[Gcl-devel] RE: [Axiom-developer] Possible GCL 2.6.7 for axiom

[Page, Bill]

On Tuesday, April 26, 2005 10:16 AM Camm Maguire wrote:

> Hi Tim!  Great conference!

I second that!

> Here's my take home understanding of the remaining issues
> re: gcl/axiom:
>
> 1) reenable run-process
> 2) display images in tcl/tk
> 3) tcl/tk working in windows
> 4) Windows socket workaround for sman/hypertex/graphics

This is consistent with my recollection of our discussions
during the Axiom sprint session. And I am willing to help
with testing and debugging on the Windows platform.

But we should be clear I think, that this is not necessarily
the shortest path to a fully functioning version of Axiom on
Windows. The linux version of Axiom does not now use tcl/tk
for Graphics and Hypertex, i.e. those parts of Axiom that are
not already ported to Microsoft Windows. Instead these currently
depend directly on the X windows libraries. These should perhaps
be re-implemented in tcl/tk linux first before attempting the
port to Windows.

Because of this X windows dependency the shortest path,
though admittedly not necessarily the best path, for fully
implementing Axiom on Windows would probably be to use Cygwin.
Currently GCL on windows uses MinGW instead of Cygwin to
compile to native Windows but unlike Cygwin MinGW does not
provided an X windows compatible environment.

I think using native Windows applications on Windows is a
wothwhile target but it is sometimes hard for applications
that originate on linux/Unix.

In principle it is possible to continue to compile AXIOMsys
as a native Windows application and to run the Hypertex and
Graphics components of Axiom under Cygwin. This would make
it unnecessary to convert any already working code to tcl/tk.
As I understand it, these additional components of Axiom
currently only require a Unix-compatible C programming
environment. So all we would need for this alternative is
step 4) above - socket compatibility between the current
native Windows version of GCL and the additional C code
compiled under Cygwin.

Of course there are drawbacks. Using *both* MSYS/MinGW and
Cygwin under Windows would further complicate the Windows
build environment for Axiom. Currently there is no Cygwin
version of GCL. Further, the X windows user interface under
Cygwin might seem a little awkward for some Microsoft
Windows users. But it should still be possible to prepare an
auto-installation binary distribution that would make much
of this transparent for users who are not interested in
building Axiom from source.

>
> These might go into a quick 2.6.7 GCL release.

However perhaps there are other reasons which would also
motivate the development of GCL tcl/tk on Microsoft Windows.

> 
> 5) axiom configure script
> 

Yes! I think that would be a great step forward. I think
Axiom should conform more closely to the de facto norm
for building open source software.

> There are a few math bugs in addition for which I'd like
> to propose fixes and test.  Should I make a tla branch?
> If so, what should it be called?

I don't think a special tla branch would be necessary
unless you contemplate drastic changes to Axiom's algebra.
(I think there is already a tla branch for that purpose.
See http://page.axiom-developer.org/zope/mathaction/ArchUsage )

If you are just testing and proposing fixes you can do that
on a local copy of Axiom and then post them to IssueTracker
on www.axiom-developer.org This would allow other people to
test them and then they could be incorporated into the main
distribution.

Regards,
Bill Page.