Bug 237731 – [Cocoa] Add a FairPlay keyserver and EME tests to LayoutTests

• Component: New Bugs
• Status: RESOLVED FIXED

[Jer Noble]

[Cocoa] Add a FairPlay keyserver and EME tests to LayoutTests

[Jer Noble]

Created attachment 454402 [details]
Patch

[Eric Carlson]

Comment on attachment 454402 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=454402&action=review

> LayoutTests/ChangeLog:23
> +        and persistable licenses. It currently only supports "modern" EME key requests, but acn

s/acn/can/

> LayoutTests/ChangeLog:261
> [phone]  Jer Noble  <[e-mail]>
> +
> +        [Cocoa] Add a FairPlay keyserver and EME tests to LayoutTests
> +        https://bugs.webkit.org/show_bug.cgi?id=237731
> +
> +        Reviewed by NOBODY (OOPS!).

Double ChangeLog

[Jonathan Bedard]

Created attachment 454500 [details]
Patch

[Jer Noble]

Created attachment 454556 [details]
Patch

[Jer Noble]

Created attachment 454558 [details]
Patch

[Jer Noble]

```
[Sun Mar 13 15:35:39.040034 2022] [cgi:error] [pid 847] [client 127.0.0.1:62020] AH01215: ModuleNotFoundError: No module named 'OpenSSL': /Volumes/Data/worker/macOS-Release-WK2-Stress-Tests-EWS/build/LayoutTests/http/tests/media/fairplay/keyserver/index.py, referer: http://127.0.0.1:8000/media/fairplay/fps-mse-unmuxed-multiple-keys.html
```

Looks like we need to auto install the OpenSSL package as well.

[Jonathan Bedard]

Comment on attachment 454558 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=454558&action=review

> LayoutTests/http/tests/media/fairplay/keyserver/keyserver/__init__.py:40
> +    from webkitpy.autoinstalled.cryptography.hazmat.primitives.ciphers import modes

`from webkitpy.autoinstalled import cryptography` should be sufficient for this.

Also, once you do this once, "from cryptography.hazmat.primitives.ciphers import modes" should work.

[Jer Noble]

Created attachment 454595 [details]
Patch

[Jer Noble]

Created attachment 454777 [details]
Patch

[Jer Noble]

Created attachment 454836 [details]
Patch

[Jer Noble]

Getting closer! The key server is now running correctly! Now to investigate the cause of the failing tests. I suspect it's just a harness problem.

[Peng Liu]

Comment on attachment 454836 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=454836&action=review

> LayoutTests/ChangeLog:23
> +        and persistable licenses. It currently only supports "modern" EME key requests, but acn

s/acn/can

> LayoutTests/http/tests/media/fairplay/eme2016.js:15
> +            method: 'POST',

Nit. The indention.

> LayoutTests/http/tests/media/fairplay/fps-hls-key-rotation.html:61
> +                        method: 'POST',

Ditto.

> LayoutTests/http/tests/media/fairplay/fps-hls-update-reject.html:52
> +                method: 'POST',

Ditto.

> LayoutTests/http/tests/media/fairplay/fps-hls.html:53
> +                method: 'POST',

Ditto.

> LayoutTests/http/tests/media/fairplay/fps-init-data-cenc.html:48
> +                    method: 'POST',

Ditto.

> LayoutTests/http/tests/media/fairplay/fps-init-data-sinf.html:45
> +                method: 'POST',

Ditto.

> LayoutTests/http/tests/media/fairplay/keyserver/keyserver/test_DFunction.py:34
> +class TestDFunction(unittest.TestCase):

Will bots run the test? :-)

[Jer Noble]

Created attachment 454868 [details]
Patch

[Jer Noble]

Created attachment 454877 [details]
Patch

[Jer Noble]

Created attachment 454879 [details]
Patch

[Xabier Rodríguez Calvar]

I think there are some changes in this patch that need to be removed as they are unrelated and there is no changelog for them, right?

[Jer Noble]

(In reply to Xabier Rodríguez Calvar from comment #16)
> I think there are some changes in this patch that need to be removed as they
> are unrelated and there is no changelog for them, right?

Yes, I should remove the r? bit on this patch; I changed some of the promise rejections inside MediaKeySession to get more information about why the testes were failing on the stress bot.

[Radar WebKit Bug Importer]

<rdar://problem/90453705>

[Jer Noble]

Okay, the Mac-wk2-stress test bot is failing for a very interesting reason: we leak every RemoteCDMInstanceSessionProxy we create in the GPU Process. So after about 800 tests, FairPlay stops allowing additional sessions to be created. I'll include a fix with the next patch.

[Jer Noble]

Created attachment 455213 [details]
Patch

[Jer Noble]

Created attachment 455565 [details]
Patch

[Jer Noble]

Created attachment 455584 [details]
Patch

[Jer Noble]

Pull request: https://github.com/WebKit/WebKit/pull/2786

[EWS]

Committed 253230@main (809ad82ca6e2): <https://commits.webkit.org/253230@main>

Reviewed commits have been landed. Closing PR #2786 and removing active labels.