(it's tuesday 19th here, that html-message took some time..)
----- Original Message -----
From: Conal Elliott
To: 'Simon Peyton-Jones'
Cc: [EMAIL PROTECTED]
Sent: Friday, April 15, 2005 7:09 AM
Subject: RE: Win32 build of ghc from cvs failed; also 6.4 source distribution
Here is another attempt, from scratch, to get and build GHC on Windows XP.
I've omitted a lot of output ("[.]") but
included some warning messages. Do you spot any goofs in my steps? Do these
messages indicate problems? BTW, I
updated cygwin this week and got a fresh mingw. - Conal
------------------------------------
just so that I don't always write about builds that fail:-)
I successfully built ghc head on win xp, with Monday's cvs,
and since my machine was rebuilt, with last week's cygwin
(out of curiosity, I did not exclude hslibs). Btw, I hope I'm
not spilling any secrets when I say that Sigbjorn has a nice
script for making movable tar-balls for windows without msi,
right there in cvs. Nightly windows builds and snapshots,
anyone?
I noticed a few oddities:
- cygwin doesn't give you a selection of "profiles", so when
I reinstall cygwin, I always have to modify the default
selection of packages for my purposes, mostly making sure
that all the tools I need will be included. I do not usually
have to do any funny stuff like replacing sh with bash!
- the online building guide seems not to have been updated
with the cvs changes? I once again forgot to add GLUT
to MSYS the first time round..
- at first, configure got hung up at various points (just
stalled, not returning or doing anything). since one of the
points was between x11 and hgl, I uninstalled all x11-related
stuff from cygwin before restarting the whole process. I
also uninstalled any mingw-related things to avoid any
accidental conflicts between cygwin's mingw and msys
mingw. after that, configure got through as expected.
- apparently, I installed only part of the doc-processing tools
in cygwin, so configure gave me a message that some
conversions won't be possible. as I only wanted html,
that looked ok to me, but when doing the distribution
stuff, hslibs/docs suddenly tried to convert docs without
having the tools available - I just commented out that part.
- it would be helful if the mk/build.mk.sample corresponded
to the default settings (with useful options in comments).
when having to add BIN_DIST=1, I naively copied and
modified that file, and was surprised by the amount of
work that was done after.
- there was an absolutely amazing amount of warnings
generated in the process.. how do you ever find the
important ones in there? I'm often not good at going
for warning-free code myself, but I do recall that as
a pragmatically important invariant:-)
As always, when it works, your build process is amazing
(have you recently counted the files in even the partial
fptools tree needed for ghc?-).
cheers,
claus
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc
