mjhjmtu

Sometimes it's convenient to use read -t 3 instead of sleep 3. How do I make it work with nohup?

nohup bash -c ' date; read -t 3; date ' | tail -n 2 nohup.out

As you can see, read -t 3 does not wait for three seconds.

read -t 3 (a ksh93 extension now also supported by zsh, bash and mksh) is meant to read one line (logical line in that lines may be continued with a trailing backslash as you don't use the -r option) from stdin into $REPLY with a 3 second timeout.

If stdin is a terminal, that will sleep for 3 seconds unless the user presses enter (and the script will be suspended with a SIGTTIN signal if it's started in background).

If it's a regular file, it will read that line from it and return straight away. If it's /dev/zero it will do a very busy read of gigabytes of zeros from there, etc.

nohup is the command you use to detach a command from a terminal. It redirects stdin to /dev/null and stdout and stderr to nohup.out. So typically you would not want to read from the terminal in that case.

read on /dev/null returns straight away with no data returned, that's what /dev/null is for.

If the purpose of using read -t is to have a kind of sleep that can be interrupted by the user (by pressing Enter) like when you want to give them the time to read a message which they can skip, then having read -t return straight away when non-interactive (like when running under nohup) would seem the right thing to do as there's no point delaying the script then.

But if you want to read from the terminal with timeout if stdin is a terminal, and sleep otherwise, then you would do:

if [ -t 0 ]; then
 read -t 3
else
 sleep 3
fi

[ -t n ] tests whether the file descriptor n (0 being stdin) refers to a terminal device.

You could do read -t 3 < /dev/tty but that would defeat the purpose of nohup by adding back the interaction with the terminal that nohup is meant to guard against.