We held a short meeting about what we plan for the further lintian
development ("Version 2.0"). Here the minutes:
(these are solely my perception of the meeting, HE and jvw might
want to add/correct some points)
Minutes of the Lintian IRC meeting
Attendees:
- Frank Lichtenheld 'djpig'
- Marc Brockschmidt 'HE'
- Jeroen van Wolfelaar 'jvw'
Place:
#debian-qa on irc.debian.org
Time:
Sunday, July 31th 17:00 CEST
Since the meeting was somehwat uncoordinated, I present the discussion
based on topic rather than how it actually happened:
HE:
- moving svn to alioth
there already exists an alioth project for lintian but joy is the only
project admin and some people are missing from it (like HE).
jvw will make the alioth move (also adding the old CVS history)
and will try to make sure we (the attendees) become project admins
- getting new maintainers into the team
most important issue probably to improve the internal documentation.
HE suggested including some sample output of the collection scripts
in the source as part of this. Depending on how much we want to change
with the "Lab", these needs to be postponed until we have a clear idea
of what we end with
- mixing lintian with pbuilder
debated, piuparts offers some of this already (possible to add missing
features like detecting broken symlinks and missing shlib dependency
problems?)
jvw and djpig not very happy with adding such functionality to lintian, too.
Perhaps a common frontend to debdiff/lintian/linda/piuparts would be better
instead of duplicating too much functionality? Other problem is requiring
root access for chroot operations. But the missing tests are certainly
worth to be implemented somewhere...
djpig:
- refining the tag classification (into something more detailed then
error/warning/info)
The general idea is to make a two-dimensional classification after
severity and significance. We agreed on the following levels:
severity: cosmetical I/policy should W/severe but !policy E/policy must E/
significance: maybe ?/pretty sure !/absolutely sure !!
suggested format of output: letter like before combined with significance
modifier, e.g. E? = error, but with a high possibility of false positives.
In the .desc file the classification should expressed numerically and
it might be a good idea to add an output format that uses this to make
parsing it easier
jvw:
- getting more modularisation, everything as included perl
no real debate over this. Almost done for the checks, needs to
be done for collections and other scripts, too.
- ditching the lab
jvw's proposal: instead of the lab, have some cache for
intermittent collection data so that, on a filled cache, you
can do cross-package stuff and lintian gets two modes, one
only doing the tests that only requires the intermittent results
of it self and one that also uses knowledge of that data for
other packages in the same suite.
jvw will make a more verbose proposal of this
Random stuff mentioned during the discussion:
- Add a banner to the lintian output that more information can be requested with -i, many people don't realise this option even exists
- we will (need to) branch out to 2.0 when/if we start with the move away
from the lab
- djpig will make a 1.23.11 upload soon [already done]
Gruesse,
--
Frank Lichtenheld <djpig@debian.org>
www: http://www.djpig.de/
Attachment:
signature.asc
Description: Digital signature