Bug 229219 – [GLIB][GStreamer] test http/tests/security/webaudio-render-remote-audio-allowed-crossorigin-redirect.html is a flaky timeout since r278004

• Status: RESOLVED
• Resolution: FIXED
• Component: Media

[Arcady Goldmints-Orlov]

The changes in r278004 appear to be causing this test to timeout, but only some of the time.

[Arcady Goldmints-Orlov]

Created attachment 435734 [details]
Gardening Patch

[EWS]

Committed r281179 (240625@main): <https://commits.webkit.org/240625@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 435734 [details].

[Radar WebKit Bug Importer]

<rdar://problem/82059333>

[Arcady Goldmints-Orlov]

Reopening to track the fix for this issue.

[Enrique Ocaña]

Created attachment 440395 [details]
Patch

[Philippe Normand]

Comment on attachment 440395 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=440395&action=review

> Source/WebCore/platform/audio/gstreamer/AudioSourceProviderGStreamer.cpp:325
> +    // Some intermediate bins are eating up the EOS message posted to the bus of the inner bin that

I wonder if setting the message-forward property to TRUE on m_audioSinkBin could help with that? Have you tried it?

[Enrique Ocaña]

Comment on attachment 440395 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=440395&action=review

>> Source/WebCore/platform/audio/gstreamer/AudioSourceProviderGStreamer.cpp:325
>> +    // Some intermediate bins are eating up the EOS message posted to the bus of the inner bin that
> 
> I wonder if setting the message-forward property to TRUE on m_audioSinkBin could help with that? Have you tried it?

I had tried that before, but it wasn't successful and I didn't committed the code on my work branch, so I couldn't recheck now to answer your comment.

I've just written code now to not only set message-forward on m_audioSinkBin, but also on any other parent bin up to the topmost GstPlaybin element, and I ensured that the property was properly set by checking the pipeline dumps. Still, the GST_MESSAGE_EOS isn't received by MediaPlayerPrivateGStreamer::handleMessage().

[Philippe Normand]

Per the doc, you should receive "element" messages on the bus:

  /**
   * GstBin:message-forward:
   *
   * Forward all children messages, even those that would normally be filtered by
   * the bin. This can be interesting when one wants to be notified of the EOS
   * state of individual elements, for example.
   *
   * The messages are converted to an ELEMENT message with the bin as the
   * source. The structure of the message is named `GstBinForwarded` and contains
   * a field named `message` that contains the original forwarded #GstMessage.
   */

[Enrique Ocaña]

I'm only able to receive the forwarded message (as an element message, as you pointed) only in the case where the test was already succeeding (with gst logs enabled here). In the case where the test fails (with gst logs disabled here) I don't get it. Enabling and disabling logs has some influence in the reproducibility of this problem, so if must be some race condition buried deep somewhere in GStreamer.

I added some debug printfs. I just print "EOS" when appsink emits EOS, but I've disabled the code to forward it to the main pipeline (the code in my previously submitted patch). The code to enable message-forward in all the bins up to GstPipeline was added. This log is from the successful case:

### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from autoaudiosink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from audio-sink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from abin
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from playsink
### WebCore::AudioSourceProviderGStreamer::handleNewDeinterleavePad(GstPad*)::<lambda(GstElement*, GstElement*)>: EOS
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from appsink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from autoaudiosink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from audio-sink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from abin
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from playsink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): media-player-0 has EOS
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GST_MESSAGE_EOS

This other log is from the failing case:

### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from autoaudiosink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from audio-sink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from abin
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from playsink
### WebCore::AudioSourceProviderGStreamer::handleNewDeinterleavePad(GstPad*)::<lambda(GstElement*, GstElement*)>: EOS
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message eos from appsink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-start from autoaudiosink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-start from audio-sink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-start from abin
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-start from playsink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from autoaudiosink1
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from audio-sink
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from abin
### void WebCore::MediaPlayerPrivateGStreamer::handleMessage(GstMessage*): GstBinForwarded element message async-done from playsink

As you see, relying on message-forward isn't enough.

[EWS]

Committed r284365 (243150@main): <https://commits.webkit.org/243150@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 440395 [details].

[Arcady Goldmints-Orlov]

Since the test is a flake, I will take a look at the next few test runs to confirm before I remove it from the expectations.

[Enrique Ocaña]

(In reply to Arcady Goldmints-Orlov from comment #11)
> Since the test is a flake, I will take a look at the next few test runs to
> confirm before I remove it from the expectations.

Argh! I forgot to remove the flake from the expectations. I'm almost sure that now it's not a flake anymore, but better have a look to it for a while, yes. Thanks!