[Pvfs2-developers] the epoll()-by-default experiment is a bust

[Pete Wyckoff]

robl at mcs.anl.gov wrote on Thu, 25 May 2006 14:28 -0500:
> On Thu, May 25, 2006 at 02:05:23PM -0500, Rob Ross wrote:
> > why don't we add a configure test that executes code to see if we get 
> > ENOSYS?
> 
> that might be a good-enough solution.  It leaves two issues:
> 
> - execution-based tests break cross compiling
> - Pete didn't like that pvfs2 libraries compiled on 2.6 systems were
>   not portable to 2.4 
> 
> For the former, we can just never use epoll if we are cross compiling
> 
> For the later, people in pw's situation can configure with
> --disable-epoll
> 
> Fair enough?

I generally don't like executing configure tests since I often don't
compile on the machine I will run on, and the cross-compilation
problem.  Most people probably do compile and run on the same
machine, though.  If you can make it so that:

    --disable-epoll disables it, no runtime test
    --enable-epoll enables it, again no test
    without one of these, configure tries to run a test code

So if I say to --enable-epoll, then configure doesn't force it back
off because it can't the test program.

		-- Pete