Please see the full report text in this upstream issue: https://phab.enlightenment.org/T7486 The fix is trivial -- add O_CLOEXEC to an open. I thought upstream might fix it, but it does not appear they will. So we might consider patching it in ports. The patch to fix is totally trivial: --- ./eio_monitor_kevent.c.orig 2019-06-17 19:34:47.015700000 -0700 +++ ./eio_monitor_kevent.c 2019-06-17 19:35:04.938381000 -0700 @@ -271,7 +271,7 @@ res = stat(monitor->path, &st); if (res) goto error; - fd = open(monitor->path, O_RDONLY); + fd = open(monitor->path, O_RDONLY | O_CLOEXEC); if (fd < 0) goto error; backend->fd = fd;
Fyi, this was addressed upstream: https://phab.enlightenment.org/rEFL3d5e73570455b19f75c1ba8223956ea15657c009