v1.152.0

[Azgaar]

Description

[netlify]

✅ Deploy Preview for afmg ready!

Name	Link
🔨 Latest commit	644534b
🔍 Latest deploy log	https://app.netlify.com/projects/afmg/deploys/6aa1abe910196d0008ffbaae
😎 Deploy Preview	https://deploy-preview-1821--afmg.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[barrulus]

Ran the branch locally (tsc / 897 vitest / biome clean; headless boot, legacy prefs adoption,
all three fixtures load → zoom/pan → SVG export → save → reload). The #1666 follow-ups
(string burg-group filters, regenerate prompt after load) are fixed and verified. Notes migrate
correctly on the fixtures and round-trip; culling and full-map export behave.

CI is red: Biome latest flags organizeImports in the three new renderer tests, and
layers.spec.ts › burgs/icons layer needs its #icons snapshot refreshed for the viewport
renderer.

Found, all reproduced in a browser unless marked:

1. Notes lost for maps saved by a 1.152.0 build before 6390273 — package.json was
   1.152.0 from the first commit, so a save from the earlier previews has version 1.152.0 and a
   full data[4]; the migration gates on isOlderThan("1.152.0") and load no longer reads the
   field. Silent, no csv offer. Gate on a non-empty data[4] instead (or bump to 1.152.1).
2. Notes Editor "Focus" offers to delete the note of any off-screen burg/marker/market —
   validateHighlightElement looks the element up in the DOM, which the viewport renderers
   no longer hold. Check pack and zoom to the entity instead.
3. Legacy state/province notes get an <h3>ShortName</h3> heading — the labels editor
   titled them with the short name; the migration compares to fullName. "Duchy of
   Crystormia" → <h3>Crystormia</h3><p>…</p>.
4. Saved dialog positions restore off-screen — no clamp on restore; a dialog dragged on a
   wide monitor and reopened in a narrower window is unreachable until cleanupData.
5. (code) Feature restore matches by cell overlap only: a named/noted lake raised to land hands
   its name+note to the enclosing island — add a type check.
6. (code) Lakes never show their note on hover: <use href="#feature_N"> has no id, the tooltip
   resolves by element id.
7. (code) The duplicate-marker-id repair now runs after the notes migration, so the renumbered
   marker loses its note.

Smaller: draw-markets.ts template writes literal false/undefined text nodes into partial
views (and the saved SVG); marker notes titled with their own id become the marker name;
markers-in-radius still writes the retired pinned attribute; contour paths add ~30 % to a
.map at interval 1 (linear in cells); Changelog vs latestPublicChanges disagree and neither
lists the markets renderer.

[barrulus]

Ran the branch at 1f84f46 over 777 saved maps (Discord attachments 2019–2026, versions
0.58b–1.150.0) and the merge base c222645, headless, fresh profile per map, then save → reload
on the branch. I have the per-map table with the Discord link of every failing file; say if you want it.

• No load regressions. 628 maps load on the branch vs 622 on master; the 6 extra are files
  master rejected (mapSize on number, options.military is not iterable, Unterminated string in JSON, getWinds) — the repair-not-reject parsing pays off. Pack counts are
  identical on all shared maps; save → reload is identical on 621/628.
• 146 maps fail on both builds, pre-existing and catalogued by error and version in the
  spec: 40× 0.70a/0.80b flatMap in Lakes.defineShoreline, 10× 1.0/1.1 neighbors.map, 9×
  striping, 7× split in parseLoadedData, 6× habitability in goods, ~12 corrupt JSON, 6
  files whose CRLF delimiters were converted to LF.

Findings:

1. Notes: 180 of 628 maps (29%) open with "Notes without an element", and 92% of the 32,389
   notes offered are stale auto-generated regiment descriptions. Old builds never pruned
   regimentS-R notes when a military was regenerated; one 1.105 map carries 19,300 notes of
   which 19,232 are dead regiments and 68 are real. Suggest dropping unmatched regiment notes
   silently (they are generator output) or splitting the count, so the dialog is about what the
   user wrote.
2. Pre-1.83 version strings compare wrongly, and the notes migration turns that into
   loss. parseMapVersion("1.61") is 1.61.0, so isOlderThan("1.7.0") is false for
   1.61–1.66 maps and the v1.7 marker migration never runs; likewise 1.11/1.22 skip the 1.3–1.7
   blocks, 1.7/1.8 re-run the 1.61–1.652 blocks, and "0.9b" (→ 0.9.0 < 0.70.0) is rejected as
   ancient while 0.70a/0.80b get through. Since eb29c5e (2024-09); before that the gates were
   numeric. In this corpus 98 of 105 loadable 1.11/1.22/1.6x maps keep their markers as SVG
   <use> elements with pack.markers empty. On master that is invisible; on the branch their
   markerElementN notes (2,174 notes on 99 maps, e.g. a 1.61 map loses 244 of 424) are offered
   as CSV and dropped. Fix outside this PR: parse compact versions monotonically
   (1.22 → 1.2.20, 1.61 → 1.6.10, 1.652 → 1.6.52, 1.7 → 1.7.0) and pass the gate strings through
   the same parser; then the v1.7 block converts the markers and noteRenames keeps the notes.
3. lake_N (232 notes on 8 maps) and roadN ids are not in ELEMENT_PATTERNS — old
   feature/route ids carrying hand-written lore.
4. LOW: a 1.22 map with a legacy #fogging nested in #fogging-cont ends up with two
   #fogging groups after save → reload.