Updates on working towards a new Debian package for ess

[Martin Maechler]

>>>>> "DE" == Dirk Eddelbuettel <edd at debian.org>
>>>>>     on Sun, 18 May 2008 17:07:30 -0500 writes:

    DE> [ Continuing to spam poor Peter G. ... ]

    DE> On 18 May 2008 at 23:53, Stefan Theussl wrote:
    DE> | > I am confused. I thought the refcard issue was fixed in svn by Martin as per
    DE> | >
    DE> | >    r3888 | maechler | 2008-05-16 04:55:32 -0500 (Fri, 16 May 2008) | 3 lines
    DE> | >
    DE> | >    Usepackage {fullpage} instead of {a4};
    DE> | >    {svn} is now also carried by ubuntu
    DE> | >
    DE> | > What did you change relative to this?
    DE> | >   
    DE> | Actually, nothing. I just thought it would be better to release the 
    DE> | latest upstream release than the current development version. I know 
    DE> | that Martin already fixed this. The refcard.tex I changed was just the 
    DE> | one from the ess.orig.tar.gz which in turn was generated from the tagged 
    DE> | 5.3.7 release.

    DE> I see. I sort-of figured it out from your tarball.  And that did not give me
    DE> a warm and fuzzy feeling. As far as I can tell, we did our bug triage
    DE> relative to the SVN, we did the test releases relative to the SVN, and we
    DE> tested relative to the SVN.  All of a sudden I saw a (arguably small) set of
    DE> diffs without explanation given ...


    DE> | >> You can download it from:
    DE> | >>
    DE> | >> http://statmath.wu-wien.ac.at/~theussl/downloads/ess.org.tar.gz

    DE> [Btw: the name is $package-$version.orig.tar.gz. not .org.tar.gz ]

    DE> | > I think I'd be more comfortable sticking with svn given that it would
    DE> | > be clearer exactly what code we released off.


Yes, definitely!   I agree very much with Dirk.
That's why it made so much sense to give you svn write access!

    DE> | > It is not uncommon to release Debian package from svn. We could call
    DE> | > this 5.3.7.svn3892. In a way, that would be more honest than calling a
    DE> | > snapshot 5.3.7 when there really is a 5.3.7 release.
    DE> | >   
    DE> | Ok, sounds good to me.

    DE> Thinking about this more today, I feel even more that we should simply
    DE> release what we currently have in SVN, but call it

    DE> 5.3.8~svn3892

Yes, indeed!
That (calling it something like "5.3.8 pre-releasish") was may
immediate reaction when starting to read the thread.

Using an exact svn revision for such a release is very much in
line with my own thinking (which of course is "biased" from the
way we do it with R).

    DE> The ~ sorts lower in Debian, so a package 5.3.8-1 will cleanly replace
    DE> this. I do the same with R pre-releases (alpha/beta/rc)

    DE> That way nothing needs to be modified from the svn tree.

which is really ideal.

    DE> Now, as I am new here, I would like to hear from the ESS guys on the list:

    DE> -- is the current SVN fit for releases ?

not always, but it is now:  AFAIK, it has no newly detected
bugs, only has some fixes and new features wrt 5.3.7.
As a matter of fact, I have thought of releasing 5.3.8 more than
once during the last weeks, but have been stopped by some
improvement proposals / feature requests that we (ESS-core) have
kind of accepted and promised to consider.


    DE> -- are you comfortable seeing it in Debian, or
    DE> -- would you rather that we took 5.3.7 as it was ?

    DE> As I argued before, pre-releases are pretty standard in Debian for many
    DE> parts, including (or even in particular) the main GNU toolchain.   

Then, I think it's fine with me to let you release

      5.3.8~svn<nn>


    DE> Thanks for any feedback!

You're welcome!
Martin