unpick the 'pass cid to event.create' workaround, comment
So now we can wait for composes to appear in PDC before doing
event creation, we don't need this workaround any more. It's a
mess, so ditch it. Also comment the situation. To expand on the
comments a bit:
Right at the moment a 'candidate' compose completes, it is only
available in a location derivable from its compose ID (not from
its label). The relevant wikitcms event's version properties,
however, are derived from the compose's label - when we actually
decide what the event should be, we get the compose's label and
say "OK, Alpha-1.5? Then the event is milestone Alpha, compose
1.5!"
When we're creating the event, we then wind up working *back*
from the event's version properties to try and get a fedfind
release instance, in order to generate the download table. We
call fedfind.release.get_release(release, milestone, compose).
For the brief period (likely a few seconds to a few minutes)
where the compose is not in PDC, fedfind will fail to find any
release for us, because there is just no way to figure out the
compose ID from the milestone and release at that point. Once
the compose is in PDC, fedfind can parse the milestone and
compose back into a label, ask PDC for the compose ID of the
compose with that label, and then find the compose based on the
compose ID.
The other way fedfind can succeed when given a label-y milestone
and compose is if the compose has been synced to stage, because
when we sync candidate composes to stage the directory name is
based on the label, not the compose ID. So fedfind can find
those just fine without needing to talk to PDC at all. But sync
usually happens some time *after* PDC import, so PDC import is
the earliest point at which we can reliably create an event for
a candidate compose and be sure the download table will work.
An alternative to all this faffing around would be to give up on
inferring the compose for wikitcms events from their version
properties and instead *require* a compose ID to be explicitly
provided at event creation time and have the compose ID for each
event stored in the wiki somehow, but that'd mean setting up some
kind of 'metadata' page in the wiki and reading that every time
we want to find out the compose for a validation event. I dunno
which is worse.