Johannes Schauer wrote:
In particular, in the "CI" column, look for those rows where the first line shows exactly "✔-" -- those packages have to be excluded. Failures are fine and even both missing is fine but a checkmark followed by a minus triggers that bug.
Neutral/missing ("⛔-") is also bad; I don't have any examples of fail/missing ("fail-").
The other way round (missing/pass, e.g. libmessage-passing-perl) is OK.Missing/missing seems to be fine, but I suspect most of those are "doesn't have autopkgtests", and don't know if that matters.
Looking at the raw debci data (e.g. https://ci.debian.net/data/autopkgtest/unstable/amd64/p/), the packages with "missing" debci data do actually have test logs, but relatively old ones. (>~ 20-30 days old, but *not* a strict age threshold, e.g. parsley-clojure vs plastex.)
This suggests the real bug may be in debci ... or that there are two bugs. (Failing to load the entire page isn't a very helpful response to one package having missing data, even if it's a bug somewhere else that the data is missing.)
It also suggests that
By going to https://ci.debian.net/user/ I can schedule new tests for the affected packages for unstable on all architectures.
would need to be repeated about that often.