[Bf-committers] configure.in et. al. missing from CVS
Mon, 23 Feb 2004 16:52:10 +0100
On Monday 23 February 2004 16:17, Michel Selten wrote:
> On Mon, 2004-02-23 at 16:02, Gregor Mück wrote:
> > Great. What an excuse.
> > Using the new system to cross-compile will probably lead to wierd and
> > complex hacks around scons functionality (at least you need to determine
> > and use two compilers - one for the host platform and one for the target
> > platform). I am not able to do that.
> > And I am not willing to invest time into a poorly designed build system
> > (scons is one in its current form) when there is an alternative availab=
> > that does not require this amount of effort. I have other work to do as
> > well.
> I'm sorry to hear your complaints after autoconf was removed from cvs.
> In my opinion, you should've told your issues when the discussion
> started, not after the actual work has been done. I'm open to all
> opinions, and I want to spend time on investigating a possible solution.
> You can read that in the mails I've sent to the list, but if I'm unaware
> of an issue, I can't investigate it.
> With regards,
It's not too late to go back :-).
The last issue that remained during my attempts was the final linking of the
binary. The rest has been fairly easy to do. At least the makesdna program
must always be compiled with the host as the target system, as it will be
executed on the host lateron. This was the only other big problem I had to
The rest of the effort is to give scons scripts the magic to find all the
components of the cross-compilation environment and find ways to call the
host compiler although the normal gcc or cc is the crosscompiler. These are
problems that were formerly implicitely solved by using configure. By
starting to work on the cross-compilation environment I have here right now I
realized where all the complexity of autoconf/automake actually comes from.
This suite attempts to solve many problems simultaneously and is actually
fairly successful at that, too.