Re: Ncat regression testing on Windows - results

[Jacek Wielemborek <wielemborekj1 () gmail com>]

2013/8/9 David Fifield <david () bamsoftware com>:
> On Fri, Aug 09, 2013 at 02:48:12PM +0200, Jacek Wielemborek wrote:
> > I kind of accidentally fixed the "--send-only client" bug. The
> > solution's in r31702.
>
> In what branch, where? Can you say more about the fix?
>
> David Fifield

Okay, sorry...

The bug was related to fselect's behavior when it just listens on
stdin. The select() isn't called then, as it's filtered out from
fd_set and we get into a loop of just calling the PeekNamedPipe. That
eats up a lot of resources until we gets some input, so calling a
usleep instead of select() makes it less CPU intensive.

I suppose it made the Cygwin tests fail because of the way the process
priorities were arranged. You pretty much can't even hit ctr+c over
SSH when it's in this loop.

For the organization stuff, I found a solution while hacking up
fselect() to work with named pipes in ncat-lua-select, then merged it
back to the trunk in r31710.
_______________________________________________
Sent through the dev mailing list
http://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/