[FFmpeg-devel] [RFC] OS support library

[Rich Felker]

On Wed, Dec 05, 2007 at 11:59:55PM +0100, Michael Niedermayer wrote:
> On Wed, Dec 05, 2007 at 05:24:30PM -0500, Rich Felker wrote:
> > On Wed, Dec 05, 2007 at 06:47:46PM +0100, Michael Niedermayer wrote:
> > > On Wed, Dec 05, 2007 at 04:47:44PM +0100, Luca Abeni wrote:
> > > > Hi V?ctor,
> > > > 
> > > > V?ctor Paesa wrote:
> > > > [...]
> > > > >> I moved lrintf() replacement to libossupport, and now I am trying to
> > > > >> test the resulting libs... Does anyone know a system that does not
> > > > >> provide lrintf()? I think cygwin can be used for such testing?
> > > > >> Does anyone know how to cross-compile from linux to cygwin (or a
> > > > >> system that does not provide lrintf())?
> > > > > 
> > > > > Cygwin has lrintf(), though it lacks llrintf().
> > > > 
> > > > Ah, ok. I must have confused lrintf() and llrintf() (BTW, how do we handle
> > > > the lack or llrintf()? If there is some workaround, can it be moved to
> > > > libossupport?)
> > > 
> > > (int64_t)rint(x) might be good enough (and is simple ...)
> > 
> > You probably meant (long long)rint((float)(x))
> 
> the function we use is llrint() not llrintf()
> about long long vs. int64_t yes long long is correcter though is there
> a system (we support) where int64_t != long long ?

Well llrint needs to return long long. If the caller wants int64_t
they can cast..

Rich