Re: "wait" loses signals

[Harald van Dijk]

On 19/02/2020 20:30, Chet Ramey wrote:
> On 2/19/20 5:29 AM, Denys Vlasenko wrote:
> > A bug report from Harald van Dijk:
> >
> > test2.sh:
> > trap 'kill $!; exit' TERM
> > { kill $$; exec sleep 9; } &
> > wait $!
> >
> > The above script ought exit quickly, and not leave a stray
> > "sleep" child:
> > (1) if "kill $$" signal is delivered before "wait",
> > then TERM trap will kill the child, and exit.
>
> This strikes me as a shaky assumption, dependent on when the shell receives
> the SIGTERM and when it runs traps. (There's nothing in POSIX that says
> when pending traps are processed. Bash runs them after commands.)

The bash documentation says traps will not be executed until the command 
completes if it receives a signal while waiting for the command to 
complete, but it does not say the same for when it receives a signal 
before waiting for a command to complete. This may be an oversight in 
the documentation.

POSIX says in the description of the trap command "Otherwise, the 
argument action shall be read and executed by the shell when one of the 
corresponding conditions arises." Because it says "when", not "after", 
if interpreted literally, it does not even allow waiting until the 
current command finishes executing. I realise that that is definitely 
not the way it is meant to be interpreted, but I am not sure what is. I 
consider the assumption that the test script is supposed to work a 
reasonable one, but it is possible that this is considered strictly a 
QoI issue.

But to be clear, regardless of what POSIX requires, I was less concerned 
with prodding other shell authors into changing their shells and more 
with seeing what I can do in my shell. I want to have a shell that is 
capable of handling scripts like this, but it is fine with me if other 
shells do not share that as a goal.

Thanks for looking into this despite your scepticism on the validity of 
the test. Your description of what happens in bash when this ends up 
sleeping probably applies to all shells that behave the same way.

Cheers,
Harald van Dijk