On 10 July 2017 at 17:20, CARVER, PAUL <[email protected]> wrote:

> I see fairly often that someone posts a “recheck no bug” comment to
> Gerrit. I assume this means that the CI failed for reasons that are
> unrelated to the specific commit and the CI needs to be re-run for that
> commit. Is there an underlying problem that is known? Are there obstacles
> to getting the root problem fixed?
>

Just a passing thought:

Supposing that you want to know whether a branch _ever_ passes, and your CI
has a spurious random failure (say) 10% of the time. The problem is that
some human will have to investigate a false-negative on 1/10 tests due to
that spurious failure. But if you would make the CI run the test (say)
three times, and report whether _any_ of them succeeded, then your
false-positive rate would drop to 1/1000 and maybe not bother anybody
anymore.

Could conceivably be a shell one-liner in e.g. Jenkins config like:

for run in 1 2 3; do ./test && exit 0; done

Could throw in a sleep or something if the errors are correlated by time,
etc.

Just a thought :-)

Cheers,
-Luke
_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org

Reply via email to