug 1073790)

The patch that fixed (1), in bug 1072798, has probably landed at the same time as your patch for the v2 signing changes.

Comment 7

[Stephen A Pohl [:spohl]]

Great! Sounds like the patch in this bug is still the right one then. Thanks for the confirmation!

Comment 8

[Myk Melez [:myk] [@mykmelez]]

Comment on attachment 8498208 [details] [diff] [review]
Patch

Looks good, r=myk.

Comment 9

[Stephen A Pohl [:spohl]]

https://hg.mozilla.org/projects/oak/rev/2444de3042c1

Waiting for inbound and/or fx-team to reopen to land there.

Comment 10

[Stephen A Pohl [:spohl]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/c772f124fed2

Comment 11

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/c772f124fed2

Comment 12

[Robert Strong (they/them - no direct email)]

Landed on aurora in the Mac V2 signing combined patch in bug 1047584