I use the script command to save the whole session, but the output in the text file is something weird, and I opened it with the cat command in the terminal, and everything became ok. How can I save it in a text file that becomes readable? This also happened with the watch command. How to fix it? The following commands (individually) result in a.txt containing weird text:

watch date | tee a.txt

script a.txt ...

I'm currently using GPUs on an Ubuntu server to run machine learning models. I often like to use the watch command in order to monitor the GPU statuses like watch -n 1 nvidia-smi. It usually works fine, but I noticed that for one of the servers I get the error:

watch: /usr/local/lib/libncursesw.so.6: no version information available (required by watch)
watch: /usr/local/lib/libncursesw.so.6: no version information available (required by watch)
Segmentation fault (core dumped)

I've tried sudo apt-get install libncursesw6 and get libncursesw6 is already the newest version (6.2-0ubuntu2).. Is there anything else I can do to fix this error? ### Edit The output of lddtree $PATH_TO_WATCH is:

watch => /usr/bin/watch (interpreter => /lib64/ld-linux-x86-64.so.2)
    libncursesw.so.6 => /usr/local/lib/libncursesw.so.6
        libtinfow.so.6 => /usr/local/lib/./libtinfow.so.6
    libtinfo.so.6 => /usr/local/lib/libtinfo.so.6
    libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6
        ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2

When I try to use Linux watch with a command which returns color output, I get the error: segmentation fault (core dumped). The command itself is the following: watch --color --no-title --interval 1 ls\ --color If I remove either --no-title or --color (from watch or ls), the error doesn't show up. What can be the reason of the error? Arch; watch from procps-ng 3.3.11; zsh 5.1.1; GNU bash 4.3.4.2

I have an account on a compute cluster which uses the SLURM scheduler. I have some jobs in the queue and I'm using the "watch" command to see their status: watch squeue -u myUserName Does constantly running this command put any significant stress on the scheduler?

The setup is I have two folders now under my current directory, busybox-1.36.1 and procps-4.0.4, and I also have a service languagetool running with DynamicUser=yes. Now observe the different behaviour: Legend: ✅ Doesn't truncate; ❌ truncates. ### busybox watch & busybox ps ✅

./busybox-1.36.1/bin/watch './busybox-1.36.1/bin/ps aux | grep languagetool'

Every 2.0s: ./busybox-1.36.1/bin/ps aux | grep languagetool                                                                           2025-03-17 23:20:02

1026972 61534    39:36 java ...languagetool...
4053857 mathiass  0:00 ./busybox-1.36.1/bin/watch ./busybox-1.36.1/bin/ps aux | grep languagetool
4053882 mathiass  0:00 sh -c -- ./busybox-1.36.1/bin/ps aux | grep languagetool
4053884 mathiass  0:00 grep languagetool

### busybox watch & procps ps ✅

./busybox-1.36.1/bin/watch './procps-4.0.4/bin/ps aux | grep languagetool'

Every 2.0s: ./procps-4.0.4/bin/ps aux | grep languagetool

languag+ 1026972  0.1 13.6 14468788 4464816 ?    Ssl  Mar04  39:38 java ...languagetool...
mathias+ 4057218  0.0  0.0   4676  2244 pts/32   S+   23:35   0:00 ./busybox-1.36.1/bin/watch ./procps-4.0.4/bin/ps aux | grep languagetool
mathias+ 4057222  0.0  0.0 231736  3788 pts/32   S+   23:35   0:00 sh -c -- ./procps-4.0.4/bin/ps aux | grep languagetool
mathias+ 4057224  0.0  0.0 230732  2520 pts/32   S+   23:35   0:00 grep languagetool

### procps watch & procps ps ❌

./procps-4.0.4/bin/watch './procps-4.0.4/bin/ps aux | grep languagetool'

Every 2.0s: ./procps-4.0.4/bin/ps aux | grep languagetool

mathias+ 4056463  0.0  0.0 230836  3320 pts/32   S+   23:31   0:00 ./procps-4.0.4/bin/watch ./procps-4.0.4/bin/ps aux | grep languagetool
mathias+ 4056540  0.0  0.0 230836  1532 pts/32   S+   23:32   0:00 ./procps-4.0.4/bin/watch ./procps-4.0.4/bin/ps aux | grep languagetool
mathias+ 4056541  0.0  0.0 231736  3556 pts/32   S+   23:32   0:00 sh -c -- ./procps-4.0.4/bin/ps aux | grep languagetool
mathias+ 4056543  0.0  0.0 230732  2244 pts/32   S+   23:32   0:00 grep languagetool

### procps watch & busybox ps ✅

./procps-4.0.4/bin/watch './busybox-1.36.1/bin/ps aux | grep languagetool'

