fix: run callbacks after setting timestampOffset

[squarebracket]

Description

Long story short, I hit a condition where non-muxed HLS wouldn't start playing sometimes. This is because it's possible to end up in a condition where one SourceUpdater gets both a timestampOffset and an appendBuffer call before the other SourceBuffer is added to the MediaSource. Since timestampOffset queues a callback, but does not trigger an updateend on the SourceBuffer, the already-queued appendBuffer call is never processed, and thus a SegmentLoader ends up waiting forever.

For easy reproduction of the problem, you can simply delay the creation of the audio SourceBuffer, like so:

diff --git i/src/source-updater.js w/src/source-updater.js
index b89993d..f0e8c4b 100644
--- i/src/source-updater.js
+++ w/src/source-updater.js
@@ -35,7 +35,13 @@ export default class SourceUpdater {
       mediaSource.addEventListener(
         'sourceopen', this.createSourceBuffer_.bind(this, mimeType, sourceBufferEmitter));
     } else {
-      this.createSourceBuffer_(mimeType, sourceBufferEmitter);
+      if (mimeType === 'audio/mp2t; codecs="mp4a.40.2"') {
+        setTimeout(() => {
+          this.createSourceBuffer_(mimeType, sourceBufferEmitter);
+        }, 2000);
+      } else {
+        this.createSourceBuffer_(mimeType, sourceBufferEmitter);
+      }
     }
   }

Specific Changes proposed

Simply call this.runCallback_() in the queued callback so that any other queued callbacks are run. I'm not sure if this is bad form; perhaps this.runCallback_() should be inside a setTimeout()?

Note: I'm adding the fix on a separate commit so you can see the change from fail -> pass.

Requirements Checklist

• Feature implemented / Bug fixed
• If necessary, more likely in a feature request than a bug fix
  • Unit Tests updated or fixed
  • Docs/guides updated
  • Example created (starter template on JSBin)
• Reviewed by Two Core Contributors

[squarebracket]

Rebased to my fork's master instead of upstream 🤦‍♂️

Let's try this again.

[gkatsev]

Seems like runCallback_ gets called directly in a few places, so, don't have a problem with it being called here.
LGTM.

[gkatsev]

tested, looks fine