On Thu, Jun 6, 2013 at 12:44 AM, Roland Mainz <[email protected]> wrote:
> On Sat, Jun 1, 2013 at 4:17 AM, Cedric Blancher
> <[email protected]> wrote:
>> I've tried to use the realtime signals in ast-ksh 20130524 to see if
>> they are reliable how but ran into instabilities again.
>> My example I used for testing is this one:
>> ------------cut--------------
> [snip]
>
> Attached (as "shrttest2_20130605.sh.txt") is an updated version of
> Cedric's test based on feedback...
> ... main differences:
> - The test now validates the _contents_ of the resulting array
> - SIGRTMIN up to SIGRTMIN+7 are now used to spread out the testing
> through multiple realtime signal channels in the attempt to weed-out
> any race conditions by having them arrive in a more or less random
> order
Attached (as "shrttest2_20130609.sh.txt") is an updated version of the
test. Main difference is that the workaround for
http://lists.research.att.com/pipermail/ast-developers/2013q2/002655.html
("[ast-developers] $ while ! wait ; do true ; done # doesn't wait for
all child processes...", e.g. I added a $ /usr/bin/true after the
while ! wait ; do true ; done #-loop to aid debugging) has been
applied and that the first dimension size of "rtar" is kept in a
seperate variable called "rtar_i2" to avoid a race condition with
signal handlers from different signals (thanks to Cedric for pointing
this out)
----
Bye,
Roland
--
__ . . __
(o.\ \/ /.o) [email protected]
\__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer
/O /==\ O\ TEL +49 641 3992797
(;O/ \/ \O;)
compound -a rtar
function rttrap
{
integer -S -a rtar_i2 # number of array elements in rtar[v]
integer v=${.sh.sig.value}
integer s=$(( rtar_i2[$v]++ ))
integer rtnum=$1
rtar[$v][$s]=(
integer pid=${.sh.sig.pid}
integer rtnum=$rtnum
typeset msg="${v}"
)
return 0
}
# POSIX gurantees at least 8 realtime signals
trap 'rttrap 0' RTMIN
trap 'rttrap 1' RTMIN+1
trap 'rttrap 2' RTMIN+2
trap 'rttrap 3' RTMIN+3
trap 'rttrap 4' RTMIN+4
trap 'rttrap 5' RTMIN+5
trap 'rttrap 6' RTMIN+6
trap 'rttrap 7' RTMIN+7
typeset m # used in child processes
integer pid p i j
# note that "numchildren" can't be higher than 4*$(ulimit -i)
integer -r numchildren=64
(( pid=$$ ))
for (( p=0 ; p < numchildren ; p++ )) ; do
{
# sleep for a moment to make sure the
# other child processes can become ready
# (sort of "poor man's" "barrier")
sleep 2
for m in 'a' 'b' 'c' 'd' 'e' 'f' ; do
kill -q $((16#$m)) -s RTMIN+6 $pid
kill -q $((16#$m)) -s RTMIN+7 $pid
kill -q $((16#$m)) -s RTMIN+4 $pid
kill -q $((16#$m)) -s RTMIN+5 $pid
kill -q $((16#$m)) -s RTMIN+2 $pid
kill -q $((16#$m)) -s RTMIN+3 $pid
kill -q $((16#$m)) -s RTMIN $pid
kill -q $((16#$m)) -s RTMIN+1 $pid
done
exit 0
} &
done
# wait for children to terminate
# we need to loop here since wait(1) may be aborted
# by signals
while ! wait ; do
true
done
# workaround the issue that "jobs -l" may not reflect the real
# up-to-date status of the child processes
# (see
http://lists.research.att.com/pipermail/ast-developers/2013q2/002655.html)
# we do this here to avoid any possible intererence from other
# job-related bugs
/usr/bin/true
# hack: force signal queue processing
true | read dummy
#
# validate results...
#
bool fail=false
if ! (( ${#rtar[@]} == 6 )) ; then
printf 'got {#rtar[@]}=%d, expected 6\n' ${#rtar[@]}
(( fail=true ))
fi
for (( i=0xa ; i <= 0xf; i++ )) ; do
if ! (( ${#rtar[$i][*]} == (numchildren*8) )) ; then
printf 'got {#rtar[%d][*]}=%d,' \
$i ${#rtar[$i][*]}
printf 'expected %d\n' $((numchildren*8))
(( fail=true ))
fi
done
for i in "${!rtar[@]}" ; do
for j in "${!rtar[$i][@]}" ; do
if [[ ! -v rtar[$i][$j].pid ]] ; then
printf 'variable %q missing.\n' \
"rtar[$i][$j].pid"
(( fail=true ))
fi
if [[ ! -v rtar[$i][$j].rtnum ]] ; then
printf 'variable %q missing.\n' \
"rtar[$i][$j].rtnum"
(( fail=true ))
fi
if [[ ! -v rtar[$i][$j].msg ]] ; then
printf 'variable %q missing.\n' \
"rtar[$i][$j].msg"
(( fail=true ))
elif [[ ! "${rtar[$i][$j].msg-}" == "$i" ]] ; then
printf 'variable %q value %q != %q.\n' \
"rtar[$i][$j].msg" \
"${rtar[$i][$j].msg-}" \
"${i}"
(( fail=true ))
fi
done
done
#(( fail )) && print -v rtar
print '#DONE.'
_______________________________________________
ast-developers mailing list
[email protected]
http://lists.research.att.com/mailman/listinfo/ast-developers
