mjhjmtu

I am scripting the installation of software on HP-UX server. Once the script starts it provides a prompt where I am to enter the install path. I need to pass the path to the script so it can continue to run. There is only 1 place in the script where this need exists.

The prompt from the script is:
Press ENTER for default path or enter path to install software:

I do not want to use the default path, so I must enter a new path. But this script will run in the background and I need to provide the path. I'm not sure of the exact response in script form.

If you can provide all the inputs when the script starts, then do so, by redirecting the program's input. That is, instead of running /path/to/installer, run

 echo '/the/path/where/to/install';
 echo 'answer to the second prompt';
 | /path/to/installer

or use a here document:

/path/to/installer <<'EOF'
/the/path/where/to/install
answer to the second prompt
EOF

If you want to interact with a program from time to time but use your terminal for other things in between, run the program in a terminal multiplexer such as Screen or tmux. With screen, start a session by running screen, then start the program. To do something else, press Ctrl+A, c to create a second window, then Ctrl+A, n to navigate between the windows. To exit Screen but leave the program running, press Ctrl+A, d (“detach”). To come back to the existing Screen session, run screen -rd (screen with no option would start a new session).

You can echo your answer into standard input of script by using a pipe.

echo "My/Path/not/default"| yourscript.sh

To provide an automatic answer, you could use one of the following:

insaller.sh < an_input_file

or

command-line | installer.sh

There is something to notice if the installer.sh script is using read -p, as in the example below:

read -p "Press ENTER for default path or enter path to install software:" answer

man bash specifies that nothing is printed if the standard input is not a terminal.

If this is your situation, then you could try this odd thing:

( sleep 30 ; printf "/my/own/pathn" ) | insaller.sh

You should adapt the number of seconds (30 in above example) to your situation.

If it happens that read -p is not used inside the install script, then you could give a try to this GNU solution:

 tempdir="$(mktemp -d)"
 mkfifo "$tempdir"/input
 touch "$tempdir"/output.log
 ./installer.sh <"$tempdir"/input >"$tempdir"/output.log 2>&1 &
 installerpid=$!
 tail --pid=$installerpid -fn 1 "$tempdir"/output.log | ( fgrep -q "Press ENTER for default path or enter path to install software:"; printf "/new/pathn" ) >> "$tempdir"/input &

 # ... do stuff


 # before ending the script, just wait that all background processes stop
 wait
 rm -f "$tempdir"/input "$tempdir"/output.log

The idea is to use 2 background command-lines, one for the install script, and one to wait the prompt and provide the answer.

A named pipe (input) and a regular file (output.log) are used for communication.

tail --pid=$installerpid -fn 1 "$tempdir"/output.log prints lines as there are written in the output.log file. It teminates when the installer script terminates.

( fgrep -q ... ; printf .. ) >> ...input: blocks until the prompt is found, and provides the new path to the install script.

Gilles solution works! You can add 'sleep' to wait for another prompt if needed, before 'echo'. Not as clean as I expected, but it works that matter. I can used my python subprocess to wrap => to copy a shell.script in remote servers and execute it from there using shell ssh command. Then the shell script will do the installation and provide input for the prompt.

NICE.

Thanks Gilles,