1178215 - Update libvpx to more recent git version (1.4.x)

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Jan Gerber]

libvpx got quite some updates in git. Would be good to update to a more recent version. Chromium uses e67d45d4ce92468ba193288b59093fef0a502662 with one patch.

Comment 1

[Jan Gerber]

Attachment: Bug1178215_update_update.py.patch

Comment 2

[Jan Gerber]

Attachment: Bug1178215_update_libvpx.patch

Comment 3

[Ralph Giles (:rillian)]

Comment on attachment 8627129 [details] [diff] [review]
Bug1178215_update_update.py.patch

Review of attachment 8627129 [details] [diff] [review]:
-----------------------------------------------------------------

Please split the update.py, patch and README_MOZILLA changes into a separate commit from the config updates which result from running the new update.py, so we have a clear point of reproducibility. r=me with that change.

Comment 4

[Ralph Giles (:rillian)]

Comment on attachment 8627130 [details] [diff] [review]
Bug1178215_update_libvpx.patch

Review of attachment 8627130 [details] [diff] [review]:
-----------------------------------------------------------------

Likewise, r=me with the update.py results split into either this commit, or a third patch.

Thanks for the update.

Comment 5

[Jan Gerber]

Attachment: Bug1178215_update_update.py.patch

Comment 6

[Jan Gerber]

Attachment: Bug1178215_update_libvpx.patch

Comment 7

[Jan Gerber]

now fully untangled, time for a try push?

Comment 8

[Ralph Giles (:rillian)]

Comment on attachment 8627371 [details] [diff] [review]
Bug1178215_update_update.py.patch

Review of attachment 8627371 [details] [diff] [review]:
-----------------------------------------------------------------

::: configure.in
@@ +5592,5 @@
>        AC_MSG_WARN([No assembler or assembly support for libvpx. Using unoptimized C routines.])
>      fi
> +
> +    dnl native libvpx no longer has vpx_mem_set_functions
> +    AC_DEFINE(MOZ_VPX_NO_MEM_REPORTING)

We should patch this back in. I filed bug 1178831 for that.

Comment 9

[Ralph Giles (:rillian)]

Thanks. Pushed to try in https://treeherder.mozilla.org/#/jobs?repo=try&revision=358cfc07c29e

Comment 10

[Ralph Giles (:rillian)]

*** No rule to make target `/home/worker/workspace/gecko/media/libvpx/vp8/common/x86/variance_mmx.c', needed by `variance_mmx.o'.  Stop.

Pushed again with CLOBBER update. Try used to auto-clobber, but maybe that's changed? Looks like the familiar mach bug with moved files.
https://treeherder.mozilla.org/#/jobs?repo=try&revision=ce5c88684b39

Comment 11

[Ralph Giles (:rillian)]

Try looks good.

Comment 12

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/fa9904c62103
https://hg.mozilla.org/integration/mozilla-inbound/rev/a5efc589a3d6

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/fa9904c62103
https://hg.mozilla.org/mozilla-central/rev/a5efc589a3d6

Comment 14

[Ralph Giles (:rillian)]

Comment on attachment 8627373 [details] [diff] [review]
Bug1178215_update_libvpx.patch

Requesting uplift of both patches for Aurora and Beta.

Approval Request Comment
[Feature/regressing bug #]: Bug 1151175
[User impact if declined]: We need to take this for security and stability.
[Describe test coverage new/current, TreeHerder]: Landed on m-c some weeks ago.
[Risks and why]: This is an update to the same third-party snapshot chrome is shipping. I don't expect any issues.
[String/UUID change made/needed]: None.

Comment 15

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8627373 [details] [diff] [review]
Bug1178215_update_libvpx.patch

This update has been on m-c for a week. We're still early enough in Beta to take a bigger change like this.

Ralph - While you don't expect any issues, how will regressions due to this change appear in the browser if they are to appear? Or, what should we test for?

Beta+ Aurora+

Comment 16

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/c3b64ca7c3c1
https://hg.mozilla.org/releases/mozilla-aurora/rev/f49571596d6f

Comment 17

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/5ebea3f7befc
https://hg.mozilla.org/releases/mozilla-beta/rev/2f025dfee1f4

Comment 18

[Ralph Giles (:rillian)]

(In reply to Lawrence Mandel [:lmandel] (use needinfo) from comment #15)

> Ralph - While you don't expect any issues, how will regressions due to this
> change appear in the browser if they are to appear? Or, what should we test
> for?

Regressions would show primarily as video corruption or crashes in desktop-to-desktop webrtc (mozilla hello) sessions. Also the same for webm playback in the <video> element, but that's not broadly used in these firefox versions.

Comment 19

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

We have performed today regression testing using Firefox 40 Beta 3 (BuildID=20150709163524) and focused on Hello functionality and Video/Audio compatibility. There were no issues encountered related to video corruption for either Hello calls between different desktop machines, or Video playback. Detailed test results are available at: https://goo.gl/bWDsSd. 

Marking this as verified for Firefox 40 based on our testing results.