fix: Fix legacy codec support by rewriting codec metadata

[cuyl]

This fixes legacy codec support by rewriting the codec metadata in the Stream objects. After capability checking, the converted codec information will be used.

[cuyl]

Test stream: https://cdn3.wowza.com/2/THFDbHhHVUJJVC9k/MEg5MFJH/hls/v7l2kt3h/playlist.m3u8

[github-actions]

Incremental code coverage: 83.33%

[joeyparrish]

I'm going to need more information to review this. Is there an associated issue filed? Why is this more correct than what we did before?

[cuyl]

@joeyparrish It's because other code will use the codecs from streams.codecs to call shaka.media.Capabilities.isTypeSupporte which can cause error.
ref lines: 361, 368

shaka-player/lib/media/media_source_engine.js

Lines 348 to 376 in 15232dd

	async init(streamsByType, sequenceMode=false) {
	const ContentType = shaka.util.ManifestParserUtils.ContentType;
	await this.mediaSourceOpen_;
	this.sequenceMode_ = sequenceMode;
	for (const contentType of streamsByType.keys()) {
	const stream = streamsByType.get(contentType);
	goog.asserts.assert(
	shaka.media.MediaSourceEngine.isStreamSupported(stream),
	'Type negotiation should happen before MediaSourceEngine.init!');
	let mimeType = shaka.util.MimeUtils.getFullType(
	stream.mimeType, stream.codecs);
	if (contentType == ContentType.TEXT) {
	this.reinitText(mimeType, sequenceMode);
	} else {
	const forceTransmux = this.config_.forceTransmux;
	if ((forceTransmux ||
	!shaka.media.Capabilities.isTypeSupported(mimeType)) &&
	shaka.media.Transmuxer.isSupported(mimeType, contentType)) {
	this.transmuxers_[contentType] =
	new shaka.media.Transmuxer(mimeType);
	mimeType =
	shaka.media.Transmuxer.convertCodecs(contentType, mimeType);
	}
	const type = mimeType + this.config_.sourceBufferExtraFeatures;
	const sourceBuffer = this.mediaSource_.addSourceBuffer(type);

[cuyl]

Feel free to try the test stream: https://cdn3.wowza.com/2/THFDbHhHVUJJVC9k/MEg5MFJH/hls/v7l2kt3h/playlist.m3u8

[cuyl]

So basically, I just store the converted correct formatted codecs for further use.

[joeyparrish]

Thanks. As a tip for next time, that all would have been useful information to include in code comments or the PR description.

[joeyparrish]

Based on your explanation, I took the liberty of writing a PR description (now for the commit log, but would have been helpful for reviewers), as well as code comments (which will be helpful for future readers of that code to avoid "cleaning up" and reverting your change).

Even better would have been to add a regression test, in case this logic is some day replaced in rewrite without preserving all the details.

[cuyl]

@joeyparrish Thanks, Sorry for the inconvenience .

[joeyparrish]

I'll merge this once the test pass is complete. (Or the other maintainers can do it if I'm too slow.)

[joeyparrish]

@joeyparrish Thanks, Sorry for the inconvenience .

No problem. I hope the explanations are helpful, and that you'll continue to contribute to Shaka Player. We appreciate your contribution!

[joeyparrish]

We have a failing test case on Tizen since this was merged:

https://github.com/shaka-project/shaka-player/actions/runs/3854362680/jobs/6568932949