Re: [Axiom-developer] compiling from source on Debian Lenny

[Tim Daly]

On 1/11/2011 2:34 PM, Alan Hutchinson wrote:
> Hello
> Thanks again for Axiom.
> I am trying to compile it from the file "axiom-sept2010-src.tgz".
> I have tried to follow all the instructions in
> http://axiom-developer.org/axiom-website/download.html
> and I think I have got them all right.
> As before, the regression tests report just a few errors.
>
> The only deviations I have consciously made from the compilation 
> instructions are additions of two bash scripts called
>   notdir
>   shell
> in
>   /usr/local/bin
> They are called early in the main Makefile.  It seems that Debian 
> Lenny does not have these programs.
> The scripts may not work as intended on all inputs, but they seem 
> adequate for this job.
Yes, these are make gnu-isms that were added a long
time ago. They are only used for trivial things.
> The result seems to be a working program.  When it starts, it announces
>                         AXIOM Computer Algebra System
>                        Version: Axiom (November 2010)
>               Timestamp: Tuesday January 11, 2011 at 14:54:35
> even though the source was dated "sept2010".  This may be because I 
> downloaded
>   //axiom.git.sourceforge.net/gitroot/axiom/axiom 
> <http://axiom.git.sourceforge.net/gitroot/axiom/axiom>
> as advised in the download web page
> http://axiom-developer.org/axiom-website/download.html
The timestamp is the time you compiled Axiom.

I am mildly surprised that it is announcing that it is
the November sources though. It isn't important because
the changes are not user-visible except in the documentation.

> Two questions and guesses:
>
> 1  Have I got a "GOLD" version of Axiom, or is this a "SILVER" version 
> with unstable updates added from the git repository?
The Gold version is released every 2 months as a stable version.
The axiom-developer download page contains Gold versions. The
github release is also a Gold version. If you visit
http://axiom-developer.org/axiom-website/releasenotes.html
you can see what changed between Gold releases.

The silver versions are at sourceforge.org and savannah.org.

The silver versions are released potentially several times a day
depending on how active I am. Silver releases are the last tested
version on my machine. They may not necessarily run everywhere.
They represent the bleeding, but tested, edge of code.

I cannot think of a good reason to use Silver releases.
They are posted so that all development is public. If you visit
http://axiom-developer.org/axiom-website/patches.html#latest
you can see what patches were applied since the last Gold release.
Scroll backward and you can see every patch applied since Nov, 2007

If you are interested in any particular patch you can click on
the date field and see the diff-Naur patch itself as in:
http://axiom-developer.org/axiom-website/patches/20110110.01.tpd.patch
which was applied and tested last night. Patches are named by the year
(2011), month (01), day (10), number that day (01), and author (tpd)

I am currently rewriting the system to use only latex and eliminate
the use of noweb. End users will not care about this change but it
is one less package dependency.


> 2  The tests I have noticed which fail are regression tests.  I have 
> not noticed the make system run any other kinds of tests.
>    Are they appropriate?  Should I be running a different test suite?
>    If so, what shell variable should be set before starting "make", 
> and to what value?
The system runs all of the available tests. Some of the tests
are failing but not in any important way. Some of the tests are
failing deliberately to test cases that should fail. I plan to
move these out of the default test suite but that is a very low
priority issue. Ignore the failing tests.

There are no interesting shell variables for user use.
You can watch more details of the build with the command line:
   make NOISE=
which turns on all of the output of the compiles and tests.
This isn't necessary as the test output lives in int/input/*.output
and the results of test success or failure is in int/input/*.regress


> Best wishes
> Alan H.
>
>
> _______________________________________________
> Axiom-developer mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/axiom-developer