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

[Tim Daly]

On 1/12/2011 6:42 AM, Alan Hutchinson wrote:
> Hello Tim
> Thanks very much for your time.
> I should have made clear: I don't want software at the "bleeding 
> edge".  I am trying to be sure that the system I run is to "Gold" 
> standard.  There are enough complications in my life already.
In that case, download the sources from axiom-developer.org
> Are you prepared to answer two more questions?
>  - It seems that "sourceforge.org <http://sourceforge.org>" and 
> "sourceforge.net <http://sourceforge.net>" are the same.  You wrote
>       The silver versions are at sourceforge.org 
> <http://sourceforge.org/> and savannah.org <http://savannah.org/>.
>    Does that mean the git repository which I copied from there will 
> contain "Silver" material?
Yes, those sites contain silver material.
>  - Can you say roughly how many tests are likely to fail?  The one 
> time I let them run to completion, there were 13 such.  The only 
> summary I still have of the test output will be attached.  One in 
> particular, using file richlog300-391, produced 164 errors from 460 
> stanzas.
The failing test cases you see are normally failing.

Many of the tasks I undertake will take a very long
time to complete due to the complex nature of the problems.
I write tests to ensure that I remember the problems that
need to be fixed. Think of it as test-driven development.

One of those tasks is the integration of Albert Rich's
pattern matching software for integration. I have created
several thousand test cases as part of this work. The
test cases are all in the richXXX files. They contain
examples which Axiom cannot compute but which will be
solved when I finish Albert's new work.

I am undertaking new research with Albert's work, trying
to integrate ideas from David Parnas to create a more
complete and theoretically sound ordered organization on
which to base some theory ideas I have.

As a policy, all Axiom test cases are run all the time.

These test cases are "known failure" cases and are not
a cause for alarm. As I said, I ought to split these
failing cases into a separate class but that is a low
priority item since the failures will eventually go away.
Some of them are marked in the regress files as ok to fail.
I will mark more of them and filter the results accordingly.

There was a long and painful discussion on the mailing
list about working in public. We agreed that there would
be "time-boxed" Gold releases of Axiom every 2 months and
that all of my work would be immediately posted in Silver.

The advocates of that policy have left to create their
own projects (where they do not follow the policy they
advocated) but the Axiom policy still exists. Thus,
you get to see "intermediate states" of work that can
take a year or more to complete.

Tim Daly