Questions regarding this column should be sent to the author at ray@cse.ucsc.edu.

Can A Pipe Have More Than One Exit?

Question: I was a subscriber of UnixWorld and now I am a reader of UnixWorld Online. I'm a Korn shell user and I have a problem in the execution of the following pipe:

 
command1 | command2

How do I save the execution status of the first command (command1) in the pipe line? I know that the exit status, after the execution of a command is saved in $?, but in the previous expression this variable takes the exit status of the second command in the pipe line (command2).

Answer: You are correct that the exit status of a pipeline is reported as the exit status of the last command in the pipeline. The exit status is reported by the shell as it executes the pipeline. It is the shell that stores the pipe's exit status in the $? variable.

What happens to the exit status of the first (or other) command? It is ignored by the shell. However, you should be able to capture it with a little ingenuity.

A command's exit status is the value returned to the process that called it. A command can only have one exit status because only a single integer is reserved by the UNIX kernel for a process's exit status. Because a pipeline has, at least, two commands, the designers of the UNIX system had to decide which exit status to report back to the shell for the command that implements a pipeline. The one they chose was the exit status of the command at the tail of the pipeline.

You can capture the exit status of the first (or other) command in a pipeline by running that command inside a shell script whose sole task is running the command and then reporting the exit status of the command executed.

Here is the code for one such script, which we'll name exit.report:

#!/bin/ksh
"$@"
echo "Exit status of command ("$@") was " $? >&2

This script executes the command line specified on its invocation command line and reports the command exit status to the standard error file to separate it from standard output, which could be redirected to a file or into a pipe.

The "$@" construct represents all the command line arguments enclosed in double quotes, so that substitution is performed, but the result is protected from further processing by the shell. The >&2 construct, which is shorthand for 1>&2, causes the output of echo to sent to standard error (which is associated with the terminal).

Here the exit.report script has been passed a simple command line (not a pipe line):

 
$ exit.report cat exit.report
#!/bin/ksh
"$@"
echo "Exit status of command ("$@") was " $? >&2
Exit status of command (cat exit.report) was 0
$ []

which it executes and then displays its exit status (0 in this example) because the cat command succeeded.

Here is an example with a pipeline:

$ exit.report cat exit.report | wc
Exit status of command (cat exit.report) was 0
3 12 64
$ []

This command line sends the output of the cat exit.report command to wc via the pipeline to give the count report. The exit status output from exit.report is sent to the standard error, not into the pipeline, so it is displayed on the terminal before the output from wc.

Here is another example where the command at the head of the pipeline fails:

$ exit.report grep Answer exit.report | wc -l
exit.report grep Answer exit.report | wc -l
Exit status of command (grep Answer exit.report) was 1
0
$ []

exit.report also works in the middle of a pipeline:

$ cat exit.report | exit.report grep Answer | wc -l
Exit status of command (grep Answer) was 1
0
$ []