Crash with new gc on Cygwin

[Stephen J. Turnbull]

Marcus Crestani writes:

 > SJT> Are these due to XEmacs changes, or due to Cygwin changes?  If
 > SJT> the latter, we really ought to try to detect them in
 > SJT> configure.
 > 
 > These are due to Cygwin changes.  There is not much to detect,
 > though:

Of course there is.  Not everybody will be willing to upgrade their
Cygwin installation just because XEmacs decides to use a technology
that doesn't work properly on older installations to conform to newer
Cygwin versions.  If we can detect that via something as simple as a
Cygwin version check, which I bet will work in 99% of current user
installations, we should do that.

 > The current default for Cygwin in configure is to use the
 > Windows-style barrier, if I get a positive report from Mats I'll
 > change that default to the POSIX barrier.

We should default to the right barrier for the user's system, perhaps
with a warning.