ding::tests::deferred_delivery_backoff_bounds_short_lived_pty_attempts fails on macOS on plain main, with no branch or patch applied.
What I ran
On origin/main at e0a17d9, on Darwin 25.6.0, Apple Silicon:
- Isolated,
cargo test -p st2 --lib -- --exact ding::tests::deferred_delivery_backoff_bounds_short_lived_pty_attempts: 3 failures in 3 runs.
- Full library suite,
cargo test -p st2 --lib: 2 failures in 2 runs, and it is the only failing test both times.
So it is not flaky and not a parallelism artefact. It fails every time, alone or with the suite.
The assertion
assertion `left == right` failed: an unsafe composer must not respawn pty peek/send children every inbox poll
left: 0
right: 1
The test expects one spawn and observes none.
Why nobody has seen it
CI is ubuntu-latest only, so every green check on this repository proves Linux and says nothing about Darwin. I found this while running a branch locally, not from a failing check.
Why it is worth fixing
Part of this fleet runs macOS, and this is the delivery path. I do not know yet whether the test encodes a Linux-specific assumption or whether the product behaves differently on Darwin. Those have very different consequences, and I am not going to guess which it is.
I am not proposing a fix here, because I did not write this code. Reporting it so it is visible.
Related: I found two more Darwin-only failures in codex_app_server while folding #254, but those are on that branch rather than on main, and hetz.st2 is working them.
ding::tests::deferred_delivery_backoff_bounds_short_lived_pty_attemptsfails on macOS on plainmain, with no branch or patch applied.What I ran
On
origin/mainate0a17d9, on Darwin 25.6.0, Apple Silicon:cargo test -p st2 --lib -- --exact ding::tests::deferred_delivery_backoff_bounds_short_lived_pty_attempts: 3 failures in 3 runs.cargo test -p st2 --lib: 2 failures in 2 runs, and it is the only failing test both times.So it is not flaky and not a parallelism artefact. It fails every time, alone or with the suite.
The assertion
The test expects one spawn and observes none.
Why nobody has seen it
CI is
ubuntu-latestonly, so every green check on this repository proves Linux and says nothing about Darwin. I found this while running a branch locally, not from a failing check.Why it is worth fixing
Part of this fleet runs macOS, and this is the delivery path. I do not know yet whether the test encodes a Linux-specific assumption or whether the product behaves differently on Darwin. Those have very different consequences, and I am not going to guess which it is.
I am not proposing a fix here, because I did not write this code. Reporting it so it is visible.
Related: I found two more Darwin-only failures in
codex_app_serverwhile folding #254, but those are on that branch rather than onmain, andhetz.st2is working them.