possible mmap() problems

[Joe Buehler]

Christopher Faylor wrote:

> Just commenting out the if in fhandler_pipe::ready_for_read which
> controls whether "fhandler_base::ready_for_read" gets called or not,
> making fhandler_base::ready_for_read always called unconditionally,
> should get you pretty close to the way things used to be.

This does indeed seem to work -- I get a lot fewer thread terminations now.
There is no noticeable impact to performance of my applications, so I will
continue using this for the moment.  I still have reliability problems,
however.  It can take a couple hours of wall clock time before anything
fails, so it will take me some time to track down the problem.
-- 
Joe Buehler