All,
I've added a section to SSN/78 about `Using the Starlink CVS
repository':
<http://www.astro.gla.ac.uk/users/norman/star/ssn78/ssn78.htx/N-
a2b5.html>. It's necessarily drafty, and has XXX at three points:
1. I seem to have lost my instructions on how to access the CVS
repository anonymously. Steve: could you patch the appropriate
instructions in here.
2. I mention bugzilla, but as I recall things aren't at present set up
so that folk can report bugs without having a cvs.starlink account. Is
that correct? Can someone fill in the details here?
3. If someone does have fixes to offer, whom do they contact? My
original idea was that the component.xml file would reliably have that
information, either as the code `owner', or at least as someone who has
worked on the code in the past and is contactable. This isn't the case
with the component.xml files at present. Do we want to make it so?
I got Steve to give a cvs.starlink account to Toby White, who's been
pushing the Fortran autoconf work ahead, so he's the first
non-Starlink, non-JAC person who can write to the repository; the
first, we hope, of many. Toby hasn't done anything yet because I
haven't yet merged in the patches he last sent me by mail. He isn't
likely to be writing to anything other than thirdparty/fsf/autoconf, on
a branch, and when he does the commit messages will be forwarded to
stardev as usual.
It might be worth thinking non-urgently about the implications of this,
and for example making sure that all or most of the repository is
covered by entries in CVSROOT/loginfo which make sure that anyone
committing either overenthusiastically or through a runaway cvs command
gets noticed by someone who has some awareness of what a commit in that
component might mean.
Also, ought we to make it a rule that everyone with an account on
cvs.starlink is also a member of stardev? In fact, they'll probably
have to be, if any commits to the top level, buildsupport/starconf, or
thirdparty/fsf are to appear on stardev via loginfo (so Steve, Toby
will have to be thus added if he's to work _visibly_ on autoconf).
That might have an impact on the nature of the discussion on stardev,
in particular stopping me making injudicious remarks (hmm).
I think we're near the point where cvs.starlink becomes a routine
production system, and the pace of SSN/78 change slows down. I'm now
looking through that document with that in mind, hence this new
section. Are there another other implications of this transition? Do
you agree that we are reaching this stage?
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|