[FFmpeg-devel] [PATCH] adding _GNU_SOURCE for udp.c compilation

[Rich Felker]

On Mon, Feb 11, 2008 at 03:30:55AM +0100, Michael Niedermayer wrote:
> On Sun, Feb 10, 2008 at 09:17:52PM -0500, Rich Felker wrote:
> > On Sun, Feb 10, 2008 at 09:14:42PM -0500, Rich Felker wrote:
> > > On Sun, Feb 10, 2008 at 07:20:31PM +0100, Nicolas George wrote:
> > > > Hi.
> > > > 
> > > > Le duodi 22 pluvi?se, an CCXVI, Reimar D?ffinger a ?crit?:
> > > > > Well in my header file (gentoo and debian) they are claimed to be part of
> > > > > POSIX1.g, so they are not under a GNU ifdef.
> > > > > Though I admit I could not find it in the POSIX specs I had lying
> > > > > around.
> > > > 
> > > > Neither NI_MAXSERV nor NI_MAXHOST appear in the Single Unix Specification
> > > > version 3, which is a superset of POSIX.
> > > > 
> > > > http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/resolv/netdb.h?annotate=1.47.2.1&cvsroot=glibc
> > > > 
> > > > shows that the check for _GNU_SOURCE has been added three weeks ago, with
> > > > the log message "Cleanup namespace.".
> > > > 
> > > > getnameinfo seems to return EAI_OVERFLOW if the buffers are not big enough.
> > > > EAI_OVERFLOW is standard, and described in the SUSv3 about netdb.h, but not
> > > > referenced in the page about getnameinfo.
> > > > 
> > > > I fear that the only absolutely reliable way of doing this is to increase
> > > > dynamically the buffer each time EAI_OVERFLOW is returned. Yet, using
> > > > NI_MAX* when they are defined as the initial size for the buffer should be
> > > > helpful nonetheless.
> > > 
> > > For the service number (port), sizeof(int)*3 is a perfectly acceptable
> > > bound. This function does not even make use of the hbuf result so it
> > > should be passing a null pointer for that rather than wasting
> > > complexity and stack space.
> > 
> > And here is the patch.
> 
> ok

committed

> and dont forget to add yourself to maintainers for udp :)

fat chance

rich =)