sync: auto-sync from AD2 at 2026-06-18 07:23:32

Author: Mike Swanson
Machine: AD2
Timestamp: 2026-06-18 07:23:32

projects/dataforth-dos/session-logs/2026-06/2026-06-18-mike-dsca-fix2-stage2-3-publish.md

@@ -1,5 +1,27 @@
 # DSCA Datasheet Fix 2 — STAGE 2 wire-in, STAGE 3 validator, publish of 68 clean models
 
+## Update: 07:23 PT — rounding-mode fix (QB single-precision) + re-publish of 85 clean models
+
+Fixed the rounding-mode issue behind the 26 last-digit-diff models. Root cause: the DOS
+QuickBASIC computed/stored values as single-precision floats, so half-boundary rounding follows
+single, not double, precision. `formatMeasuredExact` now applies `Math.fround(value)` before
+`toFixed`. Verified each boundary case against the staged golden (9.9995 -> "10.000", 46.85 ->
+"46.8", .45 -> "0.4", 3.3325 -> "3.332").
+
+STAGE 3 re-validation: FINAL-TEST CLEAN models **68 -> 85** (+17), mismatches 26 -> 9, cert matches
+2123 -> 2206. Zero regression — every remaining dirty model was already dirty; no clean model flipped.
+
+Re-pushed all 85 clean models (37,168 PASS serials): **updated=6054 unchanged=28949 created=0
+errors=0 skipped=2165**. Live site is now fround-correct across 85 DSCA models (the 6,054 updates =
+boundary corrections in the original 68 + the 17 newly-clean models).
+
+The 9 still-dirty models are NOT rounding: 4 (DSCA38-05/-1793/-19C/-19E) are Supply Current retest
+data-vintage (staged .TXT predates the DB record — not a render bug); 5 (DSCA39-01/02/05/07/1950)
+are the STAGE 1 footer-note artifact ("Standard output load..." mis-captured as a truncated row).
+A renderer-side fix for the footer note was attempted but regressed 24 clean current-output `-C`
+models (the note placement differs per golden), so it was reverted — this needs a targeted STAGE 1
+extractor fix, not a renderer change. Commit `14ee61dc`.
+
 ## User
 - **User:** Mike Swanson (mike)
 - **Machine:** AD2