RE: [Axiom-developer] Release?

[Bill Page]

Mike,

On Monday, December 13, 2004 7:11 PM you wrote:
> ...
> Perhaps it would be better if I just zipped up my local
> stable build and uploaded it so that you can merge the
> code changes into the Windows arch branch as and when you
> are able.  Let me know if you would like that.

Yes, I would be very pleased if you could do that. I will
look after merging it into the arch branch.

I suggest that before zipping up your local stable build
that you *first* prune all the irrelevant stuff out of the
axiom zips directory. Only gcl-2.6.5 and related patches
are needed for the windows build. The rest of the files
can not be used with windows but took up more than half
of the space, last time you did this kind of upload.

> 
> Regarding cat, cd etc; even if you use the respective DOS 
> commands, those commands will still choke on the forward
> slash canonical path format used by Windows GCL.

I wrote a simple cat program that does this for Windows and
it is included in the most recent version of the Windows
installer. As far as I know, it is only needed for trivial
things like )summary.

> Sure, you can knock the paths into shape before calling
> system() but that still (apart from being a hassle) leaves 
> you open to the other problems I have previously mentioned.
> For this reason I recommend either doing those commands in
> Lisp as suggested by Camm last week or as a second best/
> temporary fix - cfuns-c.

Once I have a working windows build, I can debug the cfuns-c
if necessary. But I agree that the best approach is the use
the GCL native commands were possible (everywhere I think).
I think very soon cfuns-c should be made to disappear.

> I believe that si::chdir is the GCL directory setting Lisp
> equivalent.

Yes, that works. In Axiom the following command changes the
default directory:

  )lisp (si:chdir "new/directory")

but the native Axiom command

  )cd new/directory

does not work. Odd.

> I have so far only addressed mkdir via cfuns-c if my memory
> serves, but I believe that the best solution would be
> to add ensure-directory-exists to the stable build of GCL.
>

I agree.
 
> I would personally urge you to seriously consider solving 
> those system() related issues before doing an advertised
> public release.  They aren't major problems to solve and
> could have a fairly large negative impact on user expectations
> if left untended.

Well, there is still the open source philosophy "release often
..." So far we are way behind in that process. I am really not
so sure exactly what user expectations really are, but I do know
that nearly 250 people have already downloaded the current
windows binary. I am sure that if we have more (many more?)
interested users, then most of these issues will be resolved
sooner. After all, we are not "bringing a product to market"
here! It is more like we are trying to build a developer and
user community.

> 
> Finally, thanks for the enormous effort you have put into the 
> new installer.

Actual Dan Martens did most of the work. I am just tweaking if
by applying some off-the-shelf recipes from the NSIS web site.

> 
> More next week,
> 

Till then ... I hope you have time for the upload.

Cheers,
Bill Page.