Socket timeout bug (?).

[Sergei Organov]

This time it seems to be correct.

BR,
Sergei.

Joel Sherrill <joel.sherrill at OARcorp.com> writes:
> Sergei Organov wrote:
> > 
> > "Joel Sherrill <joel at OARcorp.com>" <joel.sherrill at OARcorp.com> writes:
> > [...]
> > > I did a search for similar code in the stack and only came up
> > > with rtems/rtems_select.c (select()) which already had a similar
> > > piece of code.  It could be doing the same thing a different way
> > > though.  Anyway, I am committing this which I believe fixes this
> > > one:
> > >
> > > Index: kern/uipc_socket.c
> > > ===================================================================
> > > RCS file: /usr1/CVS/rtems/cpukit/libnetworking/kern/uipc_socket.c,v
> > > retrieving revision 1.4
> > > diff -u -r1.4 uipc_socket.c
> > > --- kern/uipc_socket.c        11 Jun 1999 14:11:41 -0000      1.4
> > > +++ kern/uipc_socket.c        14 Sep 2002 18:27:39 -0000
> > > @@ -961,6 +961,8 @@
> > >                               goto bad;
> > >                       }
> > >                       val = tv->tv_sec * hz + tv->tv_usec / tick;
> > > +                     if (val == 0)
> > > +                             val = 1;
> > 
> > I worry if the case when timeout is set to exact 0 goes through this path? If
> > so, the patch wold break "wait forever" behavior of timeout 0.
> 
> How about this version then?  It is pretty explicit about asking for
> timeout 0. 
> 
> @@ -960,7 +960,13 @@
>                                 error = EDOM;
>                                 goto bad;
>                         }
> -                       val = tv->tv_sec * hz + tv->tv_usec / tick;
> +
> +                       if ( tv->tv_sec || tv->tv_usec ) {
> +                               val = tv->tv_sec * hz + tv->tv_usec /
> tick;
> +                               if (val == 0)
> +                                       val = 1;
> +                       } else
> +                               val = 0;
> 
> > BR,
> > Sergei.
> 
> -- 
> Joel Sherrill, Ph.D.             Director of Research & Development
> joel at OARcorp.com                 On-Line Applications Research
> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
> Support Available                (256) 722-9985