axiom-developer
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Axiom-developer] Re: making ccl on cygwin


From: root
Subject: [Axiom-developer] Re: making ccl on cygwin
Date: Sun, 17 Nov 2002 19:42:38 -0500

I've uploaded the machinery to build bootsys.

In theory you should just need to type make and it
will eventually build obj/linux/bin/bootsys.
You should be able to run this image thus:
  cd obj/linux/bin
  ./bootsys

re: Cygwin. That's fine. I wasn't objecting to using
Cygwin but just passing along information.

re: diff and patch. yes. you should be able to 
   diff -Naur oldfile newfile
do man patch for details

re: using CVS. yes, we'll be using CVS to make these
changes shortly. At the moment I'm doing "forward" work
on CVS, that is, I'm uploading new code as I build it
and clean it up. I'm trying to get a clean version of
CVS and my working directories have acres of cruft that
should never get uploaded (e.g. the partial "original"
directories"). Once I get a grip on CVS we'll work 
exclusively thru CVS. For now, however, please bear
with me and use patch files.

Especially problematic about using CVS for the initial
port to Cygwin is that, to do it correctly, you need to
modify the top level Makefile with new stanzas. You should
be building on obj/cygwin, not obj/linux. I know how to
make that magic happen but I haven't sufficiently explained
it in the Makefile. I'm going to use your example (given
that you get to be the pioneer) to force me to explain
how and why to make changes while porting. 

You may notice three other things in the top level Makefile.
First, you made the contributors list. Second, there is a
FAQ section with your build questions. If you think of more
we should add them. Third, the author line is now Nicolas
Bourbaki. He apparently was a French general that screwed
up badly. In the 30s a group of French mathematicians started
using that pseudonym to publish their results. Given that 
there are many tens of contributors to Axiom it seemed 
unreasonable to attribute particular pieces to anyone and
unreasonable to attribute the pamphlet files to me. So, 
Bourbaki seems like a reasonable alternative.

I've upgraded my redhat 8 box to 3.2. I can't upgrade
the other boxes as I have to use them for other code.
Yes, 3.2 is correct but only if you use the latest
"standard" as the definition of correctness. Most of
the C and C++ code I have is badly broken by the 
language definition changes. I have a lot of things
that need fixing and I need the old compiler to keep
runnable versions of my code. 

The latest version, just to add to your confusion,
is running on GCL, not CCL. I've discovered a problem
with CCL and sent a note to Arthur. CCL isn't really
a common lisp and never pretended to be. However, it
has an obscure model of "compile-file" and I have to 
ponder how to build upon it. Just to continue development
I switched to GCL for the moment. If I've done my job
right you shouldn't notice.

You'll notice that the CVS directory is much lighter to
download. I've gone thru several download/upload iterations
so I can be sure you get a clean copy. The "original" directory
is still built but is empty. It was partially uploaded by 
mistake anyway. The annoying new/new still exists and I'll
fix that soon. The zips directory is lighter as cmucl is gone.
I'll add back an abbreviated version when I get the build to
succeed. It is a matter of some annoyance that CVS will not
delete directories.

Tim




reply via email to

[Prev in Thread] Current Thread [Next in Thread]