Reviewer Guide


Our program follows the usual OpenStack review process, albeit with some important additions (see below). See also: Your first review.

Be Professional

The PTL, with the support of the core reviewers, is ultimately responsible for holding contributors accountable for creating a positive, constructive, and productive culture. Inappropriate behavior will not be tolerated. (Why this is important?)

Do This:

  • Act professionally.

  • Treat others as friends and family.

  • Seek first to understand.

  • Be honest, transparent, and constructive.

  • Use clear, concise language.

  • Use prefixes to clarify the tone and intent of your comments.

Don’t Do This:

  • Use indecent, profane, or degrading language of any kind.

  • Hold a patch hostage for an ulterior motive, political or otherwise.

  • Abuse the review system to discuss big issues that would be better hashed out on the mailing list, in IRC, or during OpenStack Summit design sessions.

  • Engage in bullying behaviors, including but not limited to:

    • Belittling others’ opinions

    • Persistent teasing or sarcasm

    • Insulting, threatening, or yelling at someone

    • Accusing someone of being incompetent

    • Setting someone up to fail

    • Humiliating someone

    • Isolating someone from others

    • Withholding information to gain an advantage

    • Falsely accusing someone of errors

    • Sabotaging someone’s work

Reviewing Docs

When possible, enlist the help of a professional technical writer to help review each doc patch. All reviewers should familiarize themselves with OpenStack Documentation Contributor Guide. When reviewing user guide patches, please run them through Maven and proof the resulting docs before giving your +1 or +2.

Reviewing Code

When reviewing code patches, use your best judgment and seek to provide constructive feedback to the author. Compliment them on things they have done well, and highlight possible improvements. Also, dedicate as much time as necessary in order to provide a careful analysis of the code. Don’t assume that someone else will catch any issues you yourself miss; in other words, pretend you are the only person reviewing a given patch. Remember, “given enough eyeballs, all bugs are shallow” ceases to be true the moment individual reviewers become complacent.

Some things to check when reviewing code:

  • Patch aligns with project goals, and is ideally associated with a bp or bug.

  • Commit message is formatted appropriately and contains external references as needed.

  • Coding style matches guidelines given in HACKING.rst.

  • Patch is cohesive and not too big to be reviewed in a timely manner (some patches may need to be split to improve cohesion and/or reduce size).

  • Patch does what the commit message promises.

  • Algorithms are implemented correctly, and chosen appropriately.

  • Data schemas follow best practices.

  • Unit and functional tests have been included and/or updated.

  • Code contains no bugs (pay special attention to edge cases that tests may have missed).

Use Prefixes

We encourage the use of prefixes to clarify the tone and intent of your review comments. This is one way we try to mitigate misunderstandings that can lead to bad designs, bad code, and bad blood.



What the reviewer is saying



You did a nice job here, and I wanted to point that out. Keep up the good work!



I think you are missing a test for this feature, code branch, specific data input, etc.



I don’t think this code does what it was intended to do, or I think there is a general design flaw here that we need to discuss.



This is a serious security vulnerability and we better address it before merging the code.



I have a concern that this won’t be fast enough or won’t scale. Let’s discuss the issue and benchmark alternatives.



I think there is something critical here that we need to discuss this in IRC or on the mailing list before moving forward.



This doesn’t seem to be consistent with other code and with HACKING.rst



I don’t understand something. Can you clarify?



This could be modified to reduce duplication of code, data, etc. See also: Wikipedia: Don’t repeat yourself



This feature or flexibility probably isn’t needed, or isn’t worth the added complexity; if it is, we can always add the feature later. See also: Wikipedia: You aren’t gonna need it



This is a nitpick that I can live with if we want to merge without addressing it.



I’m chiming in with my opinion in response to someone else’s comment, or I just wanted to share an observation. Please take what I say with a grain of salt.



I just wanted to share some useful information.