[ Moving to [EMAIL PROTECTED] since this is now an APR issue ] On Fri, Jul 20, 2001 at 10:58:49AM -0400, Jeff Trawick wrote: > 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().
Well, first we should already be using O_NONBLOCK with pipes (I checked with gcc -E) - APR is using flag 0x80 which is Solaris's definiton for O_NONBLOCK. According to the man page, we're seeing what would happen if O_NDELAY is set - not what would happen if we had O_NONBLOCK set. Odd. Something funny is going on here or I'm incredibly stupid. I'll try and recreate this again. -- justin
