Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I have a bash script with lots of output and I would like to change that script (not create a new one) to copy all the output to a file, just like it would happen if I were piping it through tee.

I have a file script.sh:


and I would like to make a copy of STDOUT to a file script.log without having to type ./script.sh | tee script.log every time.

Thanks, Tom

share|improve this question
I am sorry for not saying this earlier, but the script is rather long, which is why I'm looking for a simple solution. Appending | tee to each line is a bit too much. –  Tom Dec 24 '09 at 15:27
add comment

6 Answers

up vote 10 down vote accepted

I couldn't get Dennis' very simple one-liner to work, so here's a far more convoluted method. I'd try his first.

As mentioned, you can use exec to redirect standard error & standard out for the entire script. Like so:
exec > $LOGFILE 2>&1 This will output all stderr and stdout to $LOGFILE.

Now, since you want to have this displayed to the console as well as a logfile, you're also going to have to use a named pipe for exec to write to, and tee to read from.
(Dennis' one-liner technically does this as well, although obviously in a different way) The pipe itself is created with mkfifo $PIPEFILE. Then do the following.

# Start tee writing to a logfile, but pulling its input from our named pipe.

# capture tee's process ID for the wait command.

# redirect the rest of the stderr and stdout to our named pipe.
exec > $PIPEFILE 2>&1

echo "Make your commands here"
echo "All their standard out will get teed."
echo "So will their standard error" >&2

# close the stderr and stdout file descriptors.
exec 1>&- 2>&-

# Wait for tee to finish since now that other end of the pipe has closed.
wait $TEEPID

If you want to be thorough, you can create and destroy the named pipe file at the start and end of your script.

For the record, I gleaned most of this from a random guy's very informative blog post: http://www.travishartwell.net/blog/2006/08/19%5F2220.tbh

share|improve this answer
This doesn't seem to work in cygwin. :-( –  The Doctor What Sep 8 '10 at 17:53
This post does a good job of educating. Thank you very much. –  Felipe Alvarez Jun 24 at 1:58
add comment

Simply add this line at the beginning of your script:

exec > >(tee -a script.log)

That will append each line sent to stdout to the file script.log, leaving the previous contents in place. If you want to start fresh every time the script is run just add rm script.log right before that exec command -or- remove the '-a' from the tee command.

Add this line to also catch all errors:

exec 2> >(tee -a scripterr.out)

** Notice the space between the '>'s above. If the space is missing, there's a good chance you won't get what you want.

share|improve this answer
Very interesting solution. –  ℝaphink Dec 23 '09 at 16:35
This is a really elegant way of doing it, compared to all the crap I posted. But when I create a script with those two lines in it, it hangs, and never exits properly. –  Christopher Karel Dec 23 '09 at 17:00
What version of Bash? –  Dennis Williamson Dec 23 '09 at 17:23
GNU bash, version 3.2.25(1) as part of a RHEL5.3 install. It appears this is the latest version from the official repositories. –  Christopher Karel Dec 23 '09 at 19:15
I just tried it under cygwin in Bash 3.2.49(23)-release and got file descriptor errors. I works in Bash 4.0.33(1)-release in Ubuntu 9.10, though, so it may be a Bash 4 thing. –  Dennis Williamson Dec 23 '09 at 19:55
show 1 more comment

If you want a copy of the output, you can use tee this way:

  do_something_that_outputs_stuff_to_STDOUT | tee script.log
  launch_applications_that_output_to_STDOUT | tee -a script.log

This will only log stdout to script.log however. If you want to make sure both the stderr and stdout are redirected, use:

  do_something_that_outputs_stuff_to_STDOUT 2>&1 | tee script.log
  launch_applications_that_output_to_STDOUT 2>&1 | tee -a script.log

You could even make it a bit nicer with a little function:


  # Wipe LOGFILE if you don't want to add to it at each run
  rm -f "$LOGFILE"

  function logcmd() {
        local cmd="$1" logfile=${LOGFILE:-script.log} # Use default value if unset earlier

        # Launch command, redirect stderr to stdout and append to $logfile
        eval '$cmd' 2>&1 | tee -a "$logfile"

  logcmd "do_something_that_outputs_stuff_to_STDOUT"
  logcmd "launch_applications_that_output_to_STDOUT"
share|improve this answer
Good answer. If he really only cares about stdout, I would just leave the 2>&1 out and just pipe it to tee. –  DaveParillo Dec 23 '09 at 16:14
That's true, I only included stderr because the previous answer from Igor cared to redirect it and I thought it was a good idea. –  ℝaphink Dec 23 '09 at 16:16
add comment

I suppose another approach is something like this:


# start grouping at the top
    # your script goes here

# and at the bottom, redirect everything from the grouped commands
} 2>&1 | tee script.log 
share|improve this answer
add comment

This is combined version of answer posted by Dennis Williamson earlier. Appends both err & std out to script.log in right order.

Add this line to the beginning of your script:

exec > >(tee -a script.log) 2>&1

Note the space between the > signs. Works for me on GNU bash, version 3.2.25(1)-release (x86_64-redhat-linux-gnu)

share|improve this answer
add comment
do_something_that_outputs_stuff_to_STDOUT > script.log 2>&1
launch_applications_that_output_to_STDOUT >> script.log 2>&1

You can see more about how it works here: http://www.xaprb.com/blog/2006/06/06/what-does-devnull-21-mean/

share|improve this answer
I added a second > to the second line in Igor's answer so that it doesn't wipeout the log written in the first line. –  Doug Harris Dec 23 '09 at 15:57
He wants a copy though, so he needs to use tee. –  ℝaphink Dec 23 '09 at 15:59
@Raphink: That's right :) –  Tom Dec 24 '09 at 15:25
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.