[ILUG] bash v4 bug with `coproc' and `wait'? (No, just bash being silly.)

[Brian Foster]

 is this a bash(1) v4 bug, or am I missing something?

 script:
=====(cut here and below)=====script `t'=====
#!/bin/bash

echo "${BASH_VERSION[@]}"

SLEEP=${SLEEP:-sleep}

report() {
	echo -n ${1:?} wait returned ${2:?}
	if [ $2 -eq ${3:?} ]; then
		echo \ ... Ok
	else
		echo \ not expected $3
	fi
}

want=17
( exit $want ) &
$SLEEP 1		# simulate some processing
wait $!
report 1st $? $want

want=42
coproc { exit 42; }
$SLEEP 1		# simulate some processing
wait $COPROC_PID
report 2nd $? $want

want=55
coproc exit $want
$SLEEP 1		# simulate some processing
wait $COPROC_PID
report 3rd $? $want

want=69
coproc ( exit $want )
$SLEEP 1		# simulate some processing
wait $COPROC_PID
report 4th $? $want
=====(cut here and below)=====script `t'=====

 output:
=====(cut here and below)=====output of above script=====
$ bash ./t
4.2.10(1)-release
1st wait returned 17 ... Ok
2nd wait returned 0 not expected 42
3rd wait returned 0 not expected 55
4th wait returned 0 not expected 69
$
$ SLEEP=: bash ./t
4.2.10(1)-release
1st wait returned 17 ... Ok
2nd wait returned 42 ... Ok
3rd wait returned 55 ... Ok
4th wait returned 69 ... Ok
$
=====(cut here and above)=====output of above script=====

 it seems to be the case that if the `wait'
 is done after the coprocess has exited,
 then its status ($?) is always 0, unlike
 background (&) processes.  huh???

 LATE UPDATE:-  ah!  no!  got it... what's
 happening is COPROC_PID is being unset,
 and so the `wait $COPROC_PID' is turning
 into a simple `wait' (no PID(s)), which
 in this case _should_ return 0.

 prove by using:  wait ${COPROC_PID:?}

 silly bash.  silly silly bash.

cheers!
	-blf-

-- 
“How many surrealists does it take to    |  Brian Foster
 change a lightbulb?  Three.  One calms  |  somewhere in south of France
 the warthog, and two fill the bathtub   |     Stop E$$o (ExxonMobil)!
 with brightly-coloured machine tools.”  |       http://www.stopesso.com