pd

pd

process defer - execute a command after a specified process (by PID) exits

this is similar to the shell's builtin wait, but it can wait for any pid, not just the child pid (also it is similar to waitpid(1) from util-linux, but pd was created before waitpid existed)

NOTE: if no cmd is provided, pd always exits with 0 regardless of the PID exit status, but if a cmd is provided, pd will exit with the exit code of that cmd

usage

pd <pid> [cmd]
        pid     process id to wait for
        cmd     optionall command to run with '/bin/sh' after <pid> exits

# standard wait can only wait for child processes
sleep 10 & wait "${!}"

# pd
pd 1  # wait for init to stop :)

sh ./wait-parent.sh

see also wait-parent.sh

to see what actually happens in this script use htop -t and extrace -t (extrace)

for example it can be used in swallow script:

#!/bin/sh --
set -ue

id="$(xdo id)"
xdo hide "${id}"

ppid="${$}"
# shellcheck disable=2016
id="${id}" pd "${ppid}" 'exec xdo show "${id:?}"' &

exec "${@}"

run-time dependencies

• x86_64 linux
  • pidfd_open syscall support (added in linux-5.3)

how to build

build-time dependencies:

• fasmg
  • tested with version kd3c
• copy of examples/x86/include from fasmg source
  • see: include/fasmg/

make FASMG=./path/to/fasmg-executable

this will create a static pd executable of 1469 bytes in size and without any run-time dependencies other than the Linux kernel