Every 2.0s: ./busybox-1.36.1/bin/ps aux | grep languagetool                                                                nixos: Mon Mar 17 23:33:31 2025

1026972 61534    39:38 java ...languagetool...
4056826 mathiass  0:00 ./procps-4.0.4/bin/watch ./busybox-1.36.1/bin/ps aux | grep languagetool
4056831 mathiass  0:00 ./procps-4.0.4/bin/watch ./busybox-1.36.1/bin/ps aux | grep languagetool
4056832 mathiass  0:00 sh -c -- ./busybox-1.36.1/bin/ps aux | grep languagetool
4056834 mathiass  0:00 grep languagetool

## Question It seems odd that in every other case the commands behave as I expect them to, except when trying to use both ps and watch from procps. Does anybody know what is going on here?

What am I doing wrong setting up following watch command? This works: ls -ltr ~/meteofetcher/data/metar/2024/12 | tail -10 | awk '{ print $9 }' | xargs -I {} grep 'LSZH' {} But this watch "ls -ltr ~/meteofetcher/data/metar/2024/12 | tail -10 | awk '{ print $9 }' | xargs -I {} grep 'LSZH' {}" only lets me watch the following error: Every 2.0s: ls -ltr ~... : Fri Dec 20 06:27:05 2024 grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. grep: invalid option -- '-' Usage: grep [OPTION]... PATTERNS [FILE]... Try 'grep --help' for more information. It's probably something to do with the quotation but I tried to do t he same with single quotes outside and double quotes inside and it still doesn't work. This produces the same error (and it shouldn't expand the $ afaik) watch 'ls -ltr ~/meteofetcher/data/metar/2024/12 | tail -10 | awk "{ print $9 }" | xargs -I {} grep "LSZH" {}' Where did I go wrong?

I want use shell and see watch on top parallel without switching to another tty. Is theres way to do that?

I usually use watch Linux utility to watch the output of a command repeatedly every n seconds, like in watch df -h /some_volume/. But I seem not to be able to use watch with a piped series of command like: $ watch ls -ltr|tail -n 1 If I do that, watch is really watching ls -ltr and the output is being passed to tail -n 1 which doesn't output anything. If I try this: $ watch (ls -ltr|tail -n 1) I get $ watch: syntax error near unexpected token `ls' And any of the following fails some reason or another: $ watch <(ls -ltr|tail -n 1) $ watch < <(ls -ltr|tail -n 1) $ watch $(ls -ltr|tail -n 1) $ watch ls -ltr|tail -n 1) And finally if do this: $ watch echo $(ls -ltr|tail -n 1) I see no change in the output at the given interval because the command inside $() is run just once and the resulting output string is always printed ("watched") as a literal. **So, how do I make the watch command work with a piped chain of commands [other that putting them inside a script]?**

I want to do some simple computation of the number of lines per minute added to a log file. I also want to store the count for each second. What I need is the output of the following command as a list which will be updated every second: watch -n1 'wc -l my.log' How can I output the 'update' of the 'watch' command as a list?

I like the watch command, but it has [its limitations](https://unix.stackexchange.com/questions/3842/how-can-i-scroll-within-the-output-of-my-watch-command) . I'm curious to know whether I could mimic the functionality of watch with less. I'm mainly looking for the ability to scroll through my directory as it dynamically gets modified via a running script.

What I'd like to have: Running a command y with watch command and have watch command update only as soon as output of command y has changed. Imagine a bash script with 3 lines of echo output

Hello world!
Hello universe!
Aloha multiverse!

I want the bash script to be watched, but updates only should happen, if said output has changed due to editing of the script and saving it afterwards, so, modification date. I fear, that watch command as usual, updating every 2 seconds by default, when being run multiple hours daily, will run down storage disk life time too fast. I mean, just imagine how many read/write operations just that command produces. I feel like this is so wasteful. It would be much better, if watch command just sleeps, and the act of saving the watched command wakes watch up, running the script once again and then sleeps again. What I mean is: The automatic equivalent of doing that handwork we all know: 1. Writing on a script, 2. then saving it, 3. then executing it once again to see if the saved changes work out as intended - repeat. I mean for seeing the change, we don't execute the command every 2 seconds, but only once, event-based. Is the price of automating that tedious evergoing switch between writing a script and executing it unavoidable evergoing interval update wearing down storage disk life by exessive growth of read and write processes? A compromise could be to highen the interval time by -n option, for example up to every 30 seconds, but can't watch command somehow do any better? I mean, I've already thought of maybe letting watch exit by -g option as soon as the file changed, but that doesn't change a thing at watch causing read operations every 2 seconds or other interval until the change has happened, keeping the storage disk busy all the time. And while loops have the same problem, don't they? Or could this be the solution: Kind of rebuilding watch command with a while loop, making it truely event-based somehow (?), and so minimizing the amount of read and write operations while developing a script? This is about efficiency, minimalism, saving ressources as much as possible.

