[Piuparts-devel] modernizing aspell/ispell packaging to become piuparts-clean

[Andreas Beckmann]

Hi,

I'm looking for help to get the remaining ispell/aspell packages updated
to the new piuparts-friendly dictionaries-common.

Since jessie we have a dictionaries-common package that simplifies
packaging ispell/aspell dictionaries and takes care of the cleanup after
the removal, thanks to Agustins work. (Please correct me, if I got the
details wrong.)

The previous packaging style involved shipping some empty files in /var
(sometimes without md5sum entries) where the postinst generated
something for the dictionary. dpkg would then remove these files on
package removal (instead having the postrm script do that).
This is not piuparts-friendly since it produces this:

1m15.9s DEBUG: Starting command: ['chroot',
'/srv/piuparts.debian.org/tmp/tmpAeMenR',
'tmp/scripts/pre_remove_40_find_missing_md5sums']
1m15.9s DUMP:
  aspell-ml: FILE WITHOUT MD5SUM /var/lib/aspell/ml.compat
  aspell-ml: FILE WITHOUT MD5SUM /var/lib/aspell/ml.rws

or that:

0m26.3s ERROR: FAIL: debsums reports modifications inside the chroot:
  /var/lib/aspell/bn.compat
  /var/lib/aspell/bn.rws

https://piuparts.debian.org/sid/missing_md5sums_issue.html
https://piuparts.debian.org/sid/debsums_mismatch_error.html

I'd like to see all the dictionary packages updated for stretch, so that
we could run piuparts on stretch without out reducing the debsums
problems to a warning (so far we had to run all stable releases with
--warn-on-debsums-errors).

Only a few bugs are filed against the affected packages so far, and I'd
like to see bugs filed with patches :-) Which is probably easy once
someone has worked himself into the modern way of packaging a
dictionary, it just needs to be applied to all of them. Hopefully.

I'm afraid, I already have too much on my own plate, so I'm looking for
someone who could handle this "transition".
Therefore I've included Tobi (who was looking into this during his DM/NM
process) and Mattia (who seems to have done a lot of QA work recently,
too) in the list of recipients.

I also tried to come up with a ben file to check which dictionaries have
a dependency on a recent enough dictionaries-common (not sure whether
the versions are correct)

is_affected = .build-depends ~ /dictionaries-common-dev/
            | .depends ~ /dictionaries-common/;
is_good = .depends ~ /dictionaries-common\s*\(\s*>=\s*1\.2[3-7]/;
is_bad  = .depends ~ /dictionaries-common\s*\(\s*>=\s*(0|1\.1|1\.2[0-2])/;


Andreas