Use timeout utility instead of double-sub-shell hell

It turned out that for incremental builds for Node.js, the test always waited
10m no matter that the build finished sooner.

It's not clear why it happened, but when debugging, it seemed like some weired
race-condition happened that let some sleeps hang.

Anyway, it's not clear whether there is a reason to not use timeout utility
that seems to serve exactly for this reason.

Author: hhorak

test-lib.sh

@@ -74,30 +74,14 @@ ct_build_image_and_parse_id() {
   local ret_val
   local dockerfile
   local command
-  local pid_build
-  local pid_sleep
   local sleep_time
   log_file="$(mktemp)"
   sleep_time="10m"
   [ -n "$1" ] && dockerfile="-f $1"
   command="$(echo "docker build --no-cache $dockerfile $2" | tr -d "'")"
-  # running command in subshell, the subshell in background, storing pid to variable
-  (
-    $command > "$log_file" 2>&1
-  ) & pid_build=$!
-  # creating second subshell with trap function on ALRM signal
-  # the subshell sleeps for 10m, then kills the first subshell
-  (
-    trap 'exit 0' ALRM; sleep "$sleep_time" && kill $pid_build
-  ) & pid_sleep=$!
-  # waiting for build subshell to finish, either with success, or killed from sleep subshell
-  wait $pid_build
+  # shellcheck disable=SC2086
+  timeout $sleep_time $command > "$log_file" 2>&1
   ret_val=$?
-  # send ALRM signal to the sleep subshell, so it exits even in case the 10mins
-  # not yet passed. If the kill was successful (the wait subshell received ALRM signal)
-  # then the build was not finished yet, so the return value is set to 1
-  kill -s ALRM $pid_sleep 2>/dev/null || ret_val=1
-
   if [ $ret_val -eq 0 ]; then
     APP_IMAGE_ID="$(tail -n 1 "$log_file")"
   fi