the PID check was racy, bonus points for not writing data at all to the
lock and not reading it meaning less disk I/O
oh, yes - by using O_CLOEXEC the lock is stale-safe
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
also not storing the host name because it serves no real purpose other than
being a possible source of problem (hostname change?), copyrighting it to
me because there is nothing relevant left of the original implementation
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
