Flaky test: process_group_cleanup reads the launcher pidfile between create and write (ParseIntError Empty)

[schickling-assistant]

Problem

codex_app_server::tests::process_group_cleanup_reaps_a_native_launcher_descendant is flaky under CI load:

thread '...' panicked at src/codex_app_server.rs:5365:
called `Result::unwrap()` on an `Err` value: ParseIntError { kind: Empty }

The descendant writes its pid with printf "%s" "$$" > "$DESCENDANT_PIDFILE". The readiness loop waits on descendant_pidfile.is_file(), which becomes true when the shell creates the file — before any bytes are written. Reading at that moment yields an empty string and the parse unwrap panics.

Observed once on run 32759546875 (PR #335); identical code passed runs 32759495299 (#333) and #334's run, confirming a race rather than a regression.

Fix

Wait for content, not existence: poll read_to_string until it yields a parseable non-empty value instead of gating on is_file().

Posted on behalf of @schickling

field	value
agent_identity	unknown
agent_persona	generalist
agent_supervisor	unavailable
agent_tool	OMP
agent_tool_version	18.0.3
agent_runtime	OMP 18.0.3
tooling_profile	dotfiles@f33cd9c-dirty