Re: test HTTP stream server - no "disconnect" signal from SoupSocket
- From: Krzysztof Konopko <krzysztof konopko youview com>
- To: Dan Winship <danw gnome org>
- Cc: libsoup-list gnome org
- Subject: Re: test HTTP stream server - no "disconnect" signal from SoupSocket
- Date: Mon, 26 Nov 2012 22:56:02 +0000
On 26/11/12 13:39, Dan Winship wrote:
> On 11/23/2012 08:46 AM, Krzysztof Konopko wrote:
>> I have some logic that I have to comply with before I can actually play
>> the pipeline and some requirements like the stream should be served
>> continuously with a *single* GET request (after some initial
>> "negotiation and probing"). So no chunked delivery is acceptable at the
>> HTTP level (unless the client explicitly makes separate Ranges requests).
>>
>> In order to acheive this I had to reach out for a pretty low level
>> SoupSocket API where I can write directly into the socket as Gst buffers
>> arrive. None of the higher level libsoup methods suited me as they tend
>> to break some of my requirements underneath, e. g. they break the stream
>> down into HTTP chunks if I use
>> soup_server_pause_message()/soup_server_unpause_message() no matter what
>> message encoding I choose.
>
> No, that's not true; tests/streaming-test.c in the source explicitly
> checks that SOUP_ENCODING_CHUNKED, SOUP_ENCODING_CONTENT_LENGTH, and
> SOUP_ENCODING_EOF all work for streaming a response from the server.
>
> Maybe you're just getting confused by the API, which still uses the
> "wrote-chunk" signal even if you're not using chunked encoding. (It
> essentially just means "finished writing the last block of data", not
> literally "wrote an HTTP/1.1 chunk".
>
> -- Dan
>
>
Thanks for replying. I'll revisit my approach as in the mean time I
fixed some problems in GStreamer elements and upgraded libsoup and the
whole solution became stable.
Cheers,
Kris
[
Date Prev][Date Next] [
Thread Prev][Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]