I'd first look into what happens if we use O_NONBLOCK instead of
O_NDELAY on Solaris. That is the simplest fix, and it avoids the need
to add extra pathlength to apr_read().
See 1.3's ap_bnonblock(). It prefers O_NONBLOCK over O_NDELAY.
I noticed that 1.3 has symbol NDELAY_PIPE_RETURNS_ZERO for some funky
platform but not for Solaris. I guess it wasn't necessary to set for
Solaris since we used O_NONBLOCK on Solaris.
Justin Erenkrantz <[EMAIL PROTECTED]> writes:
> On Fri, Jul 20, 2001 at 10:34:13AM -0400, Jeff Trawick wrote:
> > > How do we test for this in autoconf (unless we specify something in
> > > hints)?
> >
> > pipe()
> > set read handle non-blocking
> > call read() on read handle of pipe
> >
> > if rc=0, then this is Solaris^H^H^H^H^H^H^Ha platform which acts like
> > Solaris in this respect
>
> Okay, I can try and add something like this.
--
Jeff Trawick | [EMAIL PROTECTED] | PGP public key at web site:
http://www.geocities.com/SiliconValley/Park/9289/
Born in Roswell... married an alien...
