#! /bin/sh -
i=0; while [ "$i" -lt 1000 ]; do
cmdA
cmdB > tempfile || continue
cmdC < tempfile || break
i=$((i + 1))
done
rm -f tempfile
If you can't litter the current directory with temp files, you could do:
#! /bin/sh -
i=0; while [ "$i" -lt 1000 ]; do
cmdA 3>&- 4<&-
tmp=$(mktemp) || exit
exec 3> "$tmp" 4< "$tmp"
rm -f "$tmp"
cmdB >&3 3>&- 4<&- || continue
cmdC <&4 3>&- 4<&- < tempfile || break
i=$((i + 1))
done
exec 3>&- 4<&-
rm -f tempfile
Whatever you do, you'll need to store that output somewhere as you can't get the exit status of cmdB until it has terminated.
You could store it in memory instead. If cmdB's output is text, you could do:
#! /bin/sh -
i=0; while [ "$i" -lt 1000 ]; do
cmdA
output=$(cmdB > tempfile && echo .) || continue
output=${output%.}
printf %s "$output" | cmdC || break
i=$((i + 1))
done
If printf is not built in your /bin/sh (for instance if you've installed mksh or yash and made it sh), that will fail for output larger than 128KiB.
Or you could use perl instead:
#! /usr/bin/perl
for ($i = 0; $i < 1000; $i++) {
system "cmdA";
$output = `cmdB`;
next if $?;
open TO_C, "|-", "cmdC" or last;
print TO_C $output;
close TO_C;
last if $?;
}