(dropping the gnupg bug again)
Paul Gevers writes ("Re: Bug#902316: gnupg failing completely in dgit
autopkgtests"):
> Adding Martin and Antonio into the loop, as I don't know the details of
> autopkgtest so well. Question is near the bottom.
Thanks for your attention, everyone.
> PS: for you all to know, I triggered a test with git from experimental
> (already before this bug report). It is possible now with the new debci
> API. (triggers with spaces are not properly reported on the results
> page, I already knew that but I'll file a bug about it nevertheless). I
> still intend to automate this in the future, but if one wants...
>
> elbrus@respighi:~$ cat tmp.json
> [{"package": "dgit", "trigger": "try git from experimental",
> "pin-packages":[["src:git", "experimental"]]}]
> elbrus@respighi:~$ curl --cacert /etc/ssl/ca-global/ca-certificates.crt
> --header "Auth-Key: $(cat debci.key)" --form tests=@tmp.json
> "https://ci.debian.net/api/v1/test/unstable/amd64";
Would it be a sensible idea for me to upload stunt versions of dgit to
experimental with different attempts at workarounds for the gnupg bug,
and retrigger the test using some rune like that ?
Martin Pitt writes ("Re: Bug#902316: gnupg failing completely in dgit
autopkgtests"):
> Paul Gevers [2018-06-25 9:13 +0200]:
> > If I recall a discussion well, than the testbed is only shared if the
> > test dependencies are the same or a super set of previous run and
> > "breaks-testbed" is not set. Martin, do you confirm?
>
> Confirmed. This behaviour hasn't changed since day one, i. e. it's still how
> Ian originally wrote it. This makes it possible for multiple tests to share a
> testbed if you order them carefully, which cuts down the test time on runners
> where test bed set up is expensive (like on OpenStack).
I will look at the test suite metadata and see if that provides a
hint.
Paul, I guess there is no way to see per-test autopkgtest results ?
Ian.
