On Jun 11, 2025 / 08:51, Bart Van Assche wrote:
> On 6/10/25 10:29 PM, Shinichiro Kawasaki wrote:
> > On Jun 10, 2025 / 08:49, Bart Van Assche wrote:
> > > Here is an example that shows how a subshell can be used to halt a test
> > > with "set -e" if a failure occurs in such a way that error handling is
> > > still executed:
> > >
> > > $ bash -c '(set -e; false; echo "Skipped because the previous command
> > > failed"); echo "Error handling commands outside the subshell are still
> > > executed"'
> > >
> > > Error handling commands outside the subshell are still executed
> >
> > Yes, I understand it. I assume that your idea is to ask test case authors to add
> > the subshells in test cases when they want to do "set -e", right? My question is
> > how to ensure that the "set -e" is done only in the subshells. I think we need
> > to rely one code reviews. If "set -e" out of subshells are overlooked in the
> > reviews, the impact for the following test cases are left. Maybe this is not a
> > big risk and we can take it, but I wanted to know what you think about it.
>
> Can this concern be addressed by adding an unconditional set +e in
> _call_test after the $test_func call and before the _cleanup call?
I'm afraid, no. I'm thinking about this kind of test cases:
- it uses "set -e"
- it does not use subshell
- it has clean up code in test() or test_device()
('IOW, it does not register cleanup handler by _register_test_cleanup())
For such test cases, the clean up code in test() or test_device() can be skipped
by error exit by "set -e" regardless of the "set +e" in _call_test().
