[libgo] Account for 32-bit fds_bits on Solaris 2

[Ian Lance Taylor]

Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> writes:

> While debugging why several libgo tests on Solaris 2/SPARC were hanging
> in select (cf. PR go/48242, go/48243), I found that fd_set is
>
> typedef struct fd_set {
>         long    fds_bits[__howmany(FD_SETSIZE, FD_NFDBITS)];
> } fd_set;
>
> The current implementation of the FD* funcs in sysfile_posix.go assumes
> 64-bit fds_bits.  While this doesn't make a difference on little-endian
> hosts, on big-endian SPARC the wrong bits are set, leading to the
> observed hang since there is no activity on those random fds.
>
> The following patch fixes this by moving the FD* implementation to two
> new sysfile_fdset{32, 64}.go files to fix this.  There's almost
> certainly a cleaner/shorter implemenation, but this worked for me.
> While the affected tests don't hang anymore now, they still don't finish
> successfully, among others due to PR go/48222.

What an annoying problem.  Sorry about that.  But why don't we just
change to byte?

Ian