Fix dynamic ignored-directory watcher replanning and degraded status

[Bakul2006]

Pull Request

Linked issue

Closes #908

What & why

Fixes watcher planning when ignored directories are created or grow after the initial watch plan is installed.

Previously, the watch plan was calculated at startup. If an ignored directory such as node_modules/ or target/ did not exist at startup, its parent could receive a recursive watch. If the ignored directory was created later and grew beyond the existing split threshold, the watch plan was not reconsidered, allowing the ignored tree to fall under the recursive watch and potentially contribute to watcher/resource exhaustion.

This PR:

• Marks the watch plan dirty when directory changes can affect the ignored-tree split.
• Detects ignored-tree threshold crossings during shallow-watch synchronization.
• Releases obsolete watches before installing the recalculated plan.
• Preserves the existing watch adoption and budget behavior.
• Marks startup budget fallback as degraded, so crg-daemon status correctly reports degraded coverage.

Regression tests were added for dynamically created/growing ignored directories and the startup degraded-status behavior.

How it was tested

The following validations were run successfully:

python -m py_compile code_review_graph/incremental.py
git diff --check

Language/static diagnostics reported no errors.

The full pytest/lint/type-check commands could not be run in the current environment because pytest, uv, and networkx are unavailable.

uv run pytest tests/ --tb=short -q
uv run ruff check code_review_graph/
uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional

These commands should be run in a fully provisioned development environment before merging.

Checklist

• Tests added for new functionality
• All tests pass: uv run pytest tests/ --tb=short -q
• Linting passes: uv run ruff check code_review_graph/
• Type checking passes: uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
• Lines are at most 100 characters
• Docs updated where behavior changed (README, docs/, docstrings)

[github-actions]

code-review-graph review

Overall risk: 0.55 (MEDIUM) — 31 changed function(s)/class(es), 0 affected flow(s), 12 test gap(s)

Risk-scored changes

Risk	Level	Symbol	Location	Tested
0.55	medium	code_review_graph/incremental.py::_replan_if_requested	code_review_graph/incremental.py:2166	no
0.40	medium	code_review_graph/incremental.py::_plan_watch_paths_with_status	code_review_graph/incremental.py:1976	no
0.35	low	code_review_graph/incremental.py::request_replan	code_review_graph/incremental.py:2150	yes
0.35	low	code_review_graph/incremental.py::GraphUpdateHandler.dispatch	code_review_graph/incremental.py:2810	yes
0.35	low	tests/test_watch_robustness.py::TestRealObserver.Handler	tests/test_watch_robustness.py:1241	no
0.30	low	code_review_graph/incremental.py::_WatchSupervisor.__init__	code_review_graph/incremental.py:2070	yes
0.30	low	code_review_graph/incremental.py::_replace_watch	code_review_graph/incremental.py:2219	no
0.30	low	code_review_graph/incremental.py::GraphUpdateHandler	code_review_graph/incremental.py:2809	no
0.30	low	tests/test_watch_robustness.py::TestNewDirectoryAdoption.FailingReplacementObserver	tests/test_watch_robustness.py:402	yes
0.30	low	tests/test_watch_robustness.py::FailingReplacementObserver.schedule	tests/test_watch_robustness.py:403	no

Test gaps

• code_review_graph/incremental.py::_plan_watch_paths_with_status (code_review_graph/incremental.py:1976)
• code_review_graph/incremental.py::_remember_ignored_boundaries (code_review_graph/incremental.py:2154)
• code_review_graph/incremental.py::_replan_if_requested (code_review_graph/incremental.py:2166)
• code_review_graph/incremental.py::_replace_watch (code_review_graph/incremental.py:2219)
• code_review_graph/incremental.py::GraphUpdateHandler (code_review_graph/incremental.py:2809)
• ...and 7 more without direct tests

Token savings: this graph-backed report used ~43,019 fewer tokens (~89%) than reading every changed file in full (estimated, chars/4 approximation).

---

Powered by code-review-graph — local-first analysis; no code leaves the CI runner.

[Bakul2006]

Hey @tirth8205 just wanted to add on one note on validation: I added the regression tests, but I wasn't able to run the full pytest suite locally because the current environment is missing pytest, uv, and networkx.

The changes do pass py_compile, language diagnostics, and git diff --check.

Could you please run the full test suite in the project environment to verify the new watcher regression tests and the existing tests?

[tirth8205]

Changes required: pending replans are never processed when the watcher has only a recursive root and no shallow watches. Run python -m pytest tests/test_watch_robustness.py -k 'new_large_ignored_tree_replaces_recursive_root_watch or ignored_tree_growth_replans_recursive_root'; both new regressions fail because _replan_if_requested is called only inside the empty shallow-watch loop. Process that request independently and retain existing coverage until replacement registration succeeds.

[Bakul2006]

Hey @tirth8205 just did the same as you suggested me now the desired test are successfully passed

[tirth8205]

008051b fixes the original recursive-root replan tests, but existing source coverage is still lost when replacement registration raises OSError. Reproduce by growing node_modules beside src/deep/a.py and failing observer.schedule for src, or by removing the ignored tree and failing registration of a recursive root; after request_replan() and sync_watches(), later deep edits are no longer delivered. Four new preservation tests, including two real PollingObserver cases, pass main and the integration branch but fail this head. python -m mypy code_review_graph/ --ignore-missing-imports --no-strict-optional also reports five new Path/str errors in incremental.py, so the source changes still require rework.

[Bakul2006]

Hey @tirth8205 just addressed the failed replan coverage regression.

[tirth8205]

This no longer merges into staging. Conflicts in code_review_graph/incremental.py.

git fetch origin && git merge origin/staging
# resolve, then
uv run pytest tests/ -q
uv run ruff check code_review_graph/
uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
git push

When resolving:

• Rebase the watcher/replan branch onto current origin/staging and resolve code_review_graph/incremental.py: staging (82ef9d0 'match ignore rules on repo-relative update inputs') changed sync_watches() to take an ignore_patterns kwarg and to compute newly_included via _should_ignore(...previous_patter...
• After the rebase, re-run the full gate set (pytest 3.13, ruff, mypy) and re-check that tests/test_watch_robustness.py still exercises the replan + degraded-status paths against the merged sync_watches() semantics.

A large integration branch landed on staging today, which is why this drifted.

PRs now target staging, not main. Yours was retargeted already, so nothing to do there.

[tirth8205]

This no longer merges into staging. The one conflict is code_review_graph/incremental.py, and the branch is 144 commits behind.

Worth knowing before you resolve:

• code_review_graph/incremental.py: staging added a build-state checkpoint at the tail of full_build and moved the Git timeout into constants.py.

git fetch origin
git merge origin/staging
# resolve, then
uv run pytest tests/ -q
uv run ruff check code_review_graph/
uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
git push

I have not reviewed the change itself yet. That comes once it merges and the checks run against the merged state, since staging has moved a long way and the result is what matters.