glm-5.3-flash db0b1fce5c fix(local): PTY waiter reaps via libc::waitpid — ADR-004 exit encoding holds on the PTY path ...

- portable-pty 0.9.0's ExitStatus flattens a signal-terminated child to
  code 1 with the signal only as a strsignal name string, so a SIGTERM
  death reported {"code":1} — indistinguishable from a real exit 1 and
  a violation of ADR-004 decision 1's negative-code convention (review
  #004, consumer input from alkssh)
- on Unix the waiter now skips the portable-pty conversion entirely:
  libc::waitpid(pid) on the direct child, decoded WIFEXITED/WEXITSTATUS
  and WIFSIGNALED/-WTERMSIG (numeric, locale-independent); Child::wait()
  remains as the waitpid-failure fallback and the non-Unix primary
  (std::process::Child's Drop neither kills nor reaps, so the two paths
  never double-reap)
- signal features now round-trip: signal("TERM") -> -15, ADR-056 cancel
  kill (SIGHUP) -> -1 (the documented ADR-004 sentinel collision)
- tests: SIGINT -> -2, SIGTERM -> -15, SIGHUP fallback -> -1 pinned
- docs: ADR-004 decision 1, tty-local.md, ADR-005, tty-adapter.md,
  review #004 closed as resolved with disposition

Verification: cargo test (113), cargo test --all-features (138 + the
tests/pty.rs, tests/pipe.rs, tests/negotiation.rs integrations),
clippy --all-targets --all-features -D warnings, fmt, doc,
fuzz/shared corpus replay (9)

2026-11-01 12:06:23 +00:00

..

001-code-review.md

docs: add review #002 (follow-up on the 2026-09-05 resolution session)

2026-09-05 08:08:22 +00:00

002-post-session-review.md

feat: InvalidParams variant + #[non_exhaustive] TtySessionError (R5)

2026-09-05 08:45:58 +00:00

003-prepublish-review.md

docs: record v0.1.0 publish in review #003 (review closed)

2026-09-05 17:02:12 +00:00

004-pty-exit-signal-fidelity-alkssh-input.md

fix(local): PTY waiter reaps via libc::waitpid — ADR-004 exit encoding holds on the PTY path

2026-11-01 12:06:23 +00:00