Every now and then I have a need to redirect all output, both standard and error, within a shell script (bash) to a file. There are two ways — the obvious beginners’ one and the elegant gurus’ one. For starters let’s have a very simple script test.sh that generates both standard output (stdout) and error message (stderr):

#!/bin/sh

echo "Going to run /non/existent"
/non/existent

That will produce an expected result when run:

~$ ./test.sh
Going to run /non/existent
./test.sh: line 4: /non/existent: No such file or directory
~$

Now the question is how to redirect both messages into a logfile? The poor man’s approach is to create a wrapper-script test-wrapper.sh:

#!/bin/sh
exec ./test.sh > test.log 2>&1

Running that will send all the output into test.log logfile:

~$ ./test-wrapper.sh
~$ cat test.log
Going to run /non/existent
test.sh: line 4: /non/existent: No such file or directory
~$

Works a treat, but do we really need two scripts to solve such a simple problem? Of course we don’t. Let’s modify test.sh this way:

#!/bin/sh

exec > test2.log
exec 2>&1

echo "Going to run /non/existent"
/non/existent

Run it and enjoy the output redirected to a log file test2.log:

~$ ./test.sh
~$ cat test2.log
Going to run /non/existent
./test.sh: line 7: /non/existent: No such file or directory
~$

Voilà, here you go. Self contained script that sets up the redirection internally. No need for stupid wrappers anymore ;-)