[Axiom-developer] Re: axiom--windows--1 build

[root]

Bill,

>Looks good! I did a tla update of axiom--windows--1 and ran
>the build last night. Now the build completes the same way
>as Mike's original.

humpf. I started my build at 2:30am and it's only at layer 9
(of 23) of the algebra build. Windows development is going to
be painful for me.

>
>Great.
>
>So what bugs should we kill first? These are the ones
>I know of so far:
>
>1) )cd with no parameter still returns path error, but
>   should be user's HOME directory. Should GCL expand ~
>   as USERPROFILE on windows?.

I believe this will also impact the underlying lisp since
there are issues such as truename does not handle ~ properly.
Perhaps we can wait until this one is solved by GCL? 

>   Note that )cd pathname now works and so does )cd .

In your version or the --patch-5 version? I can't test
it until the compile completes which won't be until
later this afternoon (auuggghhh.. 12 hour builds!)

>2) Use of cat, rm, ls in BOOT must be replaced with lisp

This might be a good place to start. There is a new file
src/interp/port.lisp.pamphlet which would be a good place
to put this code as I'm sure every lisp will do it differently.

>3) Add NSIS installer script and build commands to Windows
>   section of Makefile.pamphlet

I'd do this first. We could probably do a 'make package' where
the 'package:' stanza was target specific. Thus on windows
it would build using the installer and on linux it would build
using tgz. 

The reason to do this first is that there is expertise about
how to build the package which no one needs to learn. Makefiles
are a very good way to encode this expertise so "it just works".

>4) Test sockets in GCL for Windows
>
>5) Compile sman and graphics. What graphics API for Windows?
>   We do not have X but is there a Windows compatible
>   library? Can we support at least postscript file
>   generation?

I looked at this for a short while. This is going to be a long,
hard problem for every system. There is the possibililty that
we could use TK, which is an interpreted X-windows like front end.
Perhaps there is a windows version of TK. 

We do need to figure out the socket stuff. Ideally we could
leverage the effort by figuring out how to do sockets from 
GCL. Once we figure that out it should be possible to move
the socket code out of C and into lisp. This would make Axiom
more portable in general. Since we're going to struggle with
this anyway we might as well do it right.

>
>Anything else to start?

On Windows the lisp compiles are getting complains that '-mcpu='
is deprecated. Use '-mtune=' or '-march='. Are you seeing the same 
message?

t