Black screen on ABR changes

[zubcoco]

Version

Media3 main branch

More version details

Issue appear with version 1.1.0 (OK with version 1.0.2)
Same on ExoPlayer, present on v2.19.0 and not with v.2.17.0

Devices that reproduce the issue

Problem is on our Box Vantiva/Technicolor (broadcom) but also reproduced on 2 differents Sony TVs.

Devices that do not reproduce the issue

Problem is not reproduced on Google TV, not on Nvidia Shield and not on Android Studio Emulator.
And also not reproduced on Samsung mobiles.

Reproducible in the demo app?

Yes

Reproduction steps

1. In the demo app, we need to set the initial bitrate estimate to 1 MBPs (to ensure the player will change profile after a while). I use the default band width meter that we can pass to the exoplayer builder.
2. Play the stream that I sent you by mail

Expected result

The media start with a low profile. After a while it play with the highest profile and the transition should be fluent.

Actual result

The media start with a low profile. After a while it play with the highest profile but during the transition we have a black screen.

Media

I'll send you the media to test.

Bug Report

• You will email the zip file produced by adb bugreport to android-media-github@google.com after filing this issue.

[tonihei]

Since you are able to replicate the problem on your device, do you think you can search for the change that broke this behavior? I think you'd need to depend on ExoPlayer/Media3 locally (see instructions here or here, depending on which namespace you are using) and then you can move the code in a more granular way to find the problematic change.

[zubcoco]

Hello,
I've found the problematic commit: 8a73d8d.

If I compile a version and I revert all lines of this commit, the problem is solve.

Would you like me to do some special tests to help your analysis?

[zubcoco]

The problem is in NalUnitUtil class, in parseSpsNalUnitPayload function.
When I comment these lines :

      if (data.readBit()) { // overscan_info_present_flag
        data.skipBit(); // overscan_appropriate_flag
      }
      if (data.readBit()) { // video_signal_type_present_flag
        data.skipBits(3); // video_format
        colorRange =
            data.readBit() ? C.COLOR_RANGE_FULL : C.COLOR_RANGE_LIMITED; // video_full_range_flag
        if (data.readBit()) { // colour_description_present_flag
          int colorPrimaries = data.readBits(8); // colour_primaries
          int transferCharacteristics = data.readBits(8); // transfer_characteristics
          data.skipBits(8); // matrix_coeffs

          colorSpace = ColorInfo.isoColorPrimariesToColorSpace(colorPrimaries);
          colorTransfer =
              ColorInfo.isoTransferCharacteristicsToColorTransfer(transferCharacteristics);
        }
      }

the problem no longer appears.

[tonihei]

It's probably not that these lines itself are wrong, it's more likely that this causes some configuration change when selecting or configuring codecs for playback.

Could you attach an EventLogger and create a bugreport with and without this change and send a link to android-media-github@google.com with "Issue 1158" in the subject? Then we may be able to see what different behavior this causes exactly. If my suspicion is correct, then maybe we can add a workaround to avoid the broken configuration.

[zubcoco]

Mail sent with requested informations.
Thanks a lot.

[tonihei]

Thanks. Assuming you still have the local checkout, could you try to comment out all four maybeSet... calls in this method:

media/libraries/common/src/main/java/androidx/media3/common/util/MediaFormatUtil.java

Line 256 in d13a0f4

maybeSetInteger(format, MediaFormat.KEY_COLOR_TRANSFER, colorInfo.colorTransfer);

to see if that fixes your problem?

[zubcoco]

No, the problem is still present with these commented lines.

[tonihei]

Thanks for the confirmation and sharing the bugreports! I think I see the problem now and will share another fix proposal soon (sorry for constantly asking to test things out, but I don't have a reproduction setup myself).

[tonihei]

The commit linked above should fix the problem. Could you verify this again with your setup? Thanks!

[zubcoco]

Yes, if I compile on the main branch I don't see the problem. Thanks a lot !