On Mon, Feb 02, 2026 at 09:22:04PM +0800, 王伟 wrote:
> #!/usr/bin/env bash
>
> test_crash ()
> {
> a=$((009+1))
> }
>
> test_crash
> if [[ $? -ne 0 ]] ; then
> echo "case 1"
> fi
>
> if ! test_crash ; then
> echo "case 2"
> fi
>
> test_crash || echo "case 3"
>
> { test_crash ; } || echo "case 4"
>
> exit 0
>
> The code above shows the different results of several methods when a
> function crashes due to a syntax error.
>
> pc@DESKTOP-0MVRMOU UCRT64 /d/code$ bash test_bash_broke.sh
> test_bash_broke.sh: line 5: 009: value too great for base (error token is
> "009")
> case 1
> test_bash_broke.sh: line 5: 009: value too great for base (error token is
> "009")
> test_bash_broke.sh: line 5: 009: value too great for base (error token is
> "009")
> test_bash_broke.sh: line 5: 009: value too great for base (error token is
> "009")
> pc@DESKTOP-0MVRMOU UCRT64 /d/code$
>
> Why is the exception branch executed only in the case of $?, and skipped in
> all other cases?
Because only in the 1st case does the shell set $? to a non-zero exit
status after the function call fails.
In the other cases, the function call itself is part of a conditional
construct, and the failure of the function call causes the conditional
construct to fail.
Note that the functon is never executed successfully. If you inserted
further commands after the line with the arithmetic error, those
commands would never be executed. The function never really returns a
value, because it never completes. In the 2nd, 3rd, and 4th cases, the
shell simply treats the failure of the function call as a failure of the
entire conditional construct.
The above is a bit of handwaving. The manual says:
If expression is invalid, bash prints a message to standard
error indicating failure, does not perform the substitution, and
does not execute the command associated with the expansion.
The "command associated with the expansion" is the function call in the
1st case, and the entire conditional construct in the other cases.
--
Matti Andreas Kähäri
Uppsala, Sweden
.
