I have a script I am trying to run with grunt using the grunt-shell plugin. This script prompts for input using read -p "enter foo" bar. When I run it, I get a cursor that correctly takes my input, but I don't see "enter foo".
I tried modifying the grunt-shell config to dump err, stderr, and stdout to the console, but they all come out empty once the command completes. The man page for read says about the -p flag "The prompt is displayed only if input is coming from a terminal.", which seems like it may be my issue since node is calling the prompt, but the read command is executing in my terminal, and I am entering input in a terminal, so why wouldn't that work?
So, my question is, is there any way to prompt for user input from the grunt-shell plugin?
Also, as a P.S., yes i know it would be very easy to just echo the prompt and deal with the input being below it, but dammit it's the principle of the thing.
I redirected my code back to /dev/tty.
Ran it as bash -c run.sh </dev/tty
Related
I have a script that runs one of the following line
sudo -u $USER $SUDOCMD &>>stdout.log
The sudo command is a realtime process that print out lots of stuff to the console.
After running the script each time, the script does not return to the command prompt. You have to press enter or ctrl + c to get back to the command prompt.
Is there a way to to do it automatically, so that I can get a return value from the script to decide whether the script runs ok or failed.
thanks.
What is probably happening here is that your script is printing binary data to the TTY rather than text to standard output/error, and this is hiding your prompt. You can for example try this:
$ PS1='\$ '
$ (printf "first line\nsecond line\r" > $(tty)) &>> output.log
The second command will result in two lines of output, the second one being "mixed in" with your prompt:
first line
$ cond line
As you can see the cursor is on the "c", but if you start typing the rest of the line is overwritten. What has happened here is the following:
You pressed Enter to run the command, so the cursor moved a line down.
The tty command prints the path to the terminal file, something like "/dev/pts/1". Writing to this file means that the output does not go to standard output (which is usually linked to the terminal) but directly to the terminal.
The subshell (similar to running the command in a shell script) ensures that the first redirect isn't overridden by the second one. So the printf output goes directly to the terminal, and nothing goes to the output log.
The terminal now proceeds to print the printf output, which ends in a carriage return. Carriage return moves the cursor to the start of the line you've already written to, so that is where your prompt appears.
By the way:
&>> redirects both standard output and standard error, contrary to your filename.
Use More Quotes™
I would recommend reading up on how to put a command in a variable
I am trying to have my terminal title change depending on what that specific window is doing. So far I have the following which will update based on directory and server.
function settitle() {
if [[ -z "$ORIG" ]]; then
ORIG=$PS1
fi
TITLE="\[\e]2;$*:$(dirs -0)\a\]" #dirs -0 is like pwd but with ~/ instead of /home/user/
PS1=${ORIG}${TITLE}
}
PROMPT_COMMAND="settitle local" # local is the server name in this case
Now, sometimes I'm in the PHP (php -a) or MySQL (mysql -u user -ppass) REPL, and I'd like the title to reflect that instead of just being whatever directory I launched the REPL from.
The best I can figure is getting the last command somehow, then figuring out what the first word is, and running an if check in settitle(). I've tried everything from here and here among other places, and while I can usually get part of it to work in the command line, non of it works in settitle(). For example.
local:~$ echo 'foobar'
foobar
local:~$ !:0
echo
# I add echo !:0 to settitle()
local:~$ source .bashrc
!:0
local:~$
A note: It should be "source", or at least "echo" from before. !:0 does not recognize itself as a command so it will repeat the last real command over and over. The "!:0" being echoed is a literal string, not the results of the command. Additionally, saving to a var does not work, and just putting the command without trying to echo/save the result gives !:0: command not found.
I don't want to make this an XY problem, so if I am barking up the wrong tree here at any step of the process, please let me know. The goal is to be able to change the title of my terminal window if I enter an REPL. How can I identify when a command will enter me into one?
Note that PROMPT_COMMAND and similar shell features are not relevant when you're in a different REPL; from the point of view of the shell, the entire REPL session is one single command. The shell prompt doesn't show up again until you exit that REPL, and that's the point at which PROMPT_COMMAND and friends are activated.
