Re: Checks for urls in packages - extend lintian
Le Mon, May 21, 2012 at 07:45:36AM +0200, Simon Kainz a écrit :
> On 05/19/12 02:39, Charles Plessy wrote:
> >
> > http://wiki.debian.org/UpstreamMetadata
> >
> > http://anonscm.debian.org/viewvc/collab-qa/packages-metadata/
> >
> > If your proposition does not fit Lintian's design goals (as it has been
> > suggested in this thread), would you like to use our platform to monitor the
> > VCS fields ? As a first step we could extend the gatherer to include the
> > debian/control files as well, and then we could write some scripts, which would
> > parse the VCS fields as provided by the Debian archive, check if they are
> > correct, and if not, check if the packages-metadata repository provides correct
> > data, indicating that the problem has been fixed in the package's VCS.
>
> Yes, i really would like to use your monitoring platform. Currently I
> interate over the list of VCS-* and Homepage entries and do some curl
> magic and/or try the get the state of source repositories. The results
> are then displayed in a list. So this more like the PTS, displaying
> status data and not something people get notifed by.
Hi Simon,
I have modified the gatherer so that it will collect the Debian source package
control files.
As a user, you can:
- Trigger the gatherer to load or refreshing a control file by accessing the
package's metadata for instance with the an URL like the following:
http://upstream-metadata.debian.net/<sourcepackagename>/ping
Note that there is a temporisation system, and by default the gatherer will
not attempt to refresh the files more than once per hour.
- Access the gathered files in http://anonscm.debian.org/viewvc/collab-qa/packages-metadata/
Note that the gathered files are pushed daily.
These two limitations (temporisation, daily push) are quite arbitrary can probably be optimiesd.
The Umegaya system is passive, so it will only collect files for which one
tries to access metadata. For the moment, I have a cron job that accesses (and
therefore forces a refresh) all the packages related to the scientific Debian
Pure Blends, plus a listener that pings the gatherer each time an upload is
announced on debian-devel-changes@lists.debian.org.
Another way to manage the metadata would be to join us maintaining
upstream-metadata.debian.net. It is hosted on blends.debian.org and I think
that I can give you access if you are interested. There, we can probably
arrange things so that you can push changes to collab-qa/packages-metadata, and
you can use the umegaya-adm command to directly manipulate the database. This
will give you the possibility to resolve conflicts between the the VCS field
provided by the Debian archive (via debcheckout), which is the reference one,
and the real location of the package's VCS.
As a developer, you can also help a lot by contributing to develop the
gatherer (Umegaya). The source code (and the manual pages for umegaya-adm
and the rest) is on Alioth: http://git.debian.org/?p=users/plessy/umegaya.git
Among the things that would be welcome, there would be to give more ways to
(trusted ?) users to maniuplate the database, like allowing to completely reset
the information for a package via HTTP DELETE, write VCS hooks that would ping
Umegaya when the control, copyright or upstream files are modified, and
implement consistency checks, which would be doing the monitoring you are
intersted in, and more.
Cheers,
--
Charles Plessy
Debian Med packaging team,
http://www.debian.org/devel/debian-med
Tsurumi, Kanagawa, Japan
Reply to: