ghc 6.2 build problem (MinGW/MSYS)

[Simon Peyton-Jones]

| > Is there any reason why cygwin is preferred environment in
comparions with
| > MSYS?
| 
| I'm not sure, but it could simply be for historical reasons.
| 
| > imho, MSYS provides much simpler & light install than cygwin,
creates win32
| > native binaries & mimic Linux configure ; make ; make install
mechanism.

I think it's historical.  I don't think MSYS was available when we began
(and I for one am still ignorant of it), whereas Cygwin was.  Now there
is also Microsoft's (free) Services For Unix (SFU) which would be
another possible choice.

We don't plan to actively support more than one build route on Windows.
Keeping GHC buildable on many systems is already a lot of work, and
given that we already have a working solution it's hard to motivate a
change.  In what way would Cygwin be better than MSYS?

Of course, you are welcome to do the port to MSYS.  Having done that
you'd be in a good position to persuade us about why it was easier.
But I'm pretty cautious about abandoning a known solution that is not
giving us problems in favour of a new one.

As Sven says, another possibility is to maintain both Cygwin and MSYS
routes.  Whether we'd want to incorporate the MSYS modifications would
depend on how invasive they were. The build system is already hard
enough to understand, and a maze of ifdefs only make it worse.  But if
it isn't too invasive, and gives benefits for at least some people, then
it's definitely possible.

But remember that we know nothing about MSYS so while we can help you
understand the build system (please read all the build-system
documentation first) you'll be pretty much on your own with
build-system/MSYS interactions.

Keep us posted!

Simon