Hi,

On Ubuntu 8.10 with binutils trunk snapshot from 20080707, Darcs fails
to build due to errors with the flags passed to gcc for compilation and
linking, specifically that -g is passed at both phases. Upstream bug at
[0].  I'm not sure whether this is a problem with darcs's build system
or with GHC, however.

A buildlog for darcs-2.0.0 (same error occurs with .2) can be found at [1].

I'd be grateful if anyone could point me to the right place to make this
change, or let me know if it is indeed a GHC problem and out of darcs's
control - this stuff is a bit foreign to me.

Thanks all,
Iain

[0] http://sourceware.org/bugzilla/show_bug.cgi?id=6656
[1]
http://launchpadlibrarian.net/15404312/buildlog_ubuntu-intrepid-i386.darcs_2.0.0-5~ppa0_FAILEDTOBUILD.txt.gz

I'm afraid I do not know what to suggest, or who to send you to.

I do not see a -g in my autoconf.mk (Ubuntu 8.04), do you?  I suppose the right
place to make this sort of change would be in configure.ac (or maybe in
autoconf.mk.in?)

Ian (Lynagh): is this something I could ask you to comment on?

Iain (Lane): if it helps, there is also the GHC trac on
http://hackage.haskell.org/trac/ghc

That build log hides the GHC commandline, so I've got no idea what's going on.


Thanks
Ian

Iain: could you try
  make VERBOSE=1

so that Ian could have an idea what is going on?

Eric Kow wrote:
> Eric Kow <eric.kow@gmail.com> added the comment:
> 
> Iain: could you try
>   make VERBOSE=1
> 
> so that Ian could have an idea what is going on?
> 

Certainly. I've uploaded a new buildlog to [0].

HTH, Iain

[0] http://orangesquash.org.uk/~laney/ftbfs_darcs

OK, so darcs is passing both -optc-g and -opta-g (twice each, in fact) to GHC.
So it looks like the problem's coming from darcs' build system.

Thanks
Ian

Ian Lynagh wrote:
> Ian Lynagh <igloo@earth.li> added the comment:
> 
> OK, so darcs is passing both -optc-g and -opta-g (twice each, in fact) to GHC.
> So it looks like the problem's coming from darcs' build system.
> 
> Thanks
> Ian
> 
> ----------
> nosy:  -iain
> 

Regardless, this problem has now been fixed by binutils upstream[0]. It
might still be a good idea for darcs to play nice too, though.

Thanks,
Iain

[0] http://sourceware.org/ml/binutils-cvs/2008-08/msg00027.html

I assume this is resolved by our switching to Cabal to build everything.  Please
let me know if this is not the case