hiesel: PTAL. I think something weird is going on with emails and unregistered CCs. Too many individual messages are showing up in the uploader's inbox.

Likely a duplicate of https://bugs.chromium.org/p/gerrit/issues/detail?id=5931

Looks like a dupe of 5931 to me, but assigning to hiesel@ to evaluate it for sure.

From what I see, this might be indented behavior (and not a duplicate of 5931).  Bug 5931  is about sending the exact same message out multiple times, while this bug seems different:

When looking at the timestamps of the actions you mentioned, all of those were individual actions (each causing a notification to be sent):

1) Sat Apr 1 00:30:17 2017 +0000: Create change
2) Sat Apr 1 00:30:27 2017 +0000: CC chromium-reviews@
3) Sat Apr 1 00:30:33 2017 +0000: CC John
4) Sat Apr 1 00:30:38 2017 +0000: CC Darin
5) Sat Apr 1 00:30:42 2017 +0000: CC Avi
6) Sat Apr 1 00:36:53 2017 +0000: Thiago presses QC dry run
[...]

> Then when I added Tommi as a reviewer, it sent two emails, one with my comments, and other adding Tommi as a reviewer.

IIRC, adding a person as a reviewer and posting comments are two different API calls in Gerrit. However, PolyGerrit shows a single UI to do both in one step. Wyatt, is PolyGerrit setting notify=none when people also post a comment next to adding a reviewer? If not, this could cause redundant messages.

> From what I see, this might be indented behavior

I see what you mean now, Patrick. This does look like partly intended behavior.

> IIRC, adding a person as a reviewer and posting comments are two different API calls in Gerrit.

They are the same API call (post review), but can result in multiple emails being sent if comments are posted at the same time that reviewers are added or removed. These emails *could* be combined, but it wouldn't be straightforward.

While adding any number of reviewers at the same time that comments and votes are posted, there should only be two (different) emails sent, so it doesn't seem like the source of redundant emails.

> I think the first 3 was unnecessary, but they were done by Gerrit, not me.

Because the timestamps are different, it looks like whatever mechanism is automatically adding the CCs is not adding them together in a batch.

> They are the same API call (post review)

Ah, right, I was thinking of PostReviewer and PostReview and forgot that PostReview can also add new reviewers.

Moved issue gerrit:5956 to now be  issue chromium:710931 .

It doesn't appear that Gerrit is doing anything different than it ever did. What we're seeing is the impact of how chromium's upload script works. It adds reviewers and CCs one-by-one. When there are dozens of CCs this will take several minutes and the change owner will receive dozens of emails.

https://chromium.googlesource.com/chromium/tools/depot_tools/+/a1bab27a2545105f08d5cd7f1d5c9bae643315d2/git_cl.py#3018

https://chromium.googlesource.com/chromium/tools/depot_tools/+/a1bab27a2545105f08d5cd7f1d5c9bae643315d2/gerrit_util.py#673

Gerrit has long supported adding reviewers via git push options. Since July 2016 the Gerrit REST API has also supported adding reviewers in a single batch through the "Set Review" endpoint:

https://gerrit-review.googlesource.com/Documentation/rest-api-changes.html#set-review

I recommend reimplementing gerrit_util.AddReviewers in terms of gerrit_util.SetReview (or at least using the same REST endpoint).