[AdaWorldAPI]

Summary

Hotfix for the live outage introduced by #123: every /api/osm/geometry/tile-bin/* request returned 400.

Root cause — WriteParams defaults max_rows_per_file to 1M, so Berlin's 2,525,052 rows were written as three Lance fragments. locate_row_column found row 0 at offset 0 of one fragment and reported the full 1.29 GB slab length against a file that only held ~40% of it; open_slab's .min() clamp then silently truncated the slice to the fragment's file size (footer bytes included), which is not a multiple of NODE_ROW_STRIDE — so RowSlab::new rejected every request. lance-graph's own soa_verbatim.rs warns about exactly this: "a split changes what an mmap offset means."

The serving contract stays as directed: hydrate once from S3 → sink into Lance ONCE → serve zero-copy lazily from the .lance bytes on disk (the mmap+offset mechanism soa_verbatim.rs proves). What changes is that every precondition that test asserts is now enforced at boot instead of assumed:

• write_lance: max_rows_per_file = rows (+ raised max_bytes_per_file) — always a single fragment.
• reopen_if_warm: also requires fragment count == 1, so the broken three-fragment Berlin dataset on volume01 reconverts once, then stays warm.
• locate_row_column: sole data file, stride-aligned run start, head anchor, plus a tail anchor (the slab's last row must sit at offset + length − 512, in bounds). A coincidental head match cannot also anchor the tail; a partial fragment fails the bounds check outright.
• open_slab: exact bounds instead of the .min() clamp — any doubt falls back to the raw .soa file rather than serving a truncated slice.
• ensure_lance_local: digest via mmap instead of fs::read — warm boots no longer pay a 1.29 GiB heap allocation just to hash the slab (relevant to the Railway RAM observation).

Test plan

• cargo check -p cockpit-server clean
• osm_lance tests: 10/10 — including new falsifiers, each verified against the failure it kills: multi-fragment split (the production outage), truncated sole fragment, tail-anchor mismatch, unaligned run start
• osm_features tests: 18/18 (run for real this time — the earlier linker-SIGBUS environment issue is resolved)
• Deploy verification: boot log should show reconverting (fragment count mismatch) once, then row column verified verbatim (sole fragment, aligned, head+tail anchors), and tiles should render again

---

Generated by Claude Code

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d09d3407-5fe6-40ef-a2c0-006ba26d922a

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_09eb6b23-90ed-477f-8684-edd6a94f6815)