RFC: Allowing ispell dictionaries hash build at postinst
Hi all,
Allowing all ispell dictionary packages to build their hash files from
package postinst is being considered. That is currently done by one of
the ispell dictionaries, and its case is explicitly mentioned in the
policy document, since it allows the sysadmin to generate the hash file
with different options.
What is being considered is to allow the same for the other dicts,
combined with some sort of semi-intelligent autorebuild behavior (that
is, not always autorebuild) and would like to know if there are
objections to this running programs from the postinst.
Advantages of this possibility:
The advantage would be to make ispell dict packages doing things this
way 'Architecture: all' and if some sort of intelligent system is
implemented to support this, that each time an ispell package with a
'compatibility level' higher is uploaded all such dicts would be
automagically rebuilt, so incompatibilities in the hash due to new
ispell versions are handled on the fly. That way any kind of ispell
buildhash parameters or format changes would make the dicts to be
autorebuilt if the ispell package raises its compat level. Note that
this is mostly useful for users of 'testing' or 'sid'.
Disadvantages of this possibility:
The disadvantage of this system would be a bit larger installed packages
and a bit longer installation time (timings show that this might not be
significant, building hash for 5 dicts in a 486 DX/4 took not much more
than a minute), as well as losing the extra testing step the
autobuilders provide. The buildhash step is probably the less prone to
errors in a ispell dict building, so probably this last is not important.
What do you think?
--
=====================================================================
Agustin Martin Domingo, Dpto. de Fisica, ETS Arquitectura Madrid,
(U. Politecnica de Madrid) tel: +34 91-336-6536, Fax: +34 91-336-6554,
email:agmartin@aq.upm.es, http://corbu.aq.upm.es/~agmartin/welcome.html
Reply to: