[csw-maintainers] Package stats collection / retrieval in checkpkg

Maciej (Matchek) Blizinski maciej at opencsw.org
Wed Jun 30 17:22:18 CEST 2010


[discussion moved from pkgsubmissions]

No dia 29 de Junho de 2010 16:47, Philip Brown <phil at opencsw.org> escreveu:
> On Mon, Jun 28, 2010 at 3:53 PM, Maciej (Matchek) Blizinski
> <maciej at opencsw.org> wrote:
>>
>>> What did you have in mind?
>>
>> I was thinking about exactly that - the package build level.  The
>> earlier a problem can be detected, the better, right?
>
> I thought you might :)
>
>>  For example:
>>
>> maciej at current9s :~/src/opencsw/gar/gar-new-checks > bin/checkpkg
>> ~/exp/*cups*sparc*
>> Analyzing collected statistics...
>> CSWcupsdev:
>> CSWcupsd:
>> CSWcupsclient:
>> CSWlibcups:
>> (...)
>
> And where are these statistics "collected"?

I wrote it up at http://wiki.opencsw.org/checkpkg#toc17  -- I've just
updated the 'Stats collection' section.

> Might it be better to define an API for accessing the "official"
> filename path repository, rather than hoping the package has been seen
> by that build machine before?
>
> Ideally, some API that allows for a common shared local cached copy,
> rather than "every maintainer has their own cache" ?

I never thought of it as something that needs caching.  The separation
between stats collection and analysis exists mainly because it makes
the development easier.  Operationally it doesn't make that much
difference for one maintainer.  The positive impact in operational
terms is the biggest when analyzing the whole catalog.

There are also data about the system files and the catalog, which are
in a sqlite database, which caches the content of
/var/sadm/install/contents.  This might be beneficial to share, but it
requires careful planning and needs to armored against race conditions
during updates.  It's something I intend to do, but it'll take time.


More information about the maintainers mailing list