find: implement -ignore_readdir_race and -noignore_readdir_race by DominicThemiya · Pull Request #903 · uutils/findutils

DominicThemiya · 2026-10-13T15:49:49Z

Closes #377

What

find rejected both options as unknown predicates:

$ find dir -ignore_readdir_race -name a
find: unknown predicate `-ignore_readdir_race'

This adds them as global options, matching GNU find 4.11.0:

They apply to the whole expression wherever they appear.
The last of -ignore_readdir_race / -noignore_readdir_race wins.
With -ignore_readdir_race, a file or directory that vanishes after its parent directory was read is not reported during the walk.
With -ignore_readdir_race, -delete on a file that has already disappeared prints nothing and evaluates to true.
Starting points that don't exist are still reported, as GNU find does.

$ mkdir -p d/b && touch d/a
$ find d -sorted -name a -exec rm -rf d/b \; -print
d/a
Error: d/b: No such file or directory (os error 2)                  # exit 1, unchanged
$ mkdir d/b
$ find d -ignore_readdir_race -sorted -name a -exec rm -rf d/b \; -print
d/a                                                                 # exit 0

How

Config gets an ignore_readdir_race flag, set while parsing the expression.
In process_dir, a walk error is skipped when the flag is set, is_not_found() is true and the depth is greater than 0, so starting points are still reported.
process_dir passes the flag to every MatcherIO, including the one used for finished_dir/finished, so DeleteMatcher reads it when it runs rather than when it is built. That is what makes the option work when it comes after -delete.
DeleteMatcher treats ErrorKind::NotFound as success when the flag is set.
Both options are listed in --help.

Not covered

GNU also silences the error when a test that needs stat (e.g. -size) hits a vanished file. Here each matcher prints its own error, so that needs a wider change (related to CodeHealth: tidy up Errors #15) and is left for a follow-up.

Tests

find::matchers::delete::tests::delete_matcher_vanished_file
delete_vanished_file_is_an_error_by_default (tests/test_find.rs)
delete_vanished_file_with_ignore_readdir_race (tests/test_find.rs)
vanished_dir_during_walk_is_an_error_by_default (tests/test_find.rs)
vanished_dir_during_walk_with_ignore_readdir_race (tests/test_find.rs)
missing_start_point_still_reported_with_ignore_readdir_race (tests/test_find.rs)

The walk tests use -sorted so that a is always visited before b.

cargo test, cargo fmt --check and cargo clippy --all-targets -- -D warnings pass locally. The GNU testsuite (util/build-gnu.sh, including the DejaGnu tests) gives the same results on main and on this branch: 495 tests, 425 passed, no test changed status.

Both options were rejected as unknown predicates. They are now accepted as global options: they apply wherever they appear in the expression, and the last one given wins, as in GNU find. With -ignore_readdir_race, -delete no longer reports a file that has already disappeared, and evaluates to true. Errors from tests that stat a vanished file (e.g. -size) are still reported; those are printed by each matcher and are left for a follow-up.

codecov · 2026-10-13T15:57:27Z

Codecov Report

❌ Patch coverage is 97.72727% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 92.60%. Comparing base (2b15f5d) to head (6dc7f51).

Files with missing lines	Patch %	Lines
src/find/matchers/delete.rs	95.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #903      +/-   ##
==========================================
+ Coverage   92.52%   92.60%   +0.08%     
==========================================
  Files          36       36              
  Lines        7694     7738      +44     
  Branches      391      392       +1     
==========================================
+ Hits         7119     7166      +47     
+ Misses        437      433       -4     
- Partials      138      139       +1

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

codspeed · 2026-10-13T16:00:14Z

Merging this PR will not alter performance

⚠️

15 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

combined_expr

prune

split_null

batched_n

split_whitespace

batched_size

size_gt

walk_all

type_f

printf

…and 5 more

✅ 20 untouched benchmarks

_{Comparing DominicThemiya:ignore-readdir-race (6dc7f51) with main (2b15f5d)}

github-actions · 2026-10-13T16:02:25Z

Commit 6dc7f51 has test result changes:

bfs testsuite:

Test results comparison:
  Current:   TOTAL: 317 / PASSED: 276 / FAILED: 35 / SKIPPED: 6
  Reference: TOTAL: 317 / PASSED: 274 / FAILED: 37 / SKIPPED: 6

