I am attempting to programmatically read the output written by an application running in a specific tmux pane, so that I can determine when to send-keys to it from a controlling process.
In particular, I would like to automatically enter a password, but I do not want to enter it until I am sure the password prompt has appeared.
My current attempt has been to use tty to find the controlling tty, and then pass it to pyserial to try to read, since it appears to be able to read tty's. Note that in the real application, I have other ways of finding out the tty.
Unfortunately, as soon as I run the following code, the target tmux pane immediately closes.
import serial
ser = serial.Serial('/dev/ttys013', timeout=1)
Is it possible to read from a pty this way?
I am running on OSX, but would appreciate a solution that works on both Linux and OSX.
Typically, with python you would use pexpect to start the program and interact with it via a pty, but if you already have a program running in tmux you could simply use tmux's pipe-pane command to save a copy in a file of what is written to the screen. For example, for a pane number 1 you can give the shell command:
tmux pipe-pane -t 1 'cat >/tmp/capture'
and then tail the file /tmp/capture. (Use tmux pipe-pane -t 1 to stop).
To avoid polling, you can use a fifo instead.
I have a BASH script that runs on bg and print messages to the user when according to system events.
My problem is that after the echo of these messages, the user need to press on the ENTER key in order to get back to the prompt line.
Adding new line to the printed messages didn't help since it still comes from the bg and not from the user shell.
Does anyone have an elegant & simple solution to get the user back to the prompt line?
I will appreciate any help here..
You don't need to press enter to get to a prompt. You are still at the prompt that was there before the output was printed. Try just entering a command and hitting enter. (You can also hit ctrl-l to clear the screen instead of hitting enter to input an empty command.)
The problem here is that the background/alternative process has no relationship to the running shell session and so it is simply writing output to wherever the terminal sticks it. The process might be able to use control sequences to control the output location (but I don't know if this actually works).
Other than that there isn't much to be done about this that I'm aware of. And it isn't a problem in any real way.
I use screen all the time and am constantly attaching/detaching from different sessions. Occasionally I'll accidentally logout/exit instead of detach, losing the buffer of my work on that project. I'd rather not have the detach (ctl-a ctl-d) and exit (ctl-d) commands be so close.
Is there a way to force screen to only ever detach instead of exit?
The main problem isn't screen, it's the shell inside of it. You can make it ignore Ctrl+D or handle it differently. For BASH, try
export IGNOREEOF=4
which means you'll have to press Ctrl+D 4 times to exit the shell.
See this question for more solutions: https://unix.stackexchange.com/questions/27588/how-can-i-keep-controld-from-disconnecting-my-session
Is it possible to move a process from one terminal to another in Linux? For example, if I have executed the "top" command via a gnome-terminal window, I want to move the output shown in the gnome-terminal window to a TTY terminal.
It's too late by the time you've already executed it, but if you recognise in advance that you might want to move it, you can use
screen top
to run top, and then detach from the running process with Ctrl-A and D. After that, you can run
screen -r
from any terminal on the same machine to reattach.
It's very powerful, and does a lot more than just that.
I am using Putty to connect to a remote server. What I want to know is if there is any way to write my commands and allow them to keep running after I close the session with Putty. The reason for this is that I do not want to keep the computer ON all the time. Is there any way to do this?.
Update with the solution
For my question as it is presented the best solution is use one of the commands provided such as nohup, because you do not have to install any additional software. But if you are in the same problem use screen, install it and use it. It is amazing.
I have selected the answer of Norman Ramsey as favourite because propose several solutions using commands and screen. But please check the other answers specially the one of PEZ, then you get an insight of what screen is able todo.
screen! It's the best thing since sliced bread. (Yeah, I know others have already suggested it, but it's so good the whole world should join in and suggest it too.)
screen is like, like, ummmm ... like using VNC or the like to connect to a GUI destop, but for command shell windows. You can have several shell "windows" open at once in the same screen session. You can do stuff like:
Start a screens session using "screen -dR" (get used to using -dR)
run some commands in one window
press CTRL-A,C to create a new window open a file there in vim
press CTRL-A,0 to go back to the first window and issue some command on the file you just edited
CTRL-A, 1 to go back to your vim session
CTRL-A, C for yet another window and maybe do "sudo - su" (because you just happen to need a full root shell)
CTRL-A, 0 and start a background process
CTRL-A, C to create yet a new window, "tail -f" the log for that background process
CTRL-A, d to disconnect your screen then CTRL-D to disconnect from the server
Go on vacation for three weeks
Log on to the server again and issue "screen -dR" to connect to your existing screen session
check the log in the the fourth window with CTRL-A, 3 (it's like you've been there watching it all the time)
CTRL-A, 1 to pick up that vim session again
I guess you're starting to get the picture now? =)
It's like magic. I've been using screen for longer than I can remember and I'm still totally amazed with how bloody great it is.
EDIT: Just want to mention there's now also tmux. Very much like screen, but has some unique features, splitting the windows being the most prominent one.
nohup, disown, and screen are all good but screen is the best because unlike the other two, screen allows you to disconnect from the remote server, keep everything running, and then reconnect later to see what is happening. With nohup and disown you can't resume interacting.
Try using GNU Screen. It allows you to have several shells open at once. And you can disconnect from those running shells (i.e. close session with Putty) and they will keep doing their thing.
What you are looking for is nohup.
See the wiki link for how to use it.
screen is the best.
Try:
screen -dmS "MyTail" tail -f /var/log/syslog
This start command in background.
Use screen -r to list, and or screen -r Mytail to enter session.
If more users need access same session, use: screen -rx MyTail, and both or more users share the session.
If you can't use screen (because, for instance, your SSH session is being programmatically driven), you can also use daemonize to run the program as a daemon.
One way that works well for me is at.
at works like cron, but for a one-time job. I used it today to download a large file without having to keep my session alive.
for example:
$ at 23:55
at> wget http://file.to.download.com/bigfile.iso
at> ^D
You pass at a time (in the future) and it gives you a prompt. You enter the commands you want to run at that time and hit ctrl+d. You can exit out of your session and it will run the commands at the specified time.
Wikipedia has more info on at.
./command & disown
ssh localhost && ./command && exit