Backfill missing dive metadata on Shearwater re-import (PRO-62)

[shaheinm]

Summary

• Root cause of PRO-62: GF settings are only extracted from the libdivecomputer binary-log parse, and dives imported before that extraction existed have NULL gf_low/gf_high. Fingerprint dedupe made this permanent: re-importing the same Shearwater Cloud .db skipped existing dives entirely, so old rows could never be repaired — and the replay sheet fell back to its 30/70 defaults.
• Fix: when a re-import encounters an already-known dive (by source fingerprint or legacy fingerprint), it now backfills metadata fields that are nil — GF low/high, deco model, salinity, surface pressure, end GF99 — from the source row before skipping. Existing non-nil values are never overwritten, and the binary log is only re-parsed when at least one target field is actually missing, so re-imports of healthy databases stay cheap.
• Adds a divesBackfilled counter to ShearwaterCloudImportResult (defaulted, source-compatible).

User action to repair existing data: re-export from Shearwater Cloud and re-import the .db — affected dives will be backfilled in place.

Closes PRO-62.

Tests

• testReimportBackfillsMissingMetadata — first import populates, fields nulled to simulate an old import, re-import restores them (divesBackfilled == 1, no duplicate dives)
• testReimportDoesNotBackfillWhenNothingMissing — fully populated dives are untouched and existing values are never overwritten (divesBackfilled == 0)
• Full Swift suite green locally; SwiftLint strict: 0 violations

Made with Cursor

[codecov]

Codecov Report

❌ Patch coverage is 85.91549% with 10 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ources/Services/ShearwaterCloudImportService.swift	85.91%	10 Missing ⚠️

📢 Thoughts on this report? Let us know!

[shaheinm]

Second-opinion review (gpt-5.3-codex) — no blockers → addressed in latest commit

Should-fixes (both addressed):

• updatedDiveIds short-circuited multi-computer groups: once row A updated a dive, row B couldn't fill remaining nil fields. → Removed the skip; the set now only counts unique dives. Test: testReimportBackfillsFromSecondComputerRow.
• Partial-merge path (existingFp != nil) didn't backfill the existing dive. → Backfill now runs after the merge write so the newly inserted fingerprints resolve and the new computer's row fills gaps. Test: testPartialMergeBackfillsMissingMetadata.

Confirmed correct by review: legacy fingerprint fallback, nil-only updates (never overwrites), transaction safety, source-compatible divesBackfilled struct change.

Full Swift suite green, SwiftLint clean.