Page MenuHomePhabricator

Gerritbot should not get subscribed to tasks
Closed, DeclinedPublic

Description

Event Timeline

Ricordisamoa updated the task description. (Show Details)
Ricordisamoa raised the priority of this task from to Needs Triage.
Ricordisamoa added a project: Phabricator.
Ricordisamoa changed Security from none to None.
Ricordisamoa added a subscriber: Ricordisamoa.
Qgil triaged this task as Lowest priority.Dec 13 2014, 3:16 PM
Qgil added a subscriber: Qgil.

I don't see an easy way to avoid this. What harm does it cause?

I think the main confusion is caused because it now looks as if @Ricordisamoa added gerritbot on purpose. The easiest solution is maybe to let gerritbot subscribe when it reports a new patchset.

because it now looks as if Ricordisamoa added gerritbot on purpose

Same for any task created that's not in a project I'm member of - I still get CC'ed on it because I've set up Herald to do so. It's just more exposed in Phab (in Bugzilla I did that via the "globalwatchers" setting but that did not get shown in a Bugzilla ticket) and I don't see a real problem with that.

because it now looks as if Ricordisamoa added gerritbot on purpose

No, because it looks like "gerritbot" is a real person, and CC statistics could be messed up.

Which "CC statistics"?

Qgil added a comment.Dec 15 2014, 11:01 AM

For what is worth, in http://korma.wmflabs.org we put an effort filtering bots from the statistics. Therefore, if we even have CC statistics, we wll try to filter bots regardless.

I checked Herald. I see no way to create a rule like "If gerritbot is added to CC, remove gerritbot".

Remembering T76993, I wonder whether we should think of a request upstream about not subscribing users automatically when they are being "just" quoted.

Qgil moved this task from To Triage to Need discussion on the Phabricator board.Dec 15 2014, 11:01 AM

@Qgil have you checked Herald recently?

Restricted Application added a subscriber: scfc. · View Herald TranscriptJun 21 2015, 12:14 PM

Herald would now allow a global rule "When: CCs include any of: gerritbot; Take these actions every time this rule maches: Remove emails from CC: gerritbot".

Still I don't understand how the current behavior creates an actual problem.

Herald would now allow a global rule "When: CCs include any of: gerritbot; Take these actions every time this rule maches: Remove emails from CC: gerritbot".

Still I don't understand how the current behavior creates an actual problem.

It prevents from viewing real subscribers at a glance.
Also Forrestbot, please.

Ricordisamoa renamed this task from gerritbot should not get subscribed to tasks when quoted to Bots should not get subscribed to tasks when quoted.Jun 27 2015, 9:04 PM
Ricordisamoa renamed this task from Bots should not get subscribed to tasks when quoted to Bots should not get subscribed to tasks.

It prevents from viewing real subscribers at a glance. Also Forrestbot, please.

In which specific situations does it actually matter if one or two of the subscribers are not human when looking at subscribers at a glance?

It prevents from viewing real subscribers at a glance. Also Forrestbot, please.

Hmm, in which specific situations does it matter if one or two of the subscribers are not human when looking at subscribers at a glance?

It prevents from viewing real subscribers at a glance. Also Forrestbot, please.

In which specific situations does it actually matter if one or two of the subscribers are not human when looking at subscribers at a glance?

One would expect that they need to be subscribed in order to perform a task. They don't.

forrestbot no longer subscribes itself: T103656#1408554.

forrestbot no longer subscribes itself: T103656#1408554.

Only because it does not comment anymore (T100945), but gerritbot needs to.

It prevents from viewing real subscribers at a glance. Also Forrestbot, please.

In which specific situations does it actually matter if one or two of the subscribers are not human when looking at subscribers at a glance?

One would expect that they need to be subscribed in order to perform a task. They don't.

Okay, but how does it actually change anything whether one would expect that or not? And what is "to perform a task" here?

One would expect that they need to be subscribed in order to perform a task. They don't.

Okay, but how does it actually change anything whether one would expect that or not?

It's misleading at best.

And what is "to perform a task" here?

For example, listening for changes to tickets the bot is subscribed to and building a live graph.

scfc added a comment.Sep 4 2015, 9:06 AM

If the purpose is to "listen for changes to tickets the bot is subscribed and build a live graph", then that is much better served by asking the bot operators for an API that exposes the tickets a bot did something to as part of its duties. Extracting that information from the Phabricator CC lists is bound to fail.

Just create a Herald rule then.

Pong. So far I just don't think that the time spent for implementation would be justified, given that I personally don't see a big problem created by the current behavior.

I can find at least 42 occurrences of 9 different users (not including myself) removing gerritbot from subscribers: mostly Krinkle, matej_suchanek and matmarex.

One minute of your time < 42 * 3 seconds of others' time

See my previous comment

greg renamed this task from Bots should not get subscribed to tasks to Gerritbot should not get subscribed to tasks.Mar 27 2016, 1:11 AM

@ReleaseTaggerBot and @Stashbot already don't subscribe themselves. Please just fix @gerritbot to behave the same.

stashbot subscribes itself, see T96848#2156077 for example.

Just ignore subscribed bots. No need to remove them because they don't receive email anyway.

Qgil removed a subscriber: Qgil.Mar 30 2016, 9:41 AM

If using maniphest.update it may be sufficient to pass the original ccPHIDs? Or maybe it is better to use maniphest.edit with comment and projects.add.

Restricted Application added a subscriber: TerraCodes. · View Herald TranscriptMay 19 2016, 3:00 PM
Aklapper moved this task from Need discussion to Misc on the Phabricator board.Oct 7 2016, 4:44 PM

I asked for a description of the root problem in T78468#1409276 and that is still missing after two years.
The only arguments have been "it's misleading" in T78468#1413128 and vague comments about "building live graphs" and some "CC statistics" (where to see those two?).

Proposing to decline this task as it's still unclear which actual problems the current behavior creates.

Aklapper closed this task as Declined.Nov 18 2016, 12:48 PM

I asked for a description of the root problem in T78468#1409276 and that is still missing after two years.
The only arguments have been "it's misleading" in T78468#1413128 and vague comments about "building live graphs" and some "CC statistics" (where to see those two?).

Proposing to decline this task as it's still unclear which actual problems the current behavior creates.

And declining due to lack of a response. :-/