Files
OrbitTv/VERIFICATION-0.21.0.md
pauljdandClaude Opus 5.5 6e2d3df225
pauljd/davidson-ci: Davidson CI / quality (push) Failing after 2m2s
pauljd/davidson-ci: Davidson CI / ai-review (push) Successful in 58s
Fix import falsely finishing on a zero total (0.21.0)
Turning the count off makes the provider return a total of zero, which
0.20 took as the real total, so it called the import finished at 65,400.
Ignores totals it didn't ask for, and repairs a box left in that state.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-01 19:55:43 +00:00

1.1 KiB

Orbit TV 0.21.0 — import position fix

Reported on the box with 0.20: position 65,400 "of zero" with only Back selectable. Cause, a bug in 0.20: with EnableTotalRecordCount=false the provider appears to return a total of zero, and the import read that as the real total, marked itself finished and disabled its buttons. Nothing was lost. It also shows the page after 65,200 was answered, so 0.20's changes got past the earlier failure, though one page isn't enough to prove it.

Changes: the first page of each run asks for the total; other pages ignore any total returned. A retry skips the count. A saved total of zero, and "finished" without a real total, are treated as not finished, so a box left in this state can carry on.

Validation: offline assembleDebug, testDebugUnitTest and lintDebug passed; 113 unit tests, zero failures or errors, no lint errors. The fixture now returns zero for a switched-off count, as the provider appears to. The saved-state repair is in CatalogueStore's SQLite code, which has no JVM test, so it needs the Xiaomi test.

Rollback: reinstall outputs/revision-0.20.0 (same database format), or revert this commit.