1239678 - Mach artifact builds are half-broken on Windows due to missing browsercomps.dll

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[:Gijs (he/him)]

Because this list of patterns:

    package_artifact_patterns = {
        'firefox/dependentlibs.list',
        'firefox/platform.ini',
        'firefox/application.ini',
        'firefox/*.dll',
        'firefox/*.exe',
    }

does not match firefox/browser/components/browsercomps.dll

but even adding 'firefox/**/*.dll' doesn't help because the basename modification that bug 1238320 made means we'll install it in the root of the install dir, instead of in browser/components/ where it belongs.

Comment 1

[:Gijs (he/him)]

Attachment: MozReview Request: Bug 1239678 - fix dll inclusion pattern on Windows and the placement of nested dlls like browsercomps and clearkey, r?nalexander

Review commit: https://reviewboard.mozilla.org/r/30899/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/30899/

Comment 2

[:Gijs (he/him)]

It would be quite neat if we could write some kind of test for these builds actually working...

Comment 3

[Nick Alexander :nalexander [he/him]]

Comment on attachment 8707870 [details]
MozReview Request: Bug 1239678 - fix dll inclusion pattern on Windows and the placement of nested dlls like browsercomps and clearkey, r?nalexander

https://reviewboard.mozilla.org/r/30899/#review27699

Oops.  Thanks Gijs!

Comment 4

[Pulsebot]

https://hg.mozilla.org/integration/fx-team/rev/8017627448ca

Comment 5

[Chris Manchester (limited bugmail, email directly)]

(In reply to :Gijs Kruitbosch from comment #2)
> It would be quite neat if we could write some kind of test for these builds
> actually working...

Sorry about the fallout here -- I should have checked with someone before landing this instead of assuming the testing I did was sufficient.

Comment 6

[:Gijs (he/him)]

(In reply to Chris Manchester [:chmanchester] from comment #5)
> (In reply to :Gijs Kruitbosch from comment #2)
> > It would be quite neat if we could write some kind of test for these builds
> > actually working...
> 
> Sorry about the fallout here -- I should have checked with someone before
> landing this instead of assuming the testing I did was sufficient.

No worries. We should really have tests for this.

Nick, can we stand this up as a taskcluster-type job that depends on a regular build having happened, or something, and smoketest that it starts and loads about:home or something? Would probably be good input into bug 1237688 anyway. Or is there a reason that's not a good idea? I might be able to look into it next week.

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/8017627448ca

Comment 8

[Nick Alexander :nalexander [he/him]]

(In reply to :Gijs Kruitbosch from comment #6)
> (In reply to Chris Manchester [:chmanchester] from comment #5)
> > (In reply to :Gijs Kruitbosch from comment #2)
> > > It would be quite neat if we could write some kind of test for these builds
> > > actually working...
> > 
> > Sorry about the fallout here -- I should have checked with someone before
> > landing this instead of assuming the testing I did was sufficient.
> 
> No worries. We should really have tests for this.
> 
> Nick, can we stand this up as a taskcluster-type job that depends on a
> regular build having happened, or something, and smoketest that it starts
> and loads about:home or something? Would probably be good input into bug
> 1237688 anyway. Or is there a reason that's not a good idea? I might be able
> to look into it next week.

This is not really easy, given the 3 platforms and the fact that some builds and tests are driven out of buildbot.  However, I've filed Bug 1240149 to track it.