Hey,<div><br></div><div>Please review the attached patch.</div><div><br></div><div>I consider it important because it properly fix libvlc_media_get_es(). We are currently advising our poor user to create a new media_player with --sout=#description which makes us look overcomplicated. Mostly because you still have to attach an event and such to know when we are done, and also because a full playback is launched, not saving CPU and memory.</div>
<div><br></div><div>Our alternative to this patch is to read the input_item's info dictionary. And parse the string in it. Hence the modification stay in libvlc umbrella and not in the core.</div><div><br></div><div>However I consider this patch as low risk:</div>
<div>- It is following how we fill the input_item's info which is an already existing string dictionary.</div><div>- Its impact in term of performance is low. We are not using a full es_format_t, but just a stripped down version of fixed size, which is what we expose in libvlc_media.</div>
<div>- It's chirugical and circumvented.</div><div>- It's fixing a real issue in libvlc.</div><div><br></div><div>If it's not accepted I'll fix libvlc_media_get_es() in 1.1 by parsing input_item's info field, which are correctly filled after preparsing for most demuxer. I am ok with this, but wanted to get the patch archived on the ML at least.</div>
<div><br></div><div>Thanks,</div><div><br></div><div>Pierre.<br><div><br></div><div><br></div></div>