Fixup callbacks deadlock when calling callback very frequently #158
Conversation
This file contains 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
A minimal test which exhibits a deadlock.
On my machine, I can get it to deadlock about 25-35% of the time.
This is with the script:
```sh
fixture=fixtures/callbacks-regression
for i in {1..50} ; do
echo
echo "Starting $i"
cargo xtask run \
--crate $fixture \
--ts-dir $fixture/generated \
--cpp-dir $fixture/generated/cpp \
$fixture/tests/bindings/test_callbacks_regression.ts
done
```
Counting the incidence of `timed out` in the output and multiplying by 2.
Increasing the work done in the callback (e.g. by uncommenting the `console.log`) gave this error. I'm fairly sure that this is a separate error, and likely an artifact of the test-harness.
```sh
-- Going to 1024, now at: 264
-- Going to 1024, now at: 265
Assertion failed: (usedDbg_ && "Stats not submitted."), function ~CollectionStats, file HadesGC.cpp, line 259.
```
This error is the reason why I can't make a test case to reproduce on CI.
For now, I'm going to treat this error as different to the deadlock.
The change here is in `UniffiCallInvoker`, reverting the implementation of `invokeBlocking` back to the Promise based implementation introduced in PR #88. The deadlock rate for the test is now 0/50. Adding any more to the implementation (e.g. uncommenting the `console.log`) causes the same error as before, but it is completely deterministic: ```sh -- Going to 1024, now at: 264 -- Going to 1024, now at: 265 Assertion failed: (usedDbg_ && "Stats not submitted."), function ~CollectionStats, file HadesGC.cpp, line 259. ``` On my machine it _always_ stops after 265/1024. I am still unsure if this is a problem with production uniff-bindgen-react-native code, an artifact of the test-harness or something weirder. I'm going to suggest that this is addressed in a separate issue.
jhugman
force-pushed
the
jhugman/callbacks-deadlock
branch
from
2291bbe to
7f61add
Compare
|
This seems like the issue I had. I experienced this both in release builds (XCode archive and run on devices from TestFlight) and locally with a debug build run directly on my iPhone 10s test device. For reference, this is the stack trace of the hung thread for me. |
zzorba
approved these changes
Nov 12, 2024
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.
Fixes #157
Minimal test: 146cd39
The change here is in
UniffiCallInvoker, reverting the implementation ofinvokeBlockingback to the Promise based implementation introduced in PR #88.The deadlock rate for the test is now 0/50.
/cc @tayrevor
Adding any more to the implementation (e.g. uncommenting the
console.log) causes the same error as before, but it is completely deterministic:On my machine it always stops after 265/1024. I am still unsure if this is a problem with production uniff-bindgen-react-native code, an artifact of the test-harness or something weirder.
I'm going to suggest that this is addressed in a separate issue.