Open Stack

Dan Prince <dprince at redhat.com> writes:

> ----- Original Message -----
>> From: "Michael Still" <mikal at stillhq.com>
>> - commenting "recheck .*"
>> - commenting "recheck migrations"
>
> With the growing interest in 3rd party testing systems would using 'recheck turbo-hipster' make more sense here?
>
> I'm fine with 'recheck migrations' in addition for turbo-hipster but it would make sense to align the recheck naming scheme with the title of the reviewer for the 3rd party testing system.

This is the can of worms I was hoping we would not open.  Or try to get
them all back into the can and close it again is perhaps the better
metaphor.

I do not think that system-specific recheck commands will be actually
that useful or important.  As I mentioned, I understand the theoretical
usefulness of being able to say "oops, I can tell this one system messed
up, let's ask it to try again".  But given that case is covered by
asking all systems to recheck, it seems harmless to say "all systems to
recheck".

I just don't think that asking developers to learn a micro-language of
commands left in gerrit comments is the best use of time.  Maybe I'm
wrong about that, but I was hoping we could try not creating it first to
see if there's really a need.  Dealing with the expected errors of a
system first coming online doesn't, in my mind, demonstrate that need.

If it turns out that asking programmers not to create a new language is
futile, yes, I'd rather we have some predictability.  Most third party
CI systems have descriptive names, so rather than saying "recheck
turbo-hipster", perhaps we should change turbo-hipster's display name to
something related to "migrations".

So _if_ we want to have system-specific recheck commands, then I think
we should ask operators to make them in the form "recheck <systemname>",
and that they output a brief sentence of help text to remind people of
that in the report message in Gerrit.

-Jim