And one engineer can (or could in 2018) put the unix name of the person sitting next to them in the "Reviewed By:" field and force push something, but it isn't/wasn't The Done Thing outside of the most extreme 3am SEV-0 scenarios.
Any two people are in a less-extreme but similar boat: if someone puts up a bad diff and someone without a stake in the code accepts it, they had better hope to hell nothing goes wrong.
SEV review is a remarkably enlightened process for what it is, but you do not want to sit there explaining to extreme-seniority people why you YOLO'd something into Presto without buy in from a Presto hacker with your manager sitting behind you already thinking about how much this is going to get harped on in calibration.
The `OWNERS` file at FB is in `hg log`, but it's there.
If I remember right, Google also had a mechanism for pushing something through outside of normal review. But it was something more disciplined than just faking a review by putting your co-workers name in a reviewed-by field.
Because it was more disciplined, those commits could be automatically marked for later review.
In the bad/fun old days of extremely small on-call loop, a less clear/organized escalation chain, and fairly "roll your own" per-team tooling, once in a blue moon a whole business unit would be down because 10k machines were crash-looping and you just needed `if (NULL == ptr) return;` on like one line to end the carnage. There are exceptions to every rule I suppose.
But in 7 years at FB I think I force slammed someone else as reviewer with merely their verbal approval in the middle of the night 2 or 3 times, and never after like, 2015.
For a long time, it was worse than that - once a change was "accepted", you could amend the commit with changes to literally any other file in the entire monorepo, and land it without any further review.
They finally added a "final review" step to ensure that someone eventually takes a look at these changes after the initial accept, but that still occurs several days after the commit.
Comments
Facebook/Meta does not have owners, so any two people in the company can land a change.
And one engineer can (or could in 2018) put the unix name of the person sitting next to them in the "Reviewed By:" field and force push something, but it isn't/wasn't The Done Thing outside of the most extreme 3am SEV-0 scenarios.
Any two people are in a less-extreme but similar boat: if someone puts up a bad diff and someone without a stake in the code accepts it, they had better hope to hell nothing goes wrong.
SEV review is a remarkably enlightened process for what it is, but you do not want to sit there explaining to extreme-seniority people why you YOLO'd something into Presto without buy in from a Presto hacker with your manager sitting behind you already thinking about how much this is going to get harped on in calibration.
The `OWNERS` file at FB is in `hg log`, but it's there.
If I remember right, Google also had a mechanism for pushing something through outside of normal review. But it was something more disciplined than just faking a review by putting your co-workers name in a reviewed-by field.
Because it was more disciplined, those commits could be automatically marked for later review.
I believe you cannot force push anymore.
Agree. Wasn’t aware of force push (2021-2022).
That is probably for the best.
In the bad/fun old days of extremely small on-call loop, a less clear/organized escalation chain, and fairly "roll your own" per-team tooling, once in a blue moon a whole business unit would be down because 10k machines were crash-looping and you just needed `if (NULL == ptr) return;` on like one line to end the carnage. There are exceptions to every rule I suppose.
But in 7 years at FB I think I force slammed someone else as reviewer with merely their verbal approval in the middle of the night 2 or 3 times, and never after like, 2015.
One can imagine that auditors were not particularly enthused about this :)
Oh anything on user data or money or illegal content shit or anything like that is a whole different ballgame.
They mail you the shit off your desk for playing games with any of that stuff, and it’s thoroughly, thoroughly checked.
For a long time, it was worse than that - once a change was "accepted", you could amend the commit with changes to literally any other file in the entire monorepo, and land it without any further review.
They finally added a "final review" step to ensure that someone eventually takes a look at these changes after the initial accept, but that still occurs several days after the commit.
Can you still add
Reviewed-by: yourself
in the commit message to skip review? That was my favorite one.