Beginning on March 24, 2021, committers@ of Chromium are no longer able to circumvent code review and OWNERS approval on CLs. The full Code Review documentation has been updated to reflect this.
Previously, these were circumventable by self-code-review and because the enforcement was done by presubmit, although rarely done by external contributors. Now, Gerrit will disallow both bypasses. Within Google, where these bypasses were more common, Googlers can find Google-specific information in the internal announcements and landing site.
Periodic updates and FAQs will be sent to [email protected] and updated on this page.
Do I need a reviewer to merge CL's to another branch, even though they were already reviewed on main?
Yes, but within 14 days of the original change you can add Rubber Stamper bot ([email protected]) as the reviewer.
Send questions about this document to [email protected]. Googlers can use an internal-specific email alias that was announced, separately.
How will major refactorings be handled? I regularly need to submit 100s of CLs across the trees; getting OWNERS approval from everyone will be too hard.
We have created a process for landing such changes: Chrome Large Scale Changes.
This process allows approved, large refactorings to bypass OWNERS for the
duration, using a special label Owners-Override
. However, these changes will
still need a second human (anyone with committers Code-Review +1
powers) to
vote.
Rubber Stamper applies the Bot-Commit label to conforming CLs, allowing them to bypass code review. It supports various benign files, clean cherry-picks, and clean reverts that should be exempt from code review.
Rubber Stamper never provides OWNERS approval, by design. It's intended to be used by those who have owners in the directory modified or who are sheriffs. If it provided both code review and OWNERS approval, that would be an abuse vector: that would allow anyone who can create a revert or cherry-pick to land it without any other person being involved (e.g. the silent revert of security patches).
Rubber Stamper auto-reviewer (described above) reviews CLs that meet strict criteria. (The list of file types is Google-internal.) For example: directories with no code.
Essentially, if we can programmatically prove that the CL is benign, then we should allow a bot to rubber-stamp it so that Gerrit allows submission. One can imagine that the classes of CLs that fit in this category would grow over time.
Yes, Rubber Stamper adds CR+1 (Browser) to clean cherry picks. Adding the bot as a reviewer to your CL will cause it to scan and approve it. [email protected] is the bot but just typing "Rubber St..." will autocomplete the full address for you. However, it doesn't provide OWNERS approval so, if you are cherry-picking a CL that you don't have OWNERS on, you can get that approval from the Release Program Manager who approved the cherry-pick.
Documentation will require code review.
There has been much discussion on this topic but senior leaders came to a majority conclusion that the quality increase in documentation from requiring code review outweighed any productivity headwinds.
We will revisit this in the future to evaluate how it is working (or not, as the case may be).
We updated the developer documentation stating that CL authors should prioritize OWNERS closer to the leaf nodes and not to use top-level owners because those folks are likely overloaded and the likelihood of a high response latency or the CL getting lost is high. OWNERS recommendations from Gerrit are in-line with this.
Yes. For break-glass scenarios, there are several folks who have the ability to direct push, including others' CLs.
While they are separate, both impact the integrity of Chrome source code and artifacts and have tangible impacts on developer workflows. For example: TBR was used to bypass OWNERS and the rollout of this policy prevented this bypass. In consultation with senior leaders, we decided that rolling both out simultaneously allowed for more streamlined communication and change management for the contributor community.