|
piuparts.debian.org / piuparts.cs.helsinki.fi
|
|
piuparts is a tool for testing that .deb packages can be installed, upgraded, and removed without problems. The
name, a variant of something suggested by Tollef Fog Heen, is short for "package installation,
upgrading and removal testing suite".
|
|
It does this by creating a minimal Debian installation in a chroot, and installing,
upgrading, and removing packages in that environment, and comparing the state of the directory tree before and after.
piuparts reports any files that have been added, removed, or modified during this process.
|
|
piuparts is ment as a quality assurance tool for people who create .deb packages to test them before they upload them to the Debian package archive. See the piuparts README for a quick intro and then read the piuparts manpage to learn about all the fancy options!
|
|
To make sure piuparts is run on all packages, piuparts.debian.org was set up as a service running on
piatti.debian.org.
This machine was generously donated by HP
to run piuparts on the Debian archive and is hosted as piuparts.cs.helsinki.fi by
the University of Helsinki, at the
Department of Computer Science
in Finland.
As this is still being polished, see the piuparts wiki page for more information on piuparts development and the piuparts setup on piatti. Better reports and statistics as well as PTS integration is planned. Join #debian-qa if you want to help.
|
|
These pages are updated daily.
|
News
|
|
2009-10-08: Reschedule testing for ~2000 failed packages in sid, which failed because of a problem when minimizing the chroot at the beginning of the piuparts tests. As of today, piuparts running on piuparts.debian.org does not minimize the chroots anymore.
|
|
2009-09-18: Reschedule testing for 17170 (successfully tested) packages in sid, to make sure they still install fine with dependency based booting enabled now in sid. Throwing away 42806 (successful) logfiles from those packages :-)
|
|
2009-09-16: Reschedule testing for 233 failing packages in sid which were affected by #545949. No packages in squeeze were affected.
|
|
2009-06-20: Failed logs are not grouped into (at the moment) seven types of known errors and one type of issues is detected in successful logs.
|
|
2009-06-06: Reschedule testing for 163 successful and 27 failing packages in sid which were affected by #530501. Once openssh 1:5.1p1-6 has reached squeeze, this will be done again with 194 packages there.
|
|
2009-05-27: Throw away all failed logs as there was a bug in piuparts leading to use a more uptodate mirror for getting the list of available packages and another for doing the tests. This lead to at least one fixed package which was incorrectly tested as failing, as an old version of the package was tested. To rule out some false positives about 1000 packages will be retested, but on this machine this will only take about a day :-)
|
|
2009-05-11: Filed #528266 and made piuparts ignore files in /tmp after purge. This got rid of 20 failures in sid and 14 in squeeze.
|
|
2009-05-06: Only believe statistics you faked yourself! Up until today piuparts used to include virtual packages (those only exist true the Provides: header) into the calculations of statistics of package states and the total number of packages. Suddenly, sid has 2444 packages less!
|
|
2009-05-01: All packages in squeeze and sid which can be tested have been tested. So it takes about one month to do a full piuparts run against one suite of the archive on this machine, that's almost 1000 packages tested per day.
|
|
2009-04-20: Deleted 86 more failed logfiles (out of 692 failures in total atm) which were due to broken packages, which most likely are temporarily uninstallable issues - a good indicator for this is that all of those failures happened in sid and none in squeeze. For the future there is a cronjob now, to notify the admins daily of such problems. In more distant future those issues should be detected and avoided.
|
2009-04-18: Deleted all 14 failed logfiles which complained about /var/games being present after purge, as this ain't an issue, see #524461.
|
|
2009-04-04: Deleted all failed logfiles so far for two reasons: until now, only three out of ten failure types where logged with a pattern to search for in the logfiles, now this is done for all ten types of failures. And second, the way of breaking circular dependencies was not bulletproof, thus there were false positives in the failures. Now it should be fine, though maybe this will lead to lots of untestable packages... we'll see.
|
|
2009-03-19: lenny2squeeze is not needed, so all logs for squeeze (as well as lenny2squeeze) were deleted. (As squeeze now includes two kinds of tests: installation and removal in squeeze, and installation in lenny, upgrade to squeeze, removal in squeeze.)
|