[MLton] cygwoes

[Matthew Fluet]

>> Unfortunately, all this suggests that there is a lot more broken
>> under the current version of Cygwin than just the virtual memory
>> management.
>
> I'm not so sure.  Keep in mind that all the regressions pass when run
> with use-mmap.  With your fix of gc.c line 1319, do all the
> regressions pass even without use-mmap?

I can't even get a working compiler on Cygwin, even with the change to 
line 1319 of gc.c.

If I set UseMmap to FALSE, then I don't get the VirtualFree error 
(obviously), I _do_ get the lowercase error, and I don't get the 
gcc/signal error or the gcc/"No child process" error.