You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
One important piece of state that bors should have up-to-date is the queue of (open) PRs. However, when a PR is closed/merged/converted to draft while the bors is offline or being redeployed, or a PR status change webhook is lost, the state in the bors DB can become outdated.
We do update the PR state after every comment on that PR that contains a known bors command, but it would be nice to also update the PRs proactively.
When the bot starts, or in the periodic refresh handler, we could update all opened PRs in our DB, and check if they are still opened. However, doing this one-by-one would be very slow and create many GitHub requests. What we could do instead is batch load PRs with a specific state (opened) from GH, which allows loading 100 PRs at once, and also batch load opened PRs from our DB. Then, when we find a PR where the opened state on GH does not match the state in DB, we update the DB state. See here for an example of how to load that data.
The text was updated successfully, but these errors were encountered:
One important piece of state that bors should have up-to-date is the queue of (open) PRs. However, when a PR is closed/merged/converted to draft while the bors is offline or being redeployed, or a PR status change webhook is lost, the state in the bors DB can become outdated.
We do update the PR state after every comment on that PR that contains a known bors command, but it would be nice to also update the PRs proactively.
When the bot starts, or in the periodic refresh handler, we could update all opened PRs in our DB, and check if they are still opened. However, doing this one-by-one would be very slow and create many GitHub requests. What we could do instead is batch load PRs with a specific state (opened) from GH, which allows loading 100 PRs at once, and also batch load opened PRs from our DB. Then, when we find a PR where the opened state on GH does not match the state in DB, we update the DB state. See here for an example of how to load that data.
The text was updated successfully, but these errors were encountered: