VMS-to-Unix Phrase Book

VMS-to-Unix Phrase Book

Return

3.6 Trapping Signals

Problem

You want to catch all possible ways a program can exit or die.

Solution

Good luck.

Discussion

Consider the following script:

#!/bin/sh
#
#    Demo a process dying because of an interrupt signal and how
#    there is no chance to even test for an exit status.

     ./bail-out.pl i 8
     echo "exit status is: $?"

#==[ EOF: bail-out-shell-test.sh ]==

When we execute this script, we get:

unix> ./bail-out-shell-test.sh
bail-out.pl: killing myself with '8' signal
unix>

Notice that the echo was never executed. This means the program not only didn't exit with an error status, in the process of dying, it took the script with it. This is because the script was not prepared to trap this interrupt signal thrown by the sample program An example of an interrupt that can kill a program dead would be a division by zero error. This particular sample program was written for Linux, so some changes would be needed to make this program's output text correct for other Unix systems. This is because such signals and their definitions vary from Unix to Unix. This sample script was designed to illustrate how a script can attempt to catch anticipated errors. Again the instructions displayed would need to be changed for systems other than Linux. For example, under Tru-64, the statement

trap 'echo "$0: expected signal";
      final=2;                                  # Our exit code for signals
      exit;                                     # we are expecting.
     ' 3 6 8 11;

generates a scripting error. It turns out that signal 11 (which is the SEGV signal under Tru-64) isn't even trappable so the shell interpreter generates a scan-time error when it sees us trying to trap that particular signal. In other words, on some systems, you may not be able to trap for all signals even if you want to try to.

3.6 Trapping Signals

Problem

Solution

Discussion

See Also