[PATCH] shell script error management in ash (set of 6 patches)

[Harald van Dijk]

Hi again,

Another bug: the exit status is not preserved.

Consider

   busybox ash -c 'trap "echo ERR" ERR; false; echo $?'

This prints ERR, and then 0, because the echo command completed 
successfully. It is supposed to print ERR, and then 1, because like EXIT 
actions, ERR actions are supposed to preserve the exit status of the 
last command before the action was invoked.

I am also somewhat concerned that there may be cases where the global 
variables "doingtrap" and "recursive" are set, but evalstring() exits 
via raise_exception() and they are never cleared, preventing future ERR 
handlers from running. I do not currently have a test case where it 
matters, but am not convinced there are no test cases where it matters.

On 21/08/2021 15:36, Roberto A. Foglietta wrote:
> Il giorno sab 21 ago 2021 alle ore 14:38 Harald van Dijk 
> <harald at gigawatt.nl <mailto:harald at gigawatt.nl>> ha scritto:
> 
>     Hi,
> 
>     In bash, the ERR trap is documented as triggering in under the exact
>     same conditions that 'set -e' would cause the shell to abort. This is
>     not what you have implemented, you have implemented it as triggering
>     whenever a simple command returns non-zero. Consider:
> 
>         trap "echo ERR" ERR
>         (:) >/access-denied
> 
>     With bash, this prints:
> 
>         bash: line 2: /access-denied: Permission denied
>         ERR
> 
>     With your patches, this prints:
> 
>         ash: can't create /access-denied: Permission denied
> 
>     That is, the ERR action is not executed when it should be. The command
>     that failed here is not a simple command.
> 
>     Consider:
> 
>         trap "echo ERR" ERR
>         false || true
> 
>     With bash, this prints nothing. With your patches, this prints ERR.
>     That
>     is, the ERR action is executed when it should not be. The command that
>     failed here is one that had its exit status tested.
> 
> 
>   The 10th patch reworks the trap ERR in such a way which is compliant 
> with bash for both these cases.
> 
>   Thank you,
> -- 
> Roberto A. Foglietta
> +39.349.33.30.697