fix(transition): advance DEADLINE repeater on terminal transition (v0.6.1)

[miles-on-nightshift]

Summary

• transition() now checks both SCHEDULED and DEADLINE for a repeater cookie
• _advance_repeater() parameterised with timestamp_attr to advance the right planning timestamp
• SCHEDULED checked before DEADLINE (matches Emacs behaviour when both carry a cookie)

Root cause

transition() only tested raw_node.scheduled for a repeater. Tasks with the cookie on DEADLINE transitioned to DONE and terminated rather than recurring.

Reported by @Amperture: nvim-orgmode habit tracker uses DEADLINE as a workaround for missing org-habit support.

Test plan

• test_transition_recurring_advances_deadline — .+1d habit, +1w plain advance, no-repeater terminal all verified
• Existing test_transition_recurring_advances_scheduled — no regression (284 tests pass)

Version

Bumps to v0.6.1 (v0.6.0 shipped 2026-09-23 before this fix landed).

Closes #17.

🤖 Generated with Claude Code

[miles-on-nightshift]

CTO pr-review (2026-09-23)

Code review: LGTM. The logic change is correct and minimal — transition() now checks deadline when scheduled has no repeater, and _advance_repeater() correctly parameterises which attribute to advance via timestamp_attr. SCHEDULED-first ordering matches Emacs behaviour. Test coverage is solid: habit (.+1d), plain advance (+1w), and no-repeater terminal all verified; existing SCHEDULED test suite passes without regression (284 tests).

Blocker fixed: All 4 Python-version CI runs failed on ruff I001 — missing blank line between stdlib and third-party imports in the new test (line 296). Fixed and pushed in 643c940. CI should rerun now.

Ready to approve once CI is green.

[miles-on-nightshift]

CI green across all 4 Python versions (3.10–3.13). PR is ready for merge by @plur9.

[miles-on-nightshift]

Rebased onto main (v0.6.0 → v0.6.1 version bump). CI re-running. Ready to merge once green — fixes the DEADLINE repeater bug reported by @Amperture.

[miles-on-nightshift]

CTO pr-review (2026-09-24)

Code is correct. transition() now checks SCHEDULED first, then DEADLINE for a repeater — SCHEDULED-first ordering matches Emacs behaviour. _advance_repeater() parameterised with timestamp_attr (default "scheduled") is the minimal correct change; getattr/setattr usage is clean and backwards-compatible.

Test coverage is solid: habit (.+1d), plain advance (+1w), no-repeater terminal — all 3 paths covered. The .+1d assertion against date.today() is intentional (habit restarts from completion). 284 existing tests pass. Sep 23 ruff I001 blocker was fixed; all 5 CI checks green (Python 3.10–3.13 + Coverage).

The commit headline reads "v0.5.5" but pyproject.toml correctly shows 0.6.0 → 0.6.1 — stale message, not a blocker.

LGTM — ready to merge. @plur9 approve and merge when convenient.