[Axiom-developer] axiom--windows--1--patch-11

[Page, Bill]

Tim,
 
As per our discussion offline, I have committed a new patch
for the windows branch that moves the awk script from the
src/Makefile.pamphlet to the src/algebra/Makefile.pamphlet
where it "logically" belongs. This script is now extracted
(by notangle) from src/algebra/Makefile.pamphlet and called
by the revised src/Makefile to explicitly add the pamphlet
rules to the new src/algebra/Makefile. All the other rules
in src/algebra/Makefile are implicit rules. As a result,
the new src/algebra/Makefile is much smaller than it used
to be and I think it is easier to understand.

Extracting the script by notangle rather than embedding
it in the Makefile eliminates most of the hieroglyphic
"line noise" to which you objected. I have also completed
and added the additional detailed documentation that
you partially drafted. A new developer can read the
src/algebra/Makefile.dvi documentation to see exactly
how this works.

I have not yet changed where the databases are built (now
it is in the etc/Makefile in order to make use of a simple
wild-card specification for the prerequisites). As you
pointed out, it could be argued that this also logically
still belongs in the src/algebra/Makefile.pamphlet. This
could be handled in a manner similar to the above with the
etc/Makefile using notangle to extract and execute this
stanza from the src/algebra/Makefile.pamphlet file. Then
at least of this documentation for building the algebra
would be in one .dvi file.

Or alternatively the database stanza could be returned to
the src/algebra/Makefile itself provided that we first
did some substitutions on the ORDER variable to construct
the explicit list of *.NRLIB/code.o dependencies.

Which of these options would you prefer? Or can we leave
this one as it is?

While doing this work, I took the opportunity to review
the Makefile.dvi files for src, src/algebra, src/input and
src/etc. Since these where in rather poor shape due to some
simple LaTeX and noweb coding problems, I presume that no
one has been paying much attention to them. In any case
I think I have corrected most of the problems with these
so that with the additional documentation, they should now
be of more use to developers.

Let me know when you have had a chance to review this
as I would like to make similar changes to axiom--main--1
branch.

My next patches to the axiom--windows--1 branch will be
to add the tm_axiom interface code and the NSIS install
program scripts. (patch-12?)

There are also some outstanding unix compatibility issues
that will require changing 'cat' and 'rm' system calls 
in the BOOTSYS/INTERP code to GCL native code to improve
portability. (patch-13?) Are there any lisp programmers
out there who would like to take this on?

Finally, we need to address the porting of the graphics
and hyperdoc code to Windows. This is a big problem if
our target remains native Windows because of the lack of
the X library under Windows. One possible way around that
might involve using the X environment provided by cygwin
under windows. For windows users this might be considered
a step backward since installing and using X under cygwin
is rather complex and somewhat "unnatural". The alternative
is to find another graphics environment for Axiom graphics
and hyperdoc - hopefully one that is more easily portable
between linux and windows. Are there any developers out
there who have some experience with this?

Regards,
Bill Page.