Page MenuHomePhabricator

Provide a way to query for commits with not "Done" (or at least unseen) comments
Closed, DuplicatePublic

Description

Right now if you have a backlog of commits going through post-review audits it is a little hard to keep track of things that still need to be done. One particular area of difficulty is tracking comments that people have made without raising a concern.
Since raising a concern is a pretty heavy-weight action (it requires an auditor to remove the raised concern at a later time somehow) it is most often desirable to just leave some comments and suggestions and then accept the Audit. The problem is that once an audit has been accepted it is pretty much impossible for the author to keep track of the comments.
We currently work around this by not accepting audits until the author had a chance to look at and respond to comments, but that's not ideal either because it still puts extra work on the auditor to close the audit at a later time and creates an uncertain state where it is not clear whether the audit is actually done.

It seems a decent way to solve this problem is to allow the author to query for commits that have comments on them that have not been marked as Done. For extra points, it would be nice to have a one-click way of marking all comments as done.

Event Timeline

This doesn't sound like a desirable solution to the problem. If something requires addressing, raising a concern is always the correct action. If it's not an urgent issue, but the author wants to address it, they can flag the audit for themselves as a reminder to follow up. Is there something about the core problem maybe I'm missing?

That almost works. The problem is that as soon as the audit is accepted it fall off of everyone's radar and there is no visibility into the comments an auditor may have made.

That's my expectation if something is accepted. That the author is trustworthy enough to follow up if needed or ignore if it was just a suggestion.

I was picturing this as more of a tool to help the author keep track of things that may require attention rather than an auditor to verify that something is done.
This also allows for a workflow where a comment is made after the audit is closed that is not worthy of reopening the audit (and potentially by a 3rd party). And it could be both "I'm late to the party and I have input" kind of scenario and "I'm an intern and I have a question about this change 3 years later"

I was picturing this as more of a tool to help the author keep track of things that may require attention rather than an auditor to verify that something is done.

This is our expected use of Flags. They already exist on most objects, and users can build/query/sort them as they like for whatever reason.

This also allows for a workflow where a comment is made after the audit is closed that is not worthy of reopening the audit (and potentially by a 3rd party). And it could be both "I'm late to the party and I have input" kind of scenario and "I'm an intern and I have a question about this change 3 years later"

The concern around adding additional options to queries is they are fundamentally undiscoverable and unlikely to be used. The scenarios you outline still require someone to have set up and monitor the queries to begin with, and most (all?) users aren't likely to do that.

Some relevant discussion in T1112, although it's more Differential-oriented than Audit-oriented.
T5722 is also probably related.

This is our expected use of Flags. They already exist on most objects, and users can build/query/sort them as they like for whatever reason.

Yes, that totally makes sense to me (and, in fact, that's what I suggested by teammates do), but even an ability to use Herald to automatically flag an audit based on comment activity would go a long way here.

I think this may have been worded to explicitly, but the real objective of this task is to whine about how it's hard to find your action items if you've forgotten to flag them yourself and no-one is yelling at you about them.