|
|
|
|
|
by barrkel
1739 days ago
|
|
FWIW, that series of small separate requests, reviewed separately, is pretty much how code review works in Google. Each change is attached to the same tracking issue which is what tracks completion, rather than any single or group change being committed. Yes, alone they don't provide actual value, but they are small (easy to review) and can be trivially / automatically rolled back (database migrations aside) and can be deployed (without being used) safely, and even when they are exercised, it may be via a switch, itself toggled with a pull request which itself can be rolled back in isolation. Whatever you think is a small change, think even smaller. Think creating a class and stubbing out many of the methods with TODOs. Delaying the merge until all commits required for the feature wouldn't work in a large monorepo - just keeping small commits up to date can be work when there's enough people working around the clock. There's no feasible way to sustain continuously rebasing in a feature branch model in a large monorepo. |
|