Re: Numbered signals for shell trap cause z/OS grief

[Bruno Haible]

I wrote:
> In fact, such a warning would make sense also in bash on Linux, for signals
> 7, 10, 12, and ≥ 16.

Thinking more about it: Such a warning in the *shell* is misplaced, because
it's perfectly OK to auto-generate a shell script for a particular machine
that runs on a particular platform. What is wrong is when a programmer
writes such a script and wants to it portably. Thus, the right place where
to produce this warning is a developer tool such as shellcheck.

I was just about to report it to the shellcheck people, when I saw that they
have implemented it already:

======================== foo.sh ========================
#!/bin/sh

trap 'rm -f foo' 0 1 2 3 7 10 12 15 17
========================================================
$ shellcheck foo.sh

In foo.sh line 3:
trap 'rm -f foo' 0 1 2 3 7 10 12 15 17
                         ^-- SC2172 (warning): Trapping signals by number is 
not well defined. Prefer signal names.
                           ^-- SC2172 (warning): Trapping signals by number is 
not well defined. Prefer signal names.
                              ^-- SC2172 (warning): Trapping signals by number 
is not well defined. Prefer signal names.
                                    ^-- SC2172 (warning): Trapping signals by 
number is not well defined. Prefer signal names.

For more information:
  https://www.shellcheck.net/wiki/SC2172 -- Trapping signals by number is not...

They warn about signal 7, 10, 12, 17, but not about 0, 1, 2, 3, 6, 9, 14, 15.

Mike Fulton wrote:
> a very subtle potential problem for z/OS.

Yes.
<https://man7.org/linux/man-pages/man7/signal.7.html>
<https://www.ibm.com/docs/en/zos/2.2.0?topic=SSLTBW_2.2.0/com.ibm.zos.v2r1.bpxb100/signals.html>

You can convince the shellcheck people to warn also about 'trap' of signal 3
(SIGQUIT vs. SIGABRT) and signal 6 (SIGABRT vs. SIGURG). Their bug tracker is
here: https://github.com/koalaman/shellcheck/issues

Bruno