[Piuparts-devel] retesting failed packages more often

[Dave Steele]

On Sun, Nov 20, 2011 at 3:25 PM, Andreas Beckmann <debian at abeckmann.de> wrote:
> On 2011-11-20 16:40, Holger Levsen wrote:
>>
...
>> ... filing bugs based on piuparts runs, as
>> this is IMO the biggest problem currently: there are too few bugs filed, thus
>> bugs are not fixed.
>
> I tried to look into this yesterday - reporting is not that easy. I
> reported 2 or three and usertagged around 8 existing ones ...
> Eventually I put too much work into making these reports...

I'll admit that I've been coming at this from a different angle.
Piuparts has automated mechanisms that work OK to flag errors back to
developers, even without bug reports. I'm thinking that the best bang
for the buck for Piuparts work comes from making sure that failures
are found in a timely manner, are assigned to the right package, and
are free from false positives and other chaff.

Over the last couple of weeks, I've submitted bug reports representing
half of the remaining packages in the sid dependency-failed-testing
queue (excluding perl), and I intend to continue. At the same time, I
truly believe that, for instance, eliminating the need for developers
to parse 30,000 spurious motd warnings provides more value to Debian
than an equal amount of time writing individual bug reports.

>> ... I'm
>> currently the only active maintainer of piuparts on piatti.debian.org. That
>> shouldn't block you from developing your ideas on another machine though!
>

On that note - how can I go about getting a snapshot of the logs on
piatti? It would take quite a while to get through the distribution
with the resources I have available.

> I currently have a setup based on piatti.git running to put stress
> testing on all my patches. It was initialized with the pass logs from
> piatti ...