Re: QA page improvements (Was: Your extremely welcome contributions to QA page)
Hi Sunil,
On Wed, Oct 28, 2020 at 10:01:38PM -0700, Sunil Mohan Adapa wrote:
> On 26/10/20 2:04 pm, Andreas Tille wrote:
>
> I have submitted merge requests for better i18n in Genshi templates and
> a different styling for the QA page.
Very cool! I really like the new layout with the smileys as marker.
The only thing is that my localhost has problems with the css/js, hmmm.
> I might work on score based sorting sometime later.
Sounds good.
> > which also contains "testing migration" information. It should not be to
> > hard to code this yourself but I learned that there is a pitfall for every
> > query. So re-using existing code is sensible.
>
> I understood the logic that DMD uses it and implemented in the QA page.
> I submitted a merge request for it too.
That code makes perfectly sense. Could you please add migration_result
to the general stats of the entry page?
> > Sound sensible - but I wonder whether there are cases where some work
> > from packagers side is left.
>
> So far, from the packages I have examined, the 'Erroneous package' is
> the message that is thrown when setting up of the test harness fails.
> This include situations when package dependencies not found,
> dependencies could not be installed, or package itself could not be
> installed.
>
> When tests start running, they pass, fail, or are skipped. Various
> messages look like this:
>
> udd=> select message, count(*) from ci group by message;
> message | count
> --------------------------------------------------------+-------
> | 105
> Tests failed, and at least one test skipped | 82
> Erroneous package | 2181
> Unexpected autopkgtest exit code 20 | 3
> No tests in this package or all skipped | 6344
> Tests passed, but at least one test skipped | 823
> elbrus | 1365
> Tests failed | 3140
> Could not run tests due to a temporary testbed failure | 178
> All tests passed | 43060
> No tests in this package | 20
>
> In case the failure is because of 'Erroneous package' there does not
> seem to additional information that could help us with when it can be
> ignored. I am not sure if we can safely ignore 'Erroneous package'
> message entirely.
I came also to the conclusion that its not save to ignore 'Erroneous
package'. May be we should really talk to CI team whether they might be
able to set a flag for more details. Those package I inspected are
passing for amd64 but failing on other architectures where some depends
do not exist. While we could query this case from UDD it would ba a
complex and potentially error prone way to fetch that information. IMHO
it would be more sensible if CI would check this in advance and would
not run the test at all if it is clearly bound to fail.
Thanks a lot for your dedication to enhance the Blends websentinel
Andreas.
--
http://fam-tille.de
Reply to: