The seating test waits for the disk, the legend names only marks that appear (v0.60.2) (#43)

CI broke on the v0.60.1 tag run: e2e "a reopened seating restores the
scroll position" failed both attempts. flushConfig only queues the
configSave, and the worker's persist of it is fire-and-forget
(engine.worker.ts `void persistNow()`), so a reload that outruns the
IndexedDB write reopens on the previous, verse-less save — the restore
lands at the top of the chapter and the final poll returns null forever.
The test now awaits rpc.flush() before reloading, exactly as its two
sibling tests already did. Mutation-checked: with the v0.60.1 bug
reintroduced (chapter-only slot write) the test still goes red at the
savedVerse assertion; with the fix restored it is green repeated 3x.

The provenance legend drops "research-grade": no shipped data can
produce the mark — quotation-pairs.json ships zero links and
text-witness.json has no runtime reader — so the legend explained a
glyph no reader can ever see. The per-chip tier_marks flask stays wired
should a graded harvest ever ship. The "weigh it" aside on the machine
item goes too (plain-copy rule). The legend is emitted by the core
producer, so both shells change at once.

Author: glendonjklassen

Cargo.lock

@@ -21,7 +21,7 @@ default-members = [
 # shells — About renders it as `engine ` + this number. The workflow refuses a tag
 # whose version disagrees with this line (and with apps/web/package.json), so a
 # release ships one number, not three.
-version = "0.60.1"
+version = "0.60.2"
 edition = "2021"
 license = "MIT"
 authors = ["Glendon Klassen <gjklassen@proton.me>"]

Cargo.toml

@@ -131,9 +131,15 @@ test("a reopened seating restores the scroll position, not just the chapter", as
     .toBeGreaterThan(0);
 
   // What the close is about to bank: this seating's slot, VERSE INCLUDED.
-  const savedVerse = await page.evaluate(() => {
+  // The flush is AWAITED before the reload: flushConfig only queues the
+  // configSave, and the worker's own persist of it is fire-and-forget — a
+  // reload that outruns the IndexedDB write reopens on the previous,
+  // verse-less save and this test goes red with the fix in place (it did,
+  // twice, on the v0.60.1 tag run).
+  const savedVerse = await page.evaluate(async () => {
     const s = (window as any).__plumbline;
     s.flushConfig();
+    await s.rpc.flush();
     return s.config.slots?.[s.slot]?.verse;
   });
   expect(savedVerse).toBeGreaterThan(1);

apps/web/e2e/session-slots.spec.ts

@@ -1,7 +1,7 @@
 {
   "name": "plumbline-web",
   "private": true,
-  "version": "0.60.1",
+  "version": "0.60.2",
   "type": "module",
   "scripts": {
     "dev": "npm run pack:i18n && vite",

apps/web/package-lock.json

@@ -695,9 +695,7 @@ fn legend() -> Block {
         Run::new("†", sz::MARK, Color::TierHuman),
         Run::new(" curated scholarship  ·  ", sz::MARK, Color::Faded),
         Run::new("≈", sz::MARK, Color::TierMachine),
-        Run::new(" machine-derived, weigh it  ·  ", sz::MARK, Color::Faded),
-        Run::new("⚗", sz::MARK, Color::TierResearch),
-        Run::new(" research-grade", sz::MARK, Color::Faded),
+        Run::new(" machine-derived", sz::MARK, Color::Faded),
     ])
 }
 

apps/web/package.json

@@ -440,8 +440,13 @@ producer emits*, not shell code.
 10. **margin notes** *(Human †)* — the verse's 1769 notes, small.
 
 A **provenance legend** closes a Full-study card once: "where this comes from:
-✝ the text · † curated scholarship · ≈ machine-derived, weigh it · ⚗
-research-grade". Weave membership + cross-references (items 5–6) and tags carry no mark (mixed /
+✝ the text · † curated scholarship · ≈ machine-derived". The ⚗ research-grade
+item was dropped from the legend: no shipped data produces the mark
+(`quotation-pairs.json` ships zero links and `text-witness.json` has no
+runtime reader), so the legend explained a glyph no reader can see. The
+per-chip `tier_marks` flask stays wired should a graded harvest ever ship.
+The "weigh it" aside on the machine item went with it (plain-copy rule).
+Weave membership + cross-references (items 5–6) and tags carry no mark (mixed /
 user-authored, not trust-weighted evidence). The producer emits it as a `Para`
 of tier-coloured runs.