On Fri, 9 Feb 2001, Henrik Nordstrom wrote:
> Are we still using O_NDELAY?
>
> I have a memory of fixing this ages ago. O_NDELAY is not standardized,
> and behaves sligthly differently on different OS:es, where quite many
> return 0 in various non-EOF conditions..
>
> Ah.. SunOS & Solaris is excluded from the use of O_NONBLOCK. Do you
> remember why (this is inherited all the way from Harvest cached.. see
> comm.c revision 1.1, way before my time)
I certainly don't remember why.
There's two ways to fix this. 1) always use O_NONBLOCK from now on
and assume that any (Solaris) system where people install Squid works
the way O_NONBLOCK should work. So if someone tries to install newer Squid
on really old Solaris on even SunOS, it might not work.
2) make a special case for Solaris 8 and probably 7? I don't think
we can detect the solaris version from preprocessor defines, so it
probably requires a configure hack.
Received on Thu Feb 15 2001 - 09:42:49 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:31 MST