[Axiom-developer] RE: Google Code repository

[Page, Bill]

Ralf, 

On Wednesday, September 20, 2006 4:53 AM you wrote:
> ...
> Bill Page wrote: 
> > The path that I had in mind to do this is starting with the
> > build-improvements branch on SVN since the new build environment
> > is based on tecnhiques that should be portable to the Windows
> > MSYS/MinGW environment and also can start with a previously
> > installed version of GCL which simplifies things significantly
> > on Windows.
> 
> The last edit on axiom--windows--1 was 2005-03-17. Wouldn't it
> make sense to merge it first with Gold (tla), then test it and
> only afterwards merge with the build-improvements? I guess that
> Gold is closer than build-improvements.

I don't think the additional differences in build-improvements
would make the merging more or less difficult. I would like to
take advantage of the new configure scripts to help detect
the differences between Linux and Windows.

> 
> I've just tried
> 
> tla get address@hidden/axiom--windows--1 
> axiom--windows--1
> 
> (one line)
> 
> cd axiom--windows--1
> tla star-merge address@hidden/axiom--main--1
> 
> Runs nicely, but gives a number of conflicts in the Makefiles
> which I feel unable to resolve. :-(
> 

I seriously doubt that star-merge would be of much help in
merging these two sources because they work on two different
systems. The changes on the Linux side might be incompativle
with Windows and vice versa even though there is no conflict
to resolve. So even if star-merge succeeded, it is likely to
cause breakages in both environments.

I think our original process for developing the Windows port
was flawed. We should not really have created a separate
branch just for Windows.

I think we really need to start again with the up to date
Linux sources: (1) Test the build of that source code on
Windows. (2) If the build fails, we can consult the diffs
with the old Windows sources to see if it contains a solution.
(3) after we update the sources then we must first check that
it still builds on Linux. (4) If it does then we can continue
to test on Windows at step (1). And iterate until we have
sources that work on both systems.

There are really not that many changes to the Axiom source
code on Windows. Most of the work done by Mike Thomas involved
patches to GCL to allow it to compile Axiom. And Most of these
have since been incorporated in gcl-2.6.8pre.

gcl-2.6.8pre does compile on Windows with one minor change. But
my first attempt to use it to compile the Axiom windows sources
fails with a data type conflict on MYCOMBINE in src/interp/
cfuns.lisp.pamphlet which is partly coded in C embedded in
Lisp and partly in Lisp.

This seems similar to a problem that Camm solved with using
2.6.8pre on linux.

Regards,
Bill Page.