cthelper.exe process creates 3.5 KB/sec of 'other' io even when the window is idle about puttycyg

[GoogleCodeExporter]

Since I happened to be in the neighborhood (see also issue 39) I put a loop 
around the call to select() in cthelper.c, and added a timeout... A timeout of 
a second or more does nothing interesting, but I/O seems to increase in inverse 
proportion to timeouts under a second, e.g. a timeout of .5 seconds causes the 
I/O to double, even though .5 seconds is an eternity for a modern PC. :) I 
tried to do a minimally-complete test case outside of puttycyg but had a tough 
time... select() on e.g. stdin/stdout didn't seem to have a problem, but I 
couldn't easily replicate the pipe created by setup_child/pty_fork/etc. Wonder 
if that's relevant...

Original comment by [email protected] on 8 Jul 2010 at 12:14

[GoogleCodeExporter]

The I/O is all being done inside the call to select().  Looking at 
winsup/cygwin/select.cc, it seems likely that we are looping in 
select_stuff::wait() -- not sure why.  We should have a pipe, a tty (master) 
and a socket; maybe one of these is difficult to wait on, or maybe the  
combination is difficult to multiplex?

I'll try and build a debugging version of cygwin1 and see if I can pin it down. 
 No promise on a timeline. :)

Original comment by medgar123 on 29 Oct 2010 at 12:02