Icecast dev - Aug 2004 - PATCH: increase network congestion resilience (SOLVED!)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Sorry. The patch is hopely here.

On Saturday 18 January 2003 03:38, Michael Smith shaped the electrons to
shout:
> > Hi,
> > 	find a patch which is an update to a patch sent months ago. Before
> > it was in net/sock.c, now I moved it to format.c, so net CVS module
> > is not affected. It polls the socket before trying to send() any byte
> > to check if the TCP buffers are full due to network congestion.
>
> Why? This should be no different from doing the send() (and handling it
Goddam!!!!

There were bugs in icecast2 while checking for recoverable error (in
several places indeed).

The program was checking for sock_recoverable() with the return value of
send(), which is a mistake and confused me because I assumed it was
right. It _must_ be done against with sock_error() instead.

Find the patch that solves the bugs.

- --
  ricardo galli       GPG id C8114D34
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+KUOLXZ2eFsgRTTQRAvtQAKDWnOoK5hNv1ilzuxVJJStE5KJgsQCfRFHC
cRtvQvDZElEoGSXzdp2J/d8=bQZY
-----END PGP SIGNATURE-----

-------------- next part --------------
A non-text attachment was scrubbed...
Name: update1.diff
Type: text/x-diff
Size: 2861 bytes
Desc: update1.diff
Url :
http://lists.xiph.org/pipermail/icecast-dev/attachments/20030118/83c6108a/update1.bin

> Goddam!!!!
>
> There were bugs in icecast2 while checking for recoverable error (in
> several places indeed).
>
> The program was checking for sock_recoverable() with the return value of
> send(), which is a mistake and confused me because I assumed it was
> right. It _must_ be done against with sock_error() instead.
>
> Find the patch that solves the bugs.
Ah! Good catch, thanks for these. That was a pretty nasty little error...
The first two fixes in this patch are correct, and have been applied.
The second two are unneccesary, since
source->format->write_buf_to_client() is
required to return less than 0 _only_ if a fatal error occurs (which with the 
first two fixes is now done correctly).

Thanks,

Mike

--- >8 ----
List archives:  http://www.xiph.org/archives/
icecast project homepage: http://www.icecast.org/
To unsubscribe from this list, send a message to
'icecast-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Saturday 18 January 2003 13:29, Michael Smith shaped the electrons to 
shout:
> applied. The second two are unneccesary, since
> source->format->write_buf_to_client() is required to return less than
0
> _only_ if a fatal error occurs (which with the first two fixes is now
> done correctly).
Damm, yes, double check, it seems I was on drugs (in fact I was taking 
care of my daughter while I was typing, which isn't healthier), thanks. 

What is your opinion of a "faststart" implementation? It's almost
ready,
just testing it extensively. It's configurable:

       <limits>
                ...
                <faststart>10</faststart>
        </limits>
(if the value is 0, faststart is disabled)

Check it out at http://mcrg.uib.es:8000/live.ogg or 
http://mcrg.uib.es:8000/high.ogg

Regards,

- -- 
  ricardo galli       GPG id C8114D34
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+KVO8XZ2eFsgRTTQRAl/9AJ0aDAf33xUpJSRXVqA0tDg7FajFeQCcCLtP
jMpoRCXEwRhLpIKxOmIY8fg=6H5c
-----END PGP SIGNATURE-----

--- >8 ----
List archives:  http://www.xiph.org/archives/
icecast project homepage: http://www.icecast.org/
To unsubscribe from this list, send a message to
'icecast-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.