I want to constantly see last 30 lines of code "nfc list ruleset" in Debian Bash. Something like this: watch -n 1 nft list ruleset | tail -n 30 But above code doesn't show last 30 lines of "nfc" command output, it shows blank screen.

When I use watch to follow a perlpod documentation i am writing, the formatting gets mangeled (see NAME and Synopsis sections): Every 2,0s: perldoc lib/Uki/Handlers/SOS.pm Tue Apr 8 10:42:43 2014 Uki::Handlers::SOS(3) User Contributed Perl DocumentationUki::Handlers::SOS(3) N NA AM ME E Uki::Handlers::SOS - Studenten-Handler S Sy yn no op ps si is s SOS-spezifische Spezialisierung des Uki::Handlers::Person-Handlers. I think this is due to nroff thinking the output terminal is not a TTY, but I don't know how to tell it that, infact, it _is_ a TTY.

I have a command (phpunit) which has a colored output. According to the watch, command I should be able to use the --color flag to allow color rendering to pass through. However, this isn't working. Is there any other ways of solving this?

Say I have a function f defined in my bashrc

function f() {
  date
}

I want to run following command to monitor the output watch f. The command failed with "**sh: f: command not found**". watch bash -c f gives the same. How to get this watch command working as intended?

Somehow I'm unable to get this working with color output, it works but no colors are shown. --- 1. I'm currently using these alias'es in my bash shell:

alias Xterm='xterm -geometry 132x60 +dc +cm -e '
   alias Watch='watch --color --differences --interval=1 --exec '
   alias xterm-watch='Xterm Watch '
   alias _nc='networkctl'
   alias nc-status='_nc --no-pager --stats --full status'
   alias nc-status-watch='xterm-watch nc-status'

1. $TERM is set to xterm-256color 1. When i enter nc-status-watch at the bash prompt, an xterm window pops-up correctly with a watch command running the networkctl command with it's arguments. But the output does not show any colors. nc-status-watch" class="img-fluid rounded" style="max-width: 100%; height: auto; margin: 10px 0;" loading="lazy"> How can I get colored output of systemd commands run with watch inside xterm?

I tried

'sensors | sudo docker stats'

'docker stats | watch sensors'

docker stats | sudo watch sensors

So I assume its not possible to display both outputs in one window? Other than installing something like glances? Any advice would be appreciated.

I'd like to monitor apps that have inotify instance in real-time with watch using something similar to:

watch "for foo in /proc/*/fd/*; do readlink -f $foo; done | grep inotify | sort | uniq -c | sort -nr"

I was hoping the output would appear similar to

for foo in /proc/*/fd/*; do readlink -f $foo; done | grep inotify | sort | uniq -c | sort -nr

but instead this fails with a message about readlink having a missing an operand. I've been using using quotes to wrap my complex shell commands to view in watch for a while now and I guess I was surprised this one didn't work. I guess watch is designed for something different, just running a command and updating the output it sees. But I can't seem to bring this command up in sh or bash with:

watch "/bin/bash for foo in /proc/*/fd/*; do readlink -f $foo; done | grep inotify | sort | uniq -c | sort -nr"
watch bash "for foo in /proc/*/fd/*; do readlink -f $foo; done | grep inotify | sort | uniq -c | sort -nr"
watch sh "for foo in /proc/*/fd/*\; do readlink -f $foo\; done | grep inotify | sort | uniq -c | sort -nr"
watch bash "for foo in /proc/*/fd/*\; do readlink -f $foo\; done | grep inotify | sort | uniq -c | sort -nr"

but this gives sh: 1: Syntax error: "do" unexpected, and/or various syntax errors. After trying this one last thing (escaping the ; with \;) but I get /bin/bash: for: No such file or directory. but am curious to know why this works and if I can avoid using a script file, eg: watch which-programs-created-inotify-instances.sh

I have a watch on jq -r '.' $(ls -t | head -1)" The goal is to apply "jq -r '.'" to the most recently modified file in directory. Unfortunately, the command used to know the last modified file is evaluated when watch is launched, but not at refresh. How should i write it, so that all of the command is reevaluated at each refresh ? Thanks

I'm currently executing the following:

watch -n 100 'sh script.sh >> /path/to/output/output.txt' &

It's been running for 30 minutes (I can verify with ps -ef | grep watch), however there's nothing being written to /path/to/output/output.txt (the file doesn't exist). script.sh executes commands that prints to terminal, so shouldn't the >> be directing everything to output.txt?