I propose the following as the new output format of piuparts:
* Standard output will report tests that have been run, and their
results. This output will be similar to the lintian output. This
output should usually be pretty short, and to the point.
* Additionally, a log file will be written, similar to the current
output. It will be named after the (first) package being tested
(or simply piuparts.log?), and placed in the current working
directory by default.
* Failed tests may _optionally_ also write dumps of the virtual
environment (chroot, eventually maybe kvm images) so things may
be debugged.
For example:
I: scenario install-remove-purge
I: environment minimal
I: test working-symlinks PASS
E: test leaves-no-extra-files FAIL /etc/foo.conf
E: test leaves-no-extra-files FAIL /var/lib/foo/
E: test leaves-no-extra-users FAIL foo
I: test install-remove PASS
What piuparts does is run a few different scenarios, and perform
tests for each scenario. The current scenarios are:
* install package, then remove it, and finally also purge it
* install package from repository, upgrade to new version provided
by user, then rmeove and finally purge
Additionally, these scenarios can be run in different kinds of environments,
such as a clean, minimal system, or one with lots of other junk installed.
The test output needs to record these.
Then there's the actual tests. They need to get a name, so they may
be easily referred to. When a test fails, it needs to report something
to make it clear what is wrong. In the sample output above, the failed
tests report the filename or user that caused the failure. Note that if
there are multiple problematic files, they are all reported, one by one
(for easy parsing).
How does this sound? It is important that those who might use piuparts
be able to understand the output.
--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
