No obvious culprit, so I'm going to disable the test.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/115ba21da04c63f6c5eb47f2f333f401737c1c68

commit 115ba21da04c63f6c5eb47f2f333f401737c1c68
Author: Adithya Srinivasan <adithyas@chromium.org>
Date: Wed May 30 17:06:51 2018

Disable Playback_VideoOnly_MP4_MDAT on Mac

Bug: 847881
Change-Id: Ida114c8e28d260642e974ab89cd2a2f96f912608
No-Try: true
Tbr: sergeyu@chromium.org
Reviewed-on: https://chromium-review.googlesource.com/1079379
Commit-Queue: Adithya Srinivasan <adithyas@chromium.org>
Reviewed-by: Adithya Srinivasan <adithyas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#562877}
[modify] https://crrev.com/115ba21da04c63f6c5eb47f2f333f401737c1c68/chrome/browser/media/encrypted_media_browsertest.cc

This appears to be browser_tests taking a long time to start up. From [1]:

[ RUN      ] MSE_ClearKey/EncryptedMediaTest.Playback_VideoOnly_MP4_MDAT/0
[27225:2563:0529/232707.580277:WARNING:notification_platform_bridge_mac.mm(510)] AlertNotificationService: XPC connection invalidated.
2018-05-29 23:27:07.694 browser_tests[27225:408511] *** Owner supplied to -[NSTrackingArea initWithRect:options:owner:userInfo:] referenced a deallocating object. Tracking area behavior is undefined. Break on NSTrackingAreaDeallocatingOwnerError to debug.
[27225:775:0529/232733.949084:WARNING:CONSOLE(0)] "Styling master document from stylesheets defined in HTML Imports is deprecated. Please refer to https://goo.gl/EGXzpw for possible migration paths.", source:  (0)
[27225:775:0529/232734.756672:INFO:CONSOLE(273)] "23:27:34.662 -  Using ClearKeyPlayer", source: http://127.0.0.1:59787/eme_player_js/utils.js (273)
[27225:775:0529/232734.756773:INFO:CONSOLE(273)] "23:27:34.663 -  Registering video event handlers.", source: http://127.0.0.1:59787/eme_player_js/utils.js (273)
...snip...
[27225:775:0529/232735.910063:INFO:CONSOLE(273)] "23:27:35.909 -  timeupdate @ 1.003425", source: http://127.0.0.1:59787/eme_player_js/utils.js (273)
[27225:775:0529/232735.910163:INFO:CONSOLE(273)] "23:27:35.909 -  waiting for video to end.", source: http://127.0.0.1:59787/eme_player_js/utils.js (273)
[27225:775:0529/232736.691521:INFO:CONSOLE(273)] "23:27:36.690 -  heartbeat #2", source: http://127.0.0.1:59787/eme_player_js/utils.js (273)
BrowserTestBase received signal: Terminated: 15. Backtrace:

So the JavaScript ran for less than 3 seconds. The video played is ~2.5 seconds long, so it was almost done when the test timed out.

However, the first message is from 23:27:07, while the JavaScript starts at 23:27:34. The test really only sets up a local web server to serve the JavaScript pages and the video (all from the same directory). No idea why the setup takes so long.

The code does log when it starts the web server and when it starts the test, Running it locally (on Linux) I get:
[101137:101137:0530/111415.671195:INFO:media_browsertest.cc(32)] 11:14:15.671 Starting HTTP server
[101137:101137:0530/111415.677377:INFO:media_browsertest.cc(48)] 11:14:15.677 Running test URL: http://127.0.0.1:41109/eme_player.html?...
However, these messages don't appear in the bot output.

From the other failure [2]:
[ RUN      ] MSE_ClearKey/EncryptedMediaTest.Playback_VideoOnly_MP4_MDAT/0
[34673:3843:0530/075252.562882:WARNING:notification_platform_bridge_mac.mm(510)] AlertNotificationService: XPC connection invalidated.
2018-05-30 07:52:52.770 browser_tests[34673:2198662] *** Owner supplied to -[NSTrackingArea initWithRect:options:owner:userInfo:] referenced a deallocating object. Tracking area behavior is undefined. Break on NSTrackingAreaDeallocatingOwnerError to debug.
BrowserTestBase received signal: Terminated: 15. Backtrace:

In this case the JavaScript never ran. Out of the 4 attempts, only 1 has logs from the JavaScript.

From the flakiness dashboard linked in #0, the test had TIMEOUT TIMEOUT PASS @ 5/29/2018 11:34:12 PM (build 3057), and TIMEOUT TIMEOUT TIMEOUT TIMEOUT @ 5/30/2018 8:35:56 AM (build 3070). Those are the only failures since 5/26/2018 10:02:29 PM (build 2974).

Out of the 100 runs listed, 68 complete in 10s or less. However, 8 take 20s or longer (3 * 20s, 24s, 26s, 27s, 2 * 30s). Bot vm73-m9 is the only one in the pool, so they all ran on the same bot.

On the other Mac bots (Mac 10.10, 10.11, and 10.12), the max run time of this test is 8s, 6s, and 5s respectively.

The MSE_ExternalClearKey versions of the same test have always passed on these same runs. The 10.13 runs are slower (max = 21s) compared with the other 3 (8s, 7s, 5s).


[1] https://logs.chromium.org/v/?s=chromium%2Fbuildbucket%2Fcr-buildbucket.appspot.com%2F8945125164712097008%2F%2B%2Fsteps%2Fbrowser_tests%2F0%2Fstdout
[2] https://logs.chromium.org/v/?s=chromium%2Fbuildbucket%2Fcr-buildbucket.appspot.com%2F8945093405151370304%2F%2B%2Fsteps%2Fbrowser_tests%2F0%2Fstdout