Building GHC in Windows (Was: Re: Failure in GHC compilation)

Esa Ilari Vuokko eivuokko at gmail.com
Tue Jun 5 07:17:13 EDT 2007


On 6/5/07, Simon Marlow <simonmarhaskell at gmail.com> wrote:
> Simon Peyton-Jones wrote:
> > I've extended the "what to look for if your build fails" section in
> >         http://hackage.haskell.org/trac/ghc/wiki/Building/Windows
> >
> > Please do keep adding to this page!  Every time you trip over something, think about whether your experience could be used to help someone else.
>
> [taken off-list]
>
> I'd say a bug report is a better place for this kind of thing.  They're just as
> searchable, and bugs can be closed more easily than editing a wiki page to
> remove out-of-date text.   The Windows wiki page is full of stuff that doesn't
> apply any more, and IMO its size is completely daunting to anyone who wants to
> build GHC on Windows.  Also the fact that there are two sets of instructions
> that differ in various details is confusing.

Whoa, there's lots of stuff there!  The page is unorganised, and I'd
guess nobody
wants to reorg it because nobody builds ghc in Windows more than one way, and
hence, nobody knows all the little tiny bits.

There's lots of choices: cygwin shell vs mingw shell; which gcc to use (cygwin,
mingw, from old ghc); which stage/install method used (stage 1/2 in-place, make
install, through bin-dist.) That gives easily dozen ways to build and use the
resulting ghc.

I've myself never (or for a long time) built ghc using cygwin, and I
never managed
to wrap bin-dist.  I have no idea why some flags are given, like
architecture-flag
(iirc it works just fine without it, in msys.)  Building docs required
cygwin for some
wierd reason I can't recall.

It's not clear to me if this needs to sorted out in Wiki/other docs or
if buildsystem needs
thorough revision in Windows.  Maybe it's all better in head (which I
haven't built for a
while now.)

Best Regards,
Esa



More information about the Cvs-ghc mailing list