You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(watch): stop re-reading the ignore rules on a read of their files - #3584
cgr watch re-read its ignore rules on any event naming .gitignore, .cgrignore or the git index, reads included. Re-reading them opens those same files, so on Linux each reload raised the opened / closed_no_write events of the next one. A single git status (VS Code's Source Control panel runs them) started a loop that re-read .gitignore every few milliseconds, forever, flooding the log with Loaded N exclude and 0 unignore patterns.
Now only a write, creation, deletion or rename of one of those files re-reads the rules (RULE_INPUT_EVENT_TYPES). Editing an ignore file, git add, git mv and git rm still apply exactly as before.
Demo
Terminal output of a reproduction script that runs the real watchdog observer and real IgnoreRules on a temporary git repo, counting reads of .gitignore over 2 seconds (no VHS recording available in this environment):
Before (main):
one read-only `cat .gitignore`, then wait 2s ...
.gitignore reads after one read: 801
touch a file + two `git status` runs (VS Code Source Control), then wait 2s ...
.gitignore reads after one read: 884
After (this branch):
one read-only `cat .gitignore`, then wait 2s ...
.gitignore reads after one read: 0
touch a file + two `git status` runs (VS Code Source Control), then wait 2s ...
.gitignore reads after one read: 2
The 2 remaining reads are the expected single reload after git status actually rewrote the index.
Unit tests pass (make test-parallel or uv run pytest -n auto -m "not integration") — ran the watcher suites only (212 passed, 25 skipped), not the full suite
New tests added — codebase_rag/tests/test_watcher_ignore_rule_reads.py: reads of each input leave the rules alone; writes, creations, deletions, closes-after-write and the index.lock → index rename still reload them; and an end-to-end test on the real inotify observer shows one read of .gitignore sets off no reloads. 7 of these fail on main.
A breaking change has ! in the title and its release notes in changes/breaking/<PR number>.md (not a breaking change)
All pre-commit checks pass (make pre-commit) — all pass except ty check, which fails locally only on unresolved imports of optional extras in untouched files; no new diagnostics
No hardcoded strings in non-config/non-constants files
No # type: ignore, cast(), Any, or object type hints
New comments and docstrings explain why, not what
Summary by CodeRabbit
Bug Fixes
Ignore rules no longer reload when their input files are only read. Changes such as edits, file creation or deletion, and relevant renames continue to trigger reloads.
The watcher re-read `.gitignore`, `.cgrignore` and the git index on any
event naming them, reads included. Re-reading them opens those same files,
so on Linux each reload raised the `opened` and `closed_no_write` events of
the next: one `git status` (VS Code's Source Control panel runs them) set
off a loop that re-read `.gitignore` every few milliseconds, for good.
Only a write, creation, deletion or rename of an input re-reads the rules
now. Edits to the ignore files and `git add`/`git mv` still apply as before.
Fixes#3583
No actionable comments were generated in the recent review.
Walkthrough
The watcher now filters filesystem events by event type and configured input paths before reloading ignore rules. New tests cover read events, relevant changes, index replacement, and Linux observer behavior.
Adds RULE_INPUT_EVENT_TYPES, combining content events with moved and closed events.
Filter reloads and test watcher behavior codebase_rag/watch.py, codebase_rag/tests/test_watcher_ignore_rule_reads.py
The watcher reloads ignore rules only when an eligible event involves a configured input. Tests cover reads, changes, index replacement, and a Linux observer.
The change prevents read-triggered ignore-rule reloads while preserving reloads for relevant file changes. No actionable merge-blocking issue remains; merge after normal checks.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title "fix(watch): stop re-reading the ignore rules on a read of their files" is a Conventional Commits title that matches the main change. The PR makes the watcher reload ignore rules only on wri…
Linked Issues check
Issue #3583 has one direct coding objective group. Its requirements are: (a) reads of .gitignore, .cgrignore, or .git/index must not trigger ignore-rule reloads; (b) ignore-rule changes must still tri…
Out of Scope Changes check
The whole-PR diff against the merge base changes three files, and each serves issue #3583.
1. codebase_rag/constants/core.py: adds RULE_INPUT_EVENT_TYPES. It defines the event types that can change a…
Description check
The description covers the required sections well. It gives a clear Summary that explains the bug (ignore-rule reloads triggered by reads, causing a loop on Linux) and the fix (reload only on write, c…
Full details: Title check
Explanation
The title "fix(watch): stop re-reading the ignore rules on a read of their files" is a Conventional Commits title that matches the main change. The PR makes the watcher reload ignore rules only on write, create, delete, or rename events, not on reads. The title names the bug fix and the affected component (watch) and the trigger (read events), so it is specific and accurate.
