[perfalert_inactive_regression] Don't needinfo on bugs with a pending needinfo and don't comment more than once

[suhaibmujahid]

Fixes#2575

@gmierz WDYT?

Checklist

• Type annotations added to new functions
• Docs added to functions touched in main classes
• Dry-run produced the expected results
• The to-be-announced tag added if this is worth announcing

[gmierz]

I'm not okay with that change. We're using those comments/needinfos to keep track of bugs that have no activity happening on them. I'm okay with it not doing a needinfo if there's already a needinfo for the author, as long as the comment is still made and we get it in the bugmail. In the next couple months, we're planning on starting to request backouts for patches that have two consecutive requests for updates with no activity. Just gathering data atm about how often it happens.

[marco-c]

We could allow further comments, but we'd need to change the wording to avoid people dismissing the bot as too noisy. For example, we could mention that a backout will happen in case of no activity.
We have to be careful with noise if we want developers to actually care. If we introduce too much noise, people will just start ignoring the bot.

[gmierz]

If we add/change a section to add a mention of a backout happening that works for me - we have been mentioning this when we reach out to devs in these situations. We just can't start requesting the backouts yet because we don't know how often this happens at the moment (seems to be a couple instances a month so far).

[gmierz]

Maybe in the first line of the template for the needinfo, it could be changed to:

It has been over {{ extra["nweeks"] * 7 }} days with no activity on this performance regression. Your patch is now under consideration for being backed out.

[suhaibmujahid]

Waht if in the second time start emailing them instead of adding more comment on the bug. We have a micnaisim in the bot to even escalate to CC the manager if this is necessary.

[gmierz]

That sounds great to me! As long as we keep seeing the bug in the bugmail summaries we get, I'm happy with that.