[MPlayer-dev-eng] [PATCH] http streaming issues

[Ivan Kalvachev]

On 6/25/08, Alexander Kanavin <alexander.kanavin at nokia.com> wrote:
> ext Dominik 'Rathann' Mierzejewski wrote:
>>> 1) Some web servers don't provide a way to seek within a file (example:
>>> Nokia N-series phones),
>>
>> Nokia phone is a web server?
>
> Yes - if you instruct the phone to share music, images and videos that it
> has over WLAN, they become available via http. It has a simple web server
> built in.
>
>>> 2) ASF streaming code relies on having first few bytes of the stream body
>>> available when it parses the http headers, however it only ensures that
>>> the
>>> entire http header is available when reading from a socket, and doesn't
>>> ensure that the beginning of the body is available as well. Some web
>>> servers, such as Windows Media Player in Windows Vista,
>>
>> Windows Media _Player_ is a web server??
>
> Yes :) If you allow collection sharing through Library->Media Sharing, it
> becomes available over http as well.
>
> The full story is that the actual URLs to play from are available over a
> separate protocol called UPnP AV Media Server in both cases, but mplayer
> isn't involved here: the URLs are retrieved by an external helper
> application, or just supplied on the command line by hand.
>
>>> send the headers
>>> and the body in separate TCP packets, so the body isn't read at all and
>>> this breaks ASF playback.
>>
>> Breaks how? Could you provide MPlayer output before and after your patch?
>> I've rarely had problems with playing ASF streams.
>
> Sure. Before the patch, mplayer does not have the stream body to inspect,
> mistakenly thinks it's a playlist, tries to parse it and exits with a
> failure. After the patch, the file is identified and played correctly. I've
> attached both outputs, have a look.
>
> Also note that this problem didn't exist before this patch:
> http://svn.mplayerhq.hu/mplayer/trunk/stream/asf_streaming.c?r1=25239&r2=25415
> because before that all URLs were treated as playable streams, even if the
> parser (mistakenly or not) decided it was a playlist. So fixing one bug
> uncovered another :)


Reimar, would you review these patches.
They look OK to me, but this is not my area of expertise.