Stamp the sweep tick where the grid rebuild reads live positions - #474
Merged
Conversation
Every other site that caches a position records when it read it: registerUnit, refreshUnit on both its branches, and the query write-back. The grid rebuild did not, and it re-buckets every unit from a fresh getPos. That only skewed a diagnostic while lastSweepTick fed nothing but the sweep's worst-age counter. Now that the query derives an entry's padding from it, a rebuild leaves every entry looking maximally stale while its cache is in fact exact, so queries pad by a whole sweep cycle and read positions they already have, until the sweep works its way around. The rule is that the stamp belongs with the read, not with the link: writing it inside linkIntoCell instead would tie freshness to bucketing and quietly depend on every future caller passing a live position.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Follow-up to #473, which merged while this commit was in flight, so it missed the squash.
rebuildSpatialIndexGridre-buckets every unit from a freshgetPos, but was the one site that cached a live position without recording when it read it. Every other one does:registerUnit,refreshUniton both branches, and the query write-back added in #473.Why it matters now
While
lastSweepTickfed nothing but the sweep's worst-age counter, a missing stamp only skewed a diagnostic. #473 made the query derive an entry's padding from it, so after a rebuild every entry looks maximally stale even though its cache was just made exact. Queries then pad by a whole sweep cycle and re-read positions they already hold, until the sweep works its way around — five ticks at 500 units and 128 per tick.Why not inside linkIntoCell
The stamp belongs with the read, not with the link. Putting it in
linkIntoCellwould tie freshness to bucketing and silently depend on every future caller passing a live position; a caller that re-links a cached position would then stamp it fresh, which is the one direction that returns wrong results rather than slow ones.Checks
grill typecheck: succeededgrill test UnitSpatialIndexTests: 5/5, and the fullgrill testsuite passed on this change before Write live positions back into the spatial index cache during a query #473 mergedStdlibIngameTests, which needs a running map