[timlichtenberg]

Summary

The board read now guards the items, but the value lists inside an item still trust hasNextPage alone. If the API returns an empty value list for one item, the run reads every field of that item as cleared and clears it at the other places. This change makes those reads detect short lists, holds back a place whose values vanish without a sign of a clear, and limits the sync to one writing run per hour.

Changes

• Board read: an issue item with no field values, or a number of values other than its totalCount, makes the board read short, so it is read again like a short item page. Every item holds at least its title.
• Issue field values: a batch that returns another number of issues than asked for, or an issue with a number of values other than its totalCount, is read again (up to 6 reads, 10 seconds apart), then the run stops before any write. Null lists and null entries count as short. An issue that lacks the value of a synced field whose latest event set it counts as unread.
• A board place of an issue whose synced fields held 2 or more values and all read empty is held back for the run. The issue fields of an issue are held back when a synced field that held a value reads empty without a removal event. A held place clears nothing and is listed in report.txt; if it reads the same in the next run, it receives the agreed values. A hand clear of all synced fields of an issue on one board is therefore undone.
• A run with apply that starts less than 50 minutes after the start of the last run that tried a write stops before the read with exit code 5. A scheduled run shows that as a notice, a run started by hand fails. Only a run that tries a write stores its start time, so a run without writes still makes no commit in the state repository. A stored time more than 5 minutes ahead of the run is a state error.
• The workflow runs once an hour at minute 11.
• README: the new rules.

Validation

• python -m pytest -q: 197 passed.
• coverage run -m pytest -q && coverage report: 99% total; __main__.py and github.py 100% of lines and branches, core.py 99% (the 2 partial branches are also on main).
• ruff check . and ruff format --check .: clean.
• I changed each new condition in turn (30 changes in all, for example != to <, the WIPE_MIN value, the removal-time test, the stamp rule, the clock check, the null handling); every change makes at least 1 test fail.
• The Sync step script with the run replaced by a fixed exit code: exit 5 gives 0 with a notice on schedule and 5 with an error on workflow_dispatch; exit 0, 2 and 4 are unchanged.
• python -m board_sync without --apply against the 3 boards and a copy of the stored state, with a personal token: all reads full, no place held back, no unread issue, state file unchanged.

Not checked

• An item with one synced value on a board can still lose it to a read that looks complete; the count rule needs 2 values, and a board gives no clear time.
• The new checks were not run with the App token the workflow uses.