Test failure on ARM32

[mazong1123]

Latest CI failed on https://github.com/microsoft/injectorppforrust/actions/runs/17475472445/job/49634194658

Error message:

     Running tests/async.rs (target/thumbv7neon-unknown-linux-gnueabihf/debug/deps/async-e2d3c686c167d777)

running 3 tests
error: test failed, to rerun pass `--test async`

Caused by:
  process didn't exit successfully: `/home/runner/work/injectorppforrust/injectorppforrust/target/thumbv7neon-unknown-linux-gnueabihf/debug/deps/async-e2d3c686c167d777 --nocapture` (signal: 11, SIGSEGV: invalid memory reference)
Error: Process completed with exit code 101.

[Feiryn]

I just saw this issue, sorry for the bug.

I currently can't reproduce the crash locally after 100 tests (but without recompiling), and the CI/CD crashed at the 3rd compilation (Feiryn#3).

I guess there is something wrong with where memory is placed at compilation time. I started 100 cargo clean + cargo test and I'll see if I can reproduce it locally

[Feiryn]

I can't make it crash locally even after a lot of compilation / tests.

On my CI/CD test PR, I published the artifacts and compared the files between a successful test and an unsuccessful one, and both files are exactly the same. So I guess it's not a compilation issue

[mazong1123]

The test is flakey so looks like the way to patching the memory somehow could resulting the undefined behavior in this target. Not sure if there's any special setting for thumbv7neon-unknown-linux-gnueabihf target.

One thing I'm suspecting is multi-threading. Have you tried modifying the Ci/CD to use single thread to see if this issue is gone? But not sure why only this target has the problem.