NutriTrace

Fixes five second-pass code-review findings against v1.1.2-dev02. Signed APK attached, Docker :dev tag will update via CI.

What was wrong in dev02

1. Applying a legacy goal template silently didn’t apply the Kilojoules goal. Templates saved under v1.1.1 or earlier store their Kilojoules goal under the pre-migration key. dev02’s onMount migration handled that on page load but not on template apply, so tapping Apply would re-plant the legacy key and the Kilojoules row would read as unconfigured. Fixed: the migration now runs on template apply too.
2. Barcode scan in add-to-diary mode was firing a redundant v3 request. lookupBarcode already returns fully-hydrated data via v3, then the pickFood hydration path re-fetched the same barcode. Fixed: barcode-scan flow marks the result as already hydrated so pickFood skips the round-trip.
3. Repeated open-save of the Kilojoules goal drifted the stored value downward. Display used 2-decimal rounding, edit used integer rounding, so a no-edit save round-tripped to a slightly different kcal value each pass. Fixed: display now uses integer rounding to match, so open-save-no-edit is a no-op after the first save.
4. Migration silently discarded days-only Kilojoules goals. Legacy entries with per-day values but no shared max/min failed the migration gate and were deleted without transferring. Fixed: migration now accepts entries with a valid days array.
5. em-dashes in Unreleased CHANGELOG. Violated the standing rule. Also removed the false “typo tolerance” claim from the search-a-licious entry (verified last session that OFF’s search-a-licious cluster returns 0 hits for fuzzy queries); accurate benefit is better relevance ranking + multilingual analysis + future-proofing off deprecated v2.

Anyone on dev01 or dev02 should upgrade

Same recommendation as before: dev builds evolve fast, the release cycle catches things static analysis missed the first time. dev03 is now on the second-pass-clean version of the fixes.

docker compose pull && docker compose up -d