Changes from main branch:
  TOTAL: +0
  PASSED: +2
  FAILED: -2

Test improvements (2):
  + gnu/ignore_readdir_race
  + gnu/ignore_readdir_race_notdir

sylvestre · 2026-10-24T08:05:00Z

            }
            Ok(entry) => {
                let mut matcher_io = matchers::MatcherIO::new(deps);
+                matcher_io.set_ignore_readdir_race(config.ignore_readdir_race);


what about the Err(err) branch just above?
a file or dir vanishing during the walk ends up there and is still reported with exit 1.
that's the main readdir race case, no? please handle it too (with is_not_found()), or change "Closes" to a reference

Good catch, thanks. Handled in 1fc1ff1: a walk error is now skipped when the option is set, is_not_found() is true and the depth is greater than 0, so missing starting points are still reported, as GNU does. Covered by vanished_dir_during_walk_is_an_error_by_default, vanished_dir_during_walk_with_ignore_readdir_race and missing_start_point_still_reported_with_ignore_readdir_race.

sylvestre · 2026-10-24T08:05:02Z

            }
            Ok(entry) => {
                let mut matcher_io = matchers::MatcherIO::new(deps);
+                matcher_io.set_ignore_readdir_race(config.ignore_readdir_race);


the MatcherIO built after the loop for finished_dir/finished doesn't get the flag, could you please set it there too?

Done in cdcd9e7.

sylvestre · 2026-10-24T08:05:04Z

+            assert_eq!(
+                deps.get_output_as_string(),
+                fix_up_slashes("./test_data/simple/subdir\n./test_data/simple/subdir/ABBBC\n")
+            );


this only checks that the option is parsed, the integration tests already cover it. do we need it?

Agreed, the integration tests cover parsing. Removed in d511206.

sylvestre · 2026-10-24T08:05:13Z

 -d / -depth              Process directory contents before directory itself
 -mount / -xdev           Do not descend into other file systems
 -noleaf                  Do not optimize by assuming 2+ hard links
+ -ignore_readdir_race     Do not report files that vanish during the search


did you run the GNU testsuite for this?
git clone https://git.savannah.gnu.org/git/findutils.git ../findutils.gnu
bash util/build-gnu.sh tests/find/.sh

Yes, I ran it locally with util/build-gnu.sh (full suite, including the DejaGnu tests) on main and on this branch: 495 tests, 425 passed in both, and no test changed status. GNU's suite has no -ignore_readdir_race test, though; the bfs suite covers it, and CI on the first commit showed gnu/ignore_readdir_race and gnu/ignore_readdir_race_notdir going from failing to passing.

…_race A file or directory removed after its parent directory was read reaches the walk error branch and was still reported with exit status 1, even with -ignore_readdir_race. Skip such "not found" errors when the option is set. Starting points (depth 0) are still reported, as GNU find does.

The MatcherIO built after the walk for finished_dir() and finished() did not get the flag, unlike the one built for each entry.

The integration tests already pass both options and check their effect.

DominicThemiya mentioned this pull request Oct 14, 2026

Implement -ignore_readdir_race and -noignore_readdir_race #377

Open

sylvestre reviewed Oct 24, 2026

View reviewed changes

DominicThemiya added 3 commits October 24, 2026 14:34

find: pass -ignore_readdir_race to the final MatcherIO

cdcd9e7

The MatcherIO built after the walk for finished_dir() and finished() did not get the flag, unlike the one built for each entry.

find: drop the parse-only -ignore_readdir_race unit test

d511206

The integration tests already pass both options and check their effect.

DominicThemiya requested a review from sylvestre October 25, 2026 01:22

Navigation Menu

Repository navigation

find: implement -ignore_readdir_race and -noignore_readdir_race - #903

find: implement -ignore_readdir_race and -noignore_readdir_race#903
DominicThemiya wants to merge 4 commits into
uutils:mainfrom
DominicThemiya:ignore-readdir-race

DominicThemiya commented Oct 13, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Oct 13, 2026

Uh oh!

codspeed Bot commented Oct 13, 2026

Uh oh!

github-actions Bot commented Oct 13, 2026

Uh oh!

sylvestre Oct 24, 2026

Uh oh!

DominicThemiya Oct 24, 2026

Uh oh!

sylvestre Oct 24, 2026

Uh oh!

DominicThemiya Oct 24, 2026

Uh oh!

sylvestre Oct 24, 2026

Uh oh!

DominicThemiya Oct 24, 2026

Uh oh!

sylvestre Oct 24, 2026

Uh oh!

DominicThemiya Oct 24, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

DominicThemiya commented Oct 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

How

Not covered

Tests

Uh oh!

codecov Bot commented Oct 13, 2026

Codecov Report

Uh oh!

codspeed Bot commented Oct 13, 2026

Merging this PR will not alter performance

Uh oh!

github-actions Bot commented Oct 13, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

DominicThemiya commented Oct 13, 2026 •

edited

Loading