Hej Ben, Lars, och Kili,
First of all, sorry for not mentioning that FreeBSD has a base system
+ ports model. As Lars has pointed out, using bash on a FreeBSD
system (whether may I say on any other BSD system?) has a fixed cost
of installing it as a build-time dependency. Building GHC 6.10.4
worked fine without using bash, and I do not see any serious reasons
to start to use it. And as Kili mentioned, if we do nothing against
the GNUism creeping into the sources, it will result in a code that
hard to be ported to non-GNU platforms in the long run.
In FreeBSD, sh has a minimal but standard (POSIX.2) interface, and it
is not just a symlink to C shell (the default shell here). I have
just installed GHC build slaves on my FreeBSD system (for both x86 and
x86_64 architectures on 8.0-RELEASE, maybe you have seen my attempts)
to turn FreeBSD into a Tier 1 platform for Haskell. In my opinion,
that would help to spot potential GNUism emerging in the sources, and
keep GHC portable. This way I can validate the sources on a platform
exotic enough to keep the heads up here. Of course, I would be glad
to help in locating and fixing the portability problems as it is
already a requirement for a Tier 1 platform. BSD systems have taught
me many about portability over the years, I think everybody wins here.
Ironically, I needed to hack my systems under the build slaves towards
the GNU direction to see other interesting things like built snapshots
:) I installed bash, and symlinked to /usr/bin/bash; I installed
libiconv, and symlinked headers and libraries to /usr/{include,lib};
finally I renamed make(1) to bmake (as it is a BSD make), installed
GNU make as gmake, and symlinked make to gmake (ouch!).
I am working on the elimination of these issues one by one, first of
them was the patch I sent here. All the other things must be
addressed is inside the build server itself, but I did not have time
to look at its sources and build a patch (or a configuration file?)
for that. Hopefully I can do it soon.
Sorry for not writing a ticket and doing a proper validation, but Kili
was right, I do not have access to all the other systems but Windows.
Regarding validation, I am expecting that others will help me out in
reviewing and verifying the patches on their favorite platform(s). My
supposed contribution here would be to test GHC for building and
validation on all combination of supported FreeBSD versions and
platforms (e.g. 6.x, 7.x, 8.x, 9.x, on both i386 and amd64 -- which
totals up in 7 different builds at the moment). For BSD developers it
is a very common habit to do so, I hope this is an accepted approach
here :)
Cheers,
:g
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