One thing you can do is alias the command you use to start the REPL so that it sets the title of the window first:
alias phpa='setttitle PHP; php -a'
alias mysqli=`settitle MySQL; mysql -u "$USER"'
or something like that.
The sequence goes like this:
PROMPT_COMMAND runs.
The shell prints its prompt.
You type the command to start a REPL
You are in the REPL. The prompt you see is printed by the REPL, not the shell, which is not involved at this point. The shell is just hanging out waiting for you to exit the REPL; it's not printing any prompts, so it's not ever running PROMPT_COMMAND.
You type commands in the REPL. No matter how many you run, it's part of a single session that the shell sees as a single command.
You exit the REPL.
PROMPT_COMMAND runs.
The shell prints its prompt.
I am writing a wrapper shell script wrapper.sh to run bunch of other already available scripts owned by other people and I cannot touch those scripts.
The problem is, there is one script that runs some db specific activities - db_perf_clean.sh. That script is normally executed manually and it prompts for a password at run time. There is no way I can supply the password to it as a parameter and I cannot modify that script. As such I know the db password and I can provide it in wrapper.sh.
Please let me know how can I run that db_perf_clean.sh script inside wrapper.sh like in a silent mode.
Sometimes a script will insist that a password be read from the tty. Often, it will read from stdin. If so, try:
echo password | db_perf_clean.sh
The above has the disadvantage that the password will appear in ps. To avoid that, hide the password in a file and use that file for stdin:
db_perf_clean.sh <file_with_password
If you want the command to be silent, you can throwaway its output:
db_perf_clean.sh <file_with_password >/dev/null 2>&1
Under bash, as opposed to generic shell, that can be slightly simplified:
db_perf_clean.sh <file_with_password &>/dev/null
I found out little different approach instead of writing a password in a file and that worked too ->
db_pass="somevalue"
sh db_perf_clean.sh<<EOM
$db_pass
EOM
How can I open a new terminal from TCL code, do some operations (e.g. ls -l), get the results of those operations and close that terminal?
Does the exec command open a new terminal and all the operations are invoked in the terminal or when I call for example "cd .." with exec, that command has nothing to do with the linux terminal and linux commands, those are just pure tcl commands that have the same name as linux standard commands?
Sounds like you want Expect.
Any command you pass to exec will be sent to the system to be executed. exec does not open a terminal window to do this: it does not need to open a GUI window like a terminal just to interact with the underlying system.
A couple of specific notes about your example commands:
parsing the output of ls or ls -l is not recommended. Suppose you have an odd but valid filename like "foo\nbar". You're better off iterating over the results of Tcl's glob command.
cd happens to be a Tcl command.
I have done my task with this:
set cvsUpdStr [exec $pathToCvsInYourSystem -qn upd]
It does not open a terminal, but it does the task:
executes a command
results is being stored in cvsUpdStr and can be used later
Also it is possible to use it with catch to understand if it was executed correctly or to avoid errors:
if {[catch {exec $pathToCvsInYourSystem -qn upd} result]} {puts $result}
I'm making a shell script and I want to know if it's possible to write directly to the command line when the script is executed ?
Example :
user#localhost:/home/user$./script.sh
... output
... another output
... another output
... last output
user#localhost:/home/user$I want to write here on the command line
I don't want to "echo" some text, I want to write directly at the prompt.
Thanks!
No, you can't do that. If you want user to invoke your provided command after your script is finished - why not just prompt user for confirmation?
If you just want the text to show up there, but not be able to do anything with it, you can do this.
File test.sh:
echo "Output"
./test2.sh &
File test2.sh:
echo "Output2"
Notice how the first script calls the second script with the & at the end.
In this case, "Output2" will be written to the prompt, but it can't be deleted and will have no effect on the next command at all. But if this is something you're doing to grab the user's attention, it would work.
In ksh:
print -s $(script)
will print to the command history. Wrap this in a function and you'll have something close to what you are asking for.
If you are using X environment install xclip and xdotool, then:
#!/bin/bash
your scripts....
echo -n your command to write 2>&1|xclip
xdotool click 2