At Wed, 26 Aug 2009 16:46:39 -0400,
John Chludzinski wrote:
> 
> > Yoshiki Ohshima yoshiki at vpri.org 
> > Wed Aug 26 20:32:00 UTC 2009 
> >
> > At Wed, 26 Aug 2009 16:28:46 -0400,
> > John Chludzinski wrote:
> > 
> >> Wouldn't Cygwin allow for greater commonality (than MinGW) within the code base? ---John
> >
> > In the current code base, do you have any particular pieces in your
> > mind where switching to Cygwin would really help to get greater
> > commonality?
> >
> > -- Yoshiki
> 
> No I don't. I've never looked thru the VM source for Win32 calls that could be POSIX calls. I just ported a POSIX app
> (OpenSolaris) to Windows (using Cygwin 1.7 beta) and was impressed with how remarkably painless that was. And being a
> Smalltalker/Squeaker, when possible, that question came to mind. ---John

  - Graphics.  You don't expect the user to run Squeak in Cygwin X.
    So, you do need a different code for it.

  - File names should be "sane" to Window users.  No /cygdrive/c/.

  - Browser plugin needs very different mechanism.

  - Memory allocation, if we try to use mmap() in cygwin to emulate
    the desirable behavior, we probably end up with fighting the
    incomplete emulation or incompatiblity more than just writing
    well-defined primitives in Win32 calls.

  - So is DLL loading.

  - Network support may or may not be easier, but probably with the
    same problem.

  - FFI, ... etc., etc.

We don't hack the VM for platform specific reasons that often, and
when we do, it is usually about writing well-defined, isolated
functions.  And having the common code means to test a change for a
platform is harder.

The bottom line:
  Don't worry about it ^^;

-- Yoshiki