Comment on attachment 684759 [details] [diff] [review] Patch (v1) Review of attachment 684759 [details] [diff] [review]: ----------------------------------------------------------------- Why do you need this? I'd prefer to not do this because we'll then expose platform specific reader's innards to the outside world, it breaks our encapsulation. Could you instead have some kind of factory (which lives in and is exported from content/media) or even just a (static?) function on MediaDecoder, which would return a MediaDecoderReader* appropriate for the MIME type or MediaResource passed in?

For web audio, I need to create decoder reader objects, not decoders. I have a class in content/media/webaudio called MediaBufferDecoder which needs to create reader objects in order to do its job, and therefore needs to have access to these headers. (Note that some of them are already exported.) I guess I can create the factory method that you're talking about, but that needs adjusting the LOCAL_INCLUDES in the Makefile in order to be able to access these headers. I'll do that if that's what you prefer. In that case, feel free to WONTFIX this bug. Thanks!

Comment on attachment 684759 [details] [diff] [review] Patch (v1) Review of attachment 684759 [details] [diff] [review]: ----------------------------------------------------------------- I guess we're damned either way. I'm concerned that when we add new backends we'll forget to add code to create readers to your Web Audio class. Let's run with this for now. I'd like to refactor and centralize our reader creation to prevent the above, but I think that's too bigger refactoring to ask you to do, so let's just do this in the meantime.

OK, thanks! https://hg.mozilla.org/integration/mozilla-inbound/rev/c35b660ef41b