Re: GNUstep - check for reuse address

[Svante Signell]

On Fri, 2016-01-08 at 21:59 +0100, Svante Signell wrote:
> On Fri, 2016-01-08 at 16:43 +0100, Samuel Thibault wrote:
> > Pino Toscano, on Fri 08 Jan 2016 16:40:08 +0100, wrote:
> > > In data venerdì 8 gennaio 2016 13:34:46, Samuel Thibault ha
> > > scritto:
> > > > Svante Signell, on Fri 08 Jan 2016 13:27:49 +0100, wrote:
> > > > > > Depends on how the test is made. SO_REUSEADDR is defined in
> > > > > > bits/socket.h but is not functional (yet).
> > > > > 
> > > > > To clarify; For pflocal. For pfinet it should work.
> > > > 
> > > > Ok.  It seems to be accepted for local sockets on Linux, but it
> > > > doesn't do anything.  I don't see what it would be supposed to mean
> > > > anyway, so gnustep should really not be using that for local sockets.
> > > 
> > > IIRC it should unlink the existing socket path before trying to
> > > bind the unix socket to the specified path -- otherwise you'd get
> > > EADDRINUSE.
> > 
> > Yes. And SO_REUSEADDR won't help there :)
> 
> Samuel, this is exactly what the SO_REUSEADDR in pflocal should do:
> Unlink the old socket and create a new one with the same name. (I
> wonder how GNU/Linux is implementing this?)

This problem is coming up again in the gccgo tests, making some of them fail.
What's wrong with implementing it also for pflocal?