Full details: Linked Issues check
Explanation
Issue #3583 has one direct coding objective group. Its requirements are: (a) reads of .gitignore, .cgrignore, or .git/index must not trigger ignore-rule reloads; (b) ignore-rule changes must still trigger reloads; (c) source monitoring must continue without excess processing or logging.
(a) Implemented. In codebase_rag/watch.py, _follow_ignore_rules now calls _apply_ignore_rules only when event.event_type is in RULE_INPUT_EVENT_TYPES. RULE_INPUT_EVENT_TYPES (codebase_rag/constants/core.py) is CONTENT_EVENT_TYPES (modified, created, deleted) plus MOVED and CLOSED. EventType.CLOSED is "closed", the close-after-write event. Read events ("opened", "closed_no_write") are excluded, so a read no longer starts a reload. Reloading only reads the files, so the reload loop is broken.
(b) Implemented. Edits (modified), creations, deletions, close-after-write, and renames (MOVED, including index.lock onto index, which git status and git add use) still reload when the path is in rules.inputs. The tests in codebase_rag/tests/test_watcher_ignore_rule_reads.py cover these cases.
(c) Unchanged. The return value of _follow_ignore_rules is unchanged, so non-ignore source events still follow the same path. Only the reload decision changed.
Tests: parametrized tests cover reads (zero reloads) and changes (one reload) for each input. A Linux end-to-end test uses the real observer and checks that one read of .gitignore causes no loads. I did not run the test suite. The PR reports 212 passed and 25 skipped for the watcher suites. The code evidence supports the requirements.
Not assessed: the issue's SQLite backend note is informational, not a coding requirement.
Full details: Out of Scope Changes check
Explanation
The whole-PR diff against the merge base changes three files, and each serves issue #3583.
codebase_rag/constants/core.py: adds RULE_INPUT_EVENT_TYPES. It defines the event types that can change an ignore rule's input. Its comment explains the link to #3583.
codebase_rag/watch.py: imports RULE_INPUT_EVENT_TYPES and gates the reload in _follow_ignore_rules on it. The docstring is updated to match. No other logic changes.
codebase_rag/tests/test_watcher_ignore_rule_reads.py: new tests for the reload behavior described in the issue. These support the change and are not unrelated.
No unrelated files or behavior changes appear in the reviewed diff. The PR description's demo and checklist items are documentation of the change, not code changes.
Full details: Description check
Explanation
The description covers the required sections well. It gives a clear Summary that explains the bug (ignore-rule reloads triggered by reads, causing a loop on Linux) and the fix (reload only on write, create, delete, or rename events via RULE_INPUT_EVENT_TYPES). It includes a Demo with before and after reproduction output, which fits the template's guidance for changes that cannot be shown as a VHS GIF. The Type of Change is marked as Bug fix, Related Issues links Fixes #3583, and the Test Plan lists new tests and manual testing. The Checklist is filled in. Gaps: the unit-test box is unchecked, and the author states only watcher suites were run (212 passed, 25 skipped), not the full suite. The author also says the ty check failure is local and unrelated. The breaking-change item is checked with a note that it is not breaking, which is acceptable. Overall the description is mostly complete.
Finishing Touches
Generate docstrings
Commit to this branch
Create a new PR
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 7 remain after this review.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
cgr watchre-read its ignore rules on any event naming.gitignore,.cgrignoreor the git index, reads included. Re-reading them opens those same files, so on Linux each reload raised theopened/closed_no_writeevents of the next one. A singlegit status(VS Code's Source Control panel runs them) started a loop that re-read.gitignoreevery few milliseconds, forever, flooding the log withLoaded N exclude and 0 unignore patterns.RULE_INPUT_EVENT_TYPES). Editing an ignore file,git add,git mvandgit rmstill apply exactly as before.Demo
Terminal output of a reproduction script that runs the real watchdog observer and real
IgnoreRuleson a temporary git repo, counting reads of.gitignoreover 2 seconds (no VHS recording available in this environment):Before (
main):After (this branch):
The 2 remaining reads are the expected single reload after
git statusactually rewrote the index.Type of Change
Related Issues
Fixes #3583
Test Plan
make test-paralleloruv run pytest -n auto -m "not integration") — ran the watcher suites only (212 passed, 25 skipped), not the full suitecodebase_rag/tests/test_watcher_ignore_rule_reads.py: reads of each input leave the rules alone; writes, creations, deletions, closes-after-write and theindex.lock→indexrename still reload them; and an end-to-end test on the real inotify observer shows one read of.gitignoresets off no reloads. 7 of these fail onmain.make test-integration, requires Docker)Checklist
!in the title and its release notes inchanges/breaking/<PR number>.md(not a breaking change)make pre-commit) — all pass exceptty check, which fails locally only on unresolved imports of optional extras in untouched files; no new diagnostics# type: ignore,cast(),Any, orobjecttype hintsSummary by CodeRabbit