README.Cygwin

README.cygwin (as of 3/21/2009):

    We have all of ROSE compiling with Cygwin now
(checked into to both internal and external SVN repositories).
I am working on getting it to pass all the tests, this is going
fairly well. 

Computer requirements:
   Previous work to compile ROSE under Cygwin didn't
work well, I expect that it is because ROSE was not 
as far along in development and Cygwin may have improved
over the years.  Tests in summer 2006 got initial Cygwin 
support in place and added libtool support to ROSE to help
make that happen (initial OSX support was added as well).
Tests in summer 2008 revisited Cygwin support and got more
working, but also had problems running out of memory, perhaps
it was the low-end machine that we used. Most of ROSE has
made to work, except Compass and some other pieces. 
Spring 2009 revisited the Cygwin support with a larger machine
(4 Gig 2 socket dual-core machine).  

Performance under Cygwin:
   Cygwin appears to be SLOW, at least compared to Linux.
In general everything using Cygwin is about 3-4 times slower
than using Linux (maybe more). That might be partly due to the 
difference machines we use for Cygwin support and Linux development.
We will reevaluate in the future.  I can't imagine doing serious
development on a Cygwin machine.

Status of compilation and testing:
   Support was added to ROSE to allow platform dependent 
compilation of the now very few parts of ROSE that remain a 
problem to Cygwin and so now the compilation of ROSE (well 
over 99% of ROSE now compiles, link, runs, and passes our 
tests).  The few parts that done (e.g. complex arithmetic 
support) are well documented as being conditionally compiled 
and/or conditionally tested using macro (__CYGWIN__) and 
Automake conditional (ROSE_BUILD_OS_IS_CYGWIN).

  In general, there are so few that I am mostly eliminating 
them from being compiled or run under Cygwin (then we can
see where we stand).  Clearly over 99% of ROSE
will compile, link and run and pass over 99% of the tests.

It all works so well that it is a mystery why anything
fails to compile, link and pass the tests.  In one case
it is because Cygwin does not appear to support
(or have) and complex.h header file, so all those
tests were eliminated from being run.


Notes:
  1) Makefile dependences cause many source files to be recompiled
     where under Linux they would just be relinked.  Not clear
     why this is.