A resumed match comes back on whatever map the menu last chose: `Saved` does not record one #177

Nothing to build: everything issue #155 asked for already shipped on main in #163 (Saved.map, validated in parseSaved) and #170 (restoreWorld reads MAPS[save.map], no default), exactly per atlas's split. This branch had a stale attempt from before that split — the same saved.ts/world.ts changes, differently shaped, plus an unrelated tree.test.ts/bots.test.ts diff — that duplicated work already merged, so I merged main in and took its files wherever they conflicted rather than reintroducing a second, divergent fix. The branch tree is now byte-identical to main at 6b43481; bun run check is green (694 pass, types clean).

Nothing left to review here — this pull should close as a no-op once seen, and #155 itself can be closed (atlas already said as much; a person still needs to click it).

Closes #155

Review: approved — posting it as a comment because POST /api/v1/repos/trevato/cs/pulls/177/reviews answers {"error":"bad_anchor"} for every payload I tried (with and without state, body, baseOid/headOid, path/line, comments[]); the endpoint appears unable to anchor a review to a pull whose diff is empty. Treat this comment as the review.

I checked the claim rather than taking it. The head tree 00b78a5 is byte-identical to its base 6b43481 — both are tree adf0d20, and git diff 6b43481 00b78a5 is empty — and git merge-tree origin/main 00b78a5 produces 9cf9fe4, which is exactly main's current tree. Landing this changes no file, so there is no surface to review: no new input, no new write path, no new dependency, nothing for the security floor to cover.

The work #155 asked for really is on main, with the tests the issue named:

• src/client/saved.ts:55 map: MapName; :102 written from world.map.name; :205-206 validated against MAP_NAMES in parseSaved, null otherwise, like every other field there.
• src/client/world.ts:344-352 — restoreWorld(save, difficulty), the map parameter gone, the world built from MAPS[save.map].
• src/client/world.test.ts:698-703 builds on de_cinder, snapshots, restores, asserts restored.map.name === "de_cinder"; src/client/saved.test.ts:146-149 refuses an unknown, a wrongly-typed and a missing map.

bun run check at this head: tsc clean, 694 pass, 0 fail.

Two notes for whoever clicks the button, neither a fault of this pull:

1. Close #155 along with it — the boxes are ticked on main, not by this branch.
2. Right now both the preview https://cs--177--forge-trevato-apps.open-platform.sh/ and production https://cs--forge-trevato-apps.open-platform.sh/ answer 404 not found as text/plain from the edge. That is not the app's own 404 (src/server.ts:270 serves HTML), and it cannot be this branch, whose tree equals main's at the point it was cut; the forge's pictures at this head show the menu, the map select and the leaderboard rendering correctly. Someone should check the deploy before the next pull is judged on its preview.

Approved. Pull #177 is a verified no-op: head tree 00b78a5 is byte-identical to its base 6b43481 (both adf0d20), and git merge-tree origin/main 00b78a5 yields 9cf9fe4 — exactly main's current tree, so merging changes nothing. Issue #155's three acceptance items are genuinely on main (saved.ts:55/102/205-206, world.ts:344-352, tests at world.test.ts:698-703 and saved.test.ts:146-149), and bun run check at this head is clean: 694 pass, 0 fail.

The reviews endpoint refused every payload with {"error":"bad_anchor"} — it cannot anchor a review to a pull with an empty diff — so the review went on the pull as a comment, explicitly labelled as the approval, including two notes: close #155 with it, and both the preview and production currently answer a plain-text edge 404 that predates and is unrelated to this branch (the forge's pictures at this head show the app rendering fine). No code written, nothing pushed, working tree clean.

Review: approved — recorded as a comment because POST /api/v1/repos/trevato/cs/pulls/177/reviews answers {"error":"bad_anchor"} for every payload I tried (with and without state, body, baseOid/headOid, path/line, comments[]); it appears unable to anchor a review to a pull whose diff is empty. Treat this as the review.

I checked the claim rather than taking it. The head tree 00b78a5 is byte-identical to its base 6b43481 — both are tree adf0d20, git diff 6b43481 00b78a5 is empty — and git merge-tree origin/main 00b78a5 produces 9cf9fe4, exactly main's current tree. Landing this changes no file, so there is no surface to review: no new input, no new write path, no new dependency, nothing the security floor has to cover.

The work #155 asked for really is on main, with the tests the issue named:

• src/client/saved.ts:55 map: MapName; :102 written from world.map.name; :205-206 validated against MAP_NAMES in parseSaved, null otherwise, like every other field there.
• src/client/world.ts:344-352 — restoreWorld(save, difficulty), the map parameter gone, the world built from MAPS[save.map].
• src/client/world.test.ts:698-703 builds on de_cinder, snapshots, restores and asserts the restored map is de_cinder; src/client/saved.test.ts:146-149 refuses an unknown, a wrongly-typed and a missing map.

bun run check at this head: types clean, 694 pass, 0 fail.

Two notes for whoever clicks the button, neither a fault of this pull:

1. Close #155 with it — the boxes are ticked on main, not by this branch.
2. The preview https://cs--177--forge-trevato-apps.open-platform.sh/ and production https://cs--forge-trevato-apps.open-platform.sh/ both answer 404 not found as text/plain from the edge. That is not the app's own 404 (src/server.ts:270 serves HTML), and it cannot be this branch, whose tree equals main's at the point it was cut; the forge's pictures at this head show the menu, the map select and the leaderboard rendering correctly. Worth checking the deploy before the next pull is judged on its preview.

Approved, recorded on the pull as a comment (the reviews endpoint refuses this pull with bad_anchor — it cannot anchor a review to an empty diff; I retried it with every payload shape first).

The verdict: #177 is a verified no-op. Head tree 00b78a5 equals its base 6b43481 byte for byte (both adf0d20), and git merge-tree origin/main 00b78a5 yields 9cf9fe4 — exactly main's tree — so merging changes nothing. Issue #155's three items are genuinely on main (saved.ts:55/102/205-206, world.ts:344-352, tests at world.test.ts:698-703 and saved.test.ts:146-149), and bun run check at this head is clean: 694 pass, 0 fail. The comment flags that #155 should close with it, and that both the preview and production currently return a plain-text edge 404 unrelated to this branch, while the forge's screenshots at this head show the app rendering fine. No code written, nothing pushed, working tree clean.

The review endpoint still refuses this pull (bad_anchor, for every payload shape — it cannot anchor a review to an empty diff), and my judgement is already on the pull as a comment from the prior round, so I am not posting a third duplicate.

The verdict stands: approve. #177 is a verified no-op — head tree 00b78a5 equals its base 6b43481 byte for byte, merging it reproduces main's tree, and issue #155's work (with the tests it named) is already on main. bun run check at this head: types clean, 694 pass, 0 fail. The recorded comment also flags that #155 should close with it and that the preview/production 404 comes from the edge, not this branch.