Re:Introducing review requests
by piotrkaminski on Wednesday December 07, 2016 @07:20PM
(#13127451)
FWIW https://reviewable.io supports arbitrarily complex review completion requirements, and the outcome gets fed into a PR status that you can make required as well. Right now the completion condition must be self-contained but I'm considering passing in designated metadata files so you could have, e.g., per-directory "owners" like Google does.
(Disclosure: I'm the founder.)
I second this. I am an xoogler and we had a file inside of folders that dictated who you needed to get approval by. When someone edited code the review tool would crawl the tree to find the nearest owners file. We couldn't merge unless someone from that file approved the change request. Worked great!
Now that we have this tool I wonder if there is an API to add a reviewer. It would be fairly trivial to add this if there was an API.
Re:Introducing review requests
by tomtheengineer on Wednesday December 07, 2016 @06:25PM
(#13127157)
+1. Gerrit has a similar way to express access permissions that differentiate between a "contributor" and a "developer": https://gerrit-review.googlesource.com/Documentation/access-...
Yes, this is important. The Kubernetes community is having to build bots that enforce this sort of workflow.
Right now the greenest guy on the team can approve a PR and it can be merged.
The real dream would be to be able to set the requirements on a per-directory basis, but I doubt that would ever happen.