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

Whither CERNLIB (, Paw, Geant3.21) in Debian? RFA / future plans

[Please follow up ONLY to debian-science]
[CC'ed to the Fedora CERNLIB maintainer and to Ubuntu MOTU-Science, for
their information]

I'm cross-posting this to debian-user on the off chance there may be
some Debian-using physicists there who don't follow debian-science,
which seems to have morphed into a developer list lately.

I'm seeking one or more people to take over maintenance of the following
FORTRAN- and physics-related source packages.


(This message does not apply to my unofficial Geant4 .debs)

Since my employment is no longer in the physics field, and family
obligations continue to reduce my amount of free time for Debian, I no
longer have time or interest to maintain these packages properly.  The
maintainer switch can be gradual (co-maintainership for "apprenticeship"
for a while is OK by me, but can't be too much of a demand on my time)
but needs to be complete well before the release of Squeeze (Lenny+1).
Please follow up to debian-science@lists.debian.org if you want them.
Nasty details below.

For the reasons below, I think these packages are much too difficult to
be maintained by QA or by a team (e.g. debian-science) that is not
specifically targeting them.  Therefore, if there are no offers, I will
request their removal from Debian after the release of Squeeze (assuming
they do not get RC bugs filed against them prior to that); until then I
will maintain them on a best-effort basis, with a note in NEWS.Debian to
the effect that they will go away after Squeeze release.

Prospective maintainers should be aware that the packages are
particularly challenging for the following reasons:

* Most of the code is written in either K&R C or FORTRAN IV, with
arch-specific #ifdef's around code for cutting-edge machines like PDP
11s.  (That is not a joke.)
* Bugs in new versions of gfortran often break various bits of code on
obscure architectures and make the test suite fail.
* The upstream build system is based on imake.
* The Debian packaging is dpatch + tarball-in-tarball, with some hacks
to make it possible to use Debian patches for building on non-Debian
* GUI code is based on Motif ... building the programs against Lesstif
seems to mostly work (with some patches) but there may be as-yet
undiscovered GUI bugs.
* There was a lot of non-free stuff that had to be cut out of the source
packages to get the existing tarballs suitable for main.  However, this
may not cause you much work, since...
* Upstream has been dead since 2006, and moribund for several years
prior to that.
* There is no upstream support for shared libraries, nor for builds on
Linux arches other than m68k or i386.  Large patches have been hacked
into the Debian source packages adding these.
* Upstream source is not 64-bit clean.  Properly fixing this would
require man-years of work.  In the meantime, a fellow named Harald Vogt
has hacked together some patches that mostly make things work on 64-bit
machines... but the patches are so fragile that they break unless
programs are statically linked against the CERNLIB libraries on 64-bit.

I will wait a week or two to hear from prospective maintainers.  If
there are any, I will email them with more info at that time.  If not,
I'll submit official RFAs to the BTS and send a note to debian-science
that PAW and friends are slated for doom.  In the meantime, prospective
maintainers may want to take a look at my "CERNLIB on Debian" pages at
http://people.debian.org/~kmccarty/cernlib/ and also at the source
packages (prepare to be horrified).

best regards,

Kevin B. McCarty <kmccarty@gmail.com>
WWW: http://www.starplot.org/
WWW: http://people.debian.org/~kmccarty/
GPG: public key ID 4F83C751

Attachment: signature.asc
Description: OpenPGP digital signature

Reply to: