Tails

#2 Updated by intrigeri 2015-07-11 09:06:41

> Note that changing the ticket state is easy, but assigning a different owner might be a bit more difficult. All that Jenkins knows about is a commit author email, not a Redmine username, so maybe the email solution won’t be able to cover every bits of the specification.

Indeed, good catch.

The sanest and easiest way I see to handle this is to assume that whatever is on the left of ‘@’ in the email address is the Redmine account name (it works for most of our usual committers); and for exceptions, maintain a file somewhere that maps committers email address to Redmine accounts name. Notes about this idea:

• What to do when reassigning the ticket fails because we guessed the Redmine user account wrongly (and there’s no corresponding entry in the override mapping file)? With the email solution, Jenkins won’t even know it has failed. And actually, nobody will, since Redmine doesn’t tell you when some email command fails. So the ticket will stay on the reviewer’s plate, which is fine: at this stage of the branch dev/review process, the reviewer is the one who is looking at the ticket, so they’re the best placed person to notice that it failed to be reassigned, and then 1. do it themselves; 2. notify us so that we add an entry in the override mapping file. Bonus: the branch’s developer will be notified of the build failure, so maybe they can make sure that the ticket is reassigned to them (unlikely it works, but who knows). Still, this leaves one corner case open: when the ticket is Ready for QA, but no assignee is set (which is our default, at least in theory). In that case we have to rely on the current RM regularly looking at unassigned Ready for QA tickets (which is part of their duty), and/or on the branch’s developer to notice that the build failed, and reassign the ticket to themselves. All that is kinda suboptimal, but let’s try without any special handling first, and readjust if it doesn’t work, OK? A simple improvement could be to Cc the last committer when emailing Redmine, which should be trivial and possibly useful; and also, perhaps Cc the current assignee (reviewer), but I expect it will be hard to find out their email address, so let’s forget that one.
• Perhaps some heuristics that use the committer’s name (as opposed to their email address) could optimize this a bit, and make the fallback to the mapping file less often needed, but IMO that’s premature optimization.

Other (crazy) ideas I’ve discarded:

• query Redmine for user account names matching a given email address; I doubt that it’s possible without writing (and maintaining) a custom Redmine plugin
• set up a LDAP directory with all needed info about project members, such as email address, Redmine account name, etc.: totally overkill in the current state of our needs

> Let’s try to setup the first one to see if it’s that easy, and if not so useable we can still fallback on the API implementation.

Full ACK!

#5 Updated by intrigeri 2015-08-07 01:28:05

As written over email:

It seems that this ticket strongly relies on our branches name containing the ticket number. That’s merely a “SHOULD” in our specs, but it seems quite easy and useful, and you flagged it for 1.5, so I’m assuming you want to take care of it soon anyway.

So, should we:

• make it clear(er) in our contributors doc that branches must be named this way?
• rename existing active branches that don’t satisfy this requirement?

Another idea: do we want to enforce this? It could be painful sometimes, but perhaps that’s still the easiest way to get a consistent dev. experience?

#7 Updated by intrigeri 2015-09-01 02:08:44

Keep it mind that this one is merely a should => I would postpone it a bit further.

#8 Updated by bertagaz 2015-09-01 03:24:12

Yes, I’ll see during the 1.7 release cycle.