Web lists-archives.com

Bug filing for autopkgtest regressions? [Was: Re: appears to break multiple autopkgtests]

Hi Chris,

[This question came up multiple times already in my e-mails about
regressions in autopkgtests.]

On 27-06-18 23:48, Chris Lamb wrote:
> (Paul, might make more sense to file bugs in future? Much easier to
> track and/or re-assign if necessary..)

I do that when I am *somewhat* confident which package should in
principal fix the regression. However, often I lack the time to dive
deeper and/or am unfamiliar with the packages involved and/or how the
autopkgtest in question works. To circumvent that problem, I send the
e-mails that we agreed upon on this list, albeit I don't due that
automatically even, because several packages have an extremely large
fall-out. This fall-out may be real (it appears to me that the current
python3-defaults may be an example of that as packages seem to be not
ready for Python3.7) or mostly just a matter of re-triggering tests as
the stack is highly in flux and should be considered together (like the
recent KDE stack update).

However, I can see the point and I prefer bugs personally as well. So
let's check what people here think of my proposal for guidance for
"by-standers" (I mean people like me that are not maintainer of either
the triggering package or one of the packages which autopkgtest regresses).

If one (me) can't determine the likely principle package that needs
fixing after some effort [1], one bug (per upload of the triggering
package) can be filed against the trigging package with the other
packages in X-Debbugs-CC and in Affects at normal severity. This bug can
contain similar text as we agreed upon earlier [2] (I'll write a
proposal if this idea is not rejected).

PS: any tips on how to handle the current python3-defaults situation is
highly appreciated.

[1] Including, but not limited to, comparing the autopkgtest logs of the
passing and failing test, looking at patterns in the fall-out, examining
recent other regressions for the same failing autopkgtest, checking for
existing FTBFS or similar bugs.

[2] https://wiki.debian.org/ContinuousIntegration/RegressionEmailInformation

Attachment: signature.asc
Description: OpenPGP digital signature