Re: [Axiom-developer] Version numbers and merging improvements

[C Y]

--- Ralf Hemmecke <address@hidden> wrote:

> > Well, the cvs and arch trees currently represent the Gold archive.
> 
> And I hope they will be gone once trunk turns into Gold. Then
> everything is on sourceforge. Who would bother with checking out
> via CVS. Any reasobable programmer has by now heard of SVN and will
> certainly be able to install it on his machine and get the Axiom
> sources via SVN.

The last time this came up (I believe) was when the AxiomSources page
was reorganized.  Tim at that time identified where Gold lived in one
of his responses:
http://lists.nongnu.org/archive/html/axiom-developer/2007-05/msg00609.html
 
> Why should we bother an invest time in keeping Arch and CVS as
> mirrors?

I certainly would not be adverse to centering on GIT with the SVN
connection, but that one isn't my call.

> If there is a maintainer (you?) who has too much time, I don't care
> if someone maintains as many mirrors as possible. I won't and I will
> ignore them.

The Gold archives were set up by Tim and he retains control of those. 
Whether he would agree with using GIT's tag system (how does the
SVN<->GIT setup handle Tags?) to identify Gold trees I don't know -
Tim?  It would reduce the upkeep and Git does appear to represent the
best technical solution (with SVN acting as a friendly face for those
who don't care for GIT).

To me the best of all possible solutions would be to center on the
GIT/SVN setup, release tarballs of Gold as it is currently defined, and
once Silver is tagged for the next Gold release retire all branches
except the GIT/SVN system.  (Bill's mercurial and darcs mirrors I don't
think would be bothered by this, unless I'm misunderstanding how they
work.)

Incidently, while I'm thinking of it, does anyone have any experience
with the qgit interface?  http://digilander.libero.it/mcostalba/

> > Well, I suppose the function that prints the Axiom header could be
> > made to check if the branch and build used to make the binary 
> > correspond to a known release, and if so print the release number.
> 
> You seem to have too much time. Lucky you.

No.  What I want is a solution that will be good enough to remove the
need or motivation to solve the problem again later.  Even if we end up
re-coding it someday for the literate process, I would like the design
of the solution to any given problem to be solid or scale easily.

I'm with Bill et. al. - I want to be figuring out how to implement
units and errors in Axiom to make the best scientific CAS ever.  But
Gaby is right - we need the tools, and since they don't exist we need
to create them.  Since we do need to, I figure we might as well learn
the lessons of Maxima and Axiom, and try and make solutions that are as
future proof as possible.

Interestingly enough, I notice that at least in wh-sandbox (and
presumably in build improvements as well) the version info is set in
configure.ac.pamphlet and this feeds on through.  Gaby, Waldek - does
autoconf read the repository information and update the appropriate
fields or do you manually update configure.ac.pamphlet with each
check-in?

Cheers,
CY


 
____________________________________________________________________________________
Now that's room service!  Choose from over 150,000 hotels
in 45,000 destinations on Yahoo! Travel to find your fit.
http://farechase.yahoo.com/promo-generic-14795097