<@U0BME18CF7S> can you please review <github.com/O...
# automations-debug
c
@fresh-machine-37269 can you please review github.com/OpenHands/software-agent-sdk/pull/5559 and feel free to approve, or request changes, or comment, as you think fit.
👀 1
f
🤖 On it! View progress here: oss-agent-canvas.ngrok.dev/conversations/…
✅ Done! Done — I reviewed and approved PR #5559 ✅ It's a clean, well-evidenced fix for the tmux socket-path overflow (#5558). The core check in
_fits_tmux_socket()
uses
os.path.realpath()
, which I confirmed is genuinely necessary: tmux resolves
$TMUX_TMPDIR
before binding, so a short symlink pointing at a deep target really does fail while a deep symlink pointing at a short target works. The 107/103-byte limits are correct (108/104 including the NUL), the
/tmp
fallback fits on both platforms (~87 bytes), and the shutdown path now correctly restores the original value instead of clearing it. Current-head CI is all green (including agent-server, workspace, and Windows tests), and the PR body shows real before/after evidence through the OpenHands live E2E workflow, with the macOS numbers honestly flagged as computed rather than reproduced. I left two non-blocking notes: the
/tmp
fallback isn't itself re-validated, and
AGENTS.md
still describes an explicit
TMUX_TMPDIR
as always honored. Neither affects approval.