[Ffmpeg-devel] Network patch for MingW users

[François Revol]

> > -#if defined(__BEOS__) || defined(__INNOTEK_LIBC__)
> > +#if defined(__BEOS__) || defined(__MINGW32__) || 
> > defined(__INNOTEK_LIBC__)
> >  typedef int socklen_t;
> >  #endif
> 
> Arrgh, more OS dependencies.  This too should be tested in configure.

That one I'll need, found out Haiku which defined __BEOS__ actually 
typedefs socklen_t :-(

> 
> >  #ifdef __MINGW32__
> >  __declspec(dllimport) void __stdcall Sleep(unsigned long 
> > dwMilliseconds);
> >  // #  include <windows.h>
> > +#define GUID microsoft_issue_GUID
> > +#include <winsock.h>
> > +#undef GUID
> >  #  define usleep(t)    Sleep((t) / 1000)
> > -#  include <fcntl.h>
> > -#  define lseek(f,p,w) _lseeki64((f), (p), (w))
> > +#  define sleep(t)     Sleep((t) * 1000)
> > +#define O_NONBLOCK FIONBIO
> > +#define fcntl(fd,b,c) { u_long arg=1L; \
> > +                        ioctlsocket(fd, c, &arg); }
> 
> I'm not convinced this is a good idea.  Someone might use fcntl() on 
> a
> real file.

Why not just define a fcntlsocket() that other OSes would just map to 
fcntl like is done for close already ? Much cleaner.

> >      if (port <= 0 || port >= 65536)
> >          goto fail;
> >  
> > +#ifdef __MINGW32__
> > +    init_winsock();
> > +#endif
> 
> Aaaaiiiieeee!  NO MORE OF THIS. PLEASE.

Also that's totally definitely non threadsafe.

Fran?ois.