Solve issues with mutexes while running tests on MacOS M1 #677
+20
−11
Conversation
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
|
|
JoanFM
changed the title
alonre24
reviewed
May 22, 2025
JoanFM
changed the title
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #677 +/- ##
=======================================
Coverage 96.24% 96.24%
=======================================
Files 112 112
Lines 6278 6278
=======================================
Hits 6042 6042
Misses 236 236 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
alonre24
reviewed
May 22, 2025
alonre24
approved these changes
May 22, 2025
|
Backport failed for Please cherry-pick the changes locally and resolve any conflicts. git fetch origin 0.7
git worktree add -d .worktree/backport-677-to-0.7 origin/0.7
cd .worktree/backport-677-to-0.7
git switch --create backport-677-to-0.7
git cherry-pick -x 205f6bc2913e19442801cb78350616cf80919306 |
|
Backport failed for Please cherry-pick the changes locally and resolve any conflicts. git fetch origin 0.8
git worktree add -d .worktree/backport-677-to-0.8 origin/0.8
cd .worktree/backport-677-to-0.8
git switch --create backport-677-to-0.8
git cherry-pick -x 205f6bc2913e19442801cb78350616cf80919306 |
|
Successfully created backport PR for |
github-actions bot
pushed a commit
that referenced
this pull request
May 22, 2025
* refactor to use RAII when locking mutexes * capture mutex shared pointer by value to ensure validity at del time * Avoid deadlock involving GIL, and indexGuard * revert change to barrier and guard scoped locks * add small comment (cherry picked from commit 205f6bc)
github-merge-queue bot
pushed a commit
that referenced
this pull request
May 26, 2025
Solve issues with mutexes while running tests on MacOS M1 (#677) * refactor to use RAII when locking mutexes * capture mutex shared pointer by value to ensure validity at del time * Avoid deadlock involving GIL, and indexGuard * revert change to barrier and guard scoped locks * add small comment (cherry picked from commit 205f6bc) Co-authored-by: Joan Fontanals <jfontanalsmartinez@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Describe the changes in the pull request
2 issues are handled in this PR:
1- An
abortwas identified while running a test related toPyBatchIterator. The abort happened when the test was being shutdown. The error comes from thedeleterpassed to thePyBatchIteratorshared_ptrwhich captures thePyIndexby reference. The problem is that there is no guarantee that thePyIndexlifetime will survive theIterator. In that deleter amutexis released (which could be in an invalid state). The solution is to have themutexbe wrapped in a shared_ptr that will share the ownership with thePyBatchIteratordeleter as it captures it by value (incrementing its reference count).2- Deadlock found on
tests/flow/test_hnsw_parallel.py::test_parallel_insert_batch_search:indexGuardlocked in shared mode, and tries to recover the GIL in the getNextResultindexGuardexclusively, while N-1 has the indexGuard in shared mode. (WAIT)indexGuardin shared mode (It goes in the queue) but cannot get it in shared mode (I guess is System dependant if they would schedule a reader to bypass the writer ) potential starvation. Thread N+1 has the GIL.This is solved by releasing the GIL in
Thread N+1before asking forindexGuard.