How to wait in bash for several subprocesses to finish, and return exit code !=0 when any subprocess ends with code !=0?

wait also (optionally) takes the PID of the process to wait for, and with $! you get the PID of the last command launched in the background. Modify the loop to store the PID of each spawned sub-process into an array, and then loop again waiting on each PID.

# run processes and store pids in array
for i in $n_procs; do
    ./procs[${i}] &
    pids[${i}]=$!
done

# wait for all pids
for pid in ${pids[*]}; do
    wait $pid
done

---

http://jeremy.zawodny.com/blog/archives/010717.html :

#!/bin/bash

FAIL=0

echo "starting"

./sleeper 2 0 &
./sleeper 2 1 &
./sleeper 3 0 &
./sleeper 2 0 &

for job in `jobs -p`
do
echo $job
    wait $job || let "FAIL+=1"
done

echo $FAIL

if [ "$FAIL" == "0" ];
then
echo "YAY!"
else
echo "FAIL! ($FAIL)"
fi

---

Here is simple example using wait.

Run some processes:

$ sleep 10 &
$ sleep 10 &
$ sleep 20 &
$ sleep 20 &

Then wait for them with wait command:

$ wait < <(jobs -p)

^{Or just wait (without arguments) for all.}

This will wait for all jobs in the background are completed.

^{If the -n option is supplied, waits for the next job to terminate and returns its exit status.}

See: help wait and help jobs for syntax.

However the downside is that this will return on only the status of the last ID, so you need to check the status for each subprocess and store it in the variable.

Or make your calculation function to create some file on failure (empty or with fail log), then check of that file if exists, e.g.

$ sleep 20 && true || tee fail &
$ sleep 20 && false || tee fail &
$ wait < <(jobs -p)
$ test -f fail && echo Calculation failed.

---

How about simply:

#!/bin/bash

pids=""

for i in `seq 0 9`; do
   doCalculations $i &
   pids="$pids $!"
done

wait $pids

...code continued here ...

Update:

As pointed by multiple commenters, the above waits for all processes to be completed before continuing, but does not exit and fail if one of them fails, it can be made to do with the following modification suggested by @Bryan, @SamBrightman, and others:

#!/bin/bash

pids=""
RESULT=0


for i in `seq 0 9`; do
   doCalculations $i &
   pids="$pids $!"
done

for pid in $pids; do
    wait $pid || let "RESULT=1"
done

if [ "$RESULT" == "1" ];
    then
       exit 1
fi

...code continued here ...

---

If you have GNU Parallel installed you can do:

# If doCalculations is a function
export -f doCalculations
seq 0 9 | parallel doCalculations {}

GNU Parallel will give you exit code:

0 - All jobs ran without error.

1-253 - Some of the jobs failed. The exit status gives the number of failed jobs

254 - More than 253 jobs failed.

255 - Other error.

Watch the intro videos to learn more: http://pi.dk/1

---

Here's what I've come up with so far. I would like to see how to interrupt the sleep command if a child terminates, so that one would not have to tune WAITALL_DELAY to one's usage.

waitall() { # PID...
  ## Wait for children to exit and indicate whether all exited with 0 status.
  local errors=0
  while :; do
    debug "Processes remaining: $*"
    for pid in "$@"; do
      shift
      if kill -0 "$pid" 2>/dev/null; then
        debug "$pid is still alive."
        set -- "$@" "$pid"
      elif wait "$pid"; then
        debug "$pid exited with zero exit status."
      else
        debug "$pid exited with non-zero exit status."
        ((++errors))
      fi
    done
    (("$#" > 0)) || break
    # TODO: how to interrupt this sleep when a child terminates?
    sleep ${WAITALL_DELAY:-1}
   done
  ((errors == 0))
}

debug() { echo "DEBUG: $*" >&2; }

pids=""
for t in 3 5 4; do 
  sleep "$t" &
  pids="$pids $!"
done
waitall $pids

---

To parallelize this...

for i in $(whatever_list) ; do
   do_something $i
done

Translate it to this...

for i in $(whatever_list) ; do echo $i ; done | ## execute in parallel...
   (
   export -f do_something ## export functions (if needed)
   export PATH ## export any variables that are required
   xargs -I{} --max-procs 0 bash -c ' ## process in batches...
      {
      echo "processing {}" ## optional
      do_something {}
      }' 
   )

If an error occurs in one process, it won't interrupt the other processes, but it will result in a non-zero exit code from the sequence as a whole.

Exporting functions and variables may or may not be necessary, in any particular case.

You can set --max-procs based on how much parallelism you want (0 means "all at once").

GNU Parallel offers some additional features when used in place of xargs -- but it isn't always installed by default.

The for loop isn't strictly necessary in this example since echo $i is basically just regenerating the output of $(whatever_list). I just think the use of the for keyword makes it a little easier to see what is going on.

Bash string handling can be confusing -- I have found that using single quotes works best for wrapping non-trivial scripts.

You can easily interrupt the entire operation (using ^C or similar), unlike the the more direct approach to Bash parallelism.

Here's a simplified working example...

for i in {0..5} ; do echo $i ; done |xargs -I{} --max-procs 2 bash -c '
   {
   echo sleep {}
   sleep 2s
   }'

---

This is something that I use:

#wait for jobs
for job in `jobs -p`; do wait ${job}; done

---

I see lots of good examples listed on here, wanted to throw mine in as well.

#! /bin/bash

items="1 2 3 4 5 6"
pids=""

for item in $items; do
    sleep $item &
    pids+="$! "
done

for pid in $pids; do
    wait $pid
    if [ $? -eq 0 ]; then
        echo "SUCCESS - Job $pid exited with a status of $?"
    else
        echo "FAILED - Job $pid exited with a status of $?"
    fi
done

I use something very similar to start/stop servers/services in parallel and check each exit status. Works great for me. Hope this helps someone out!

---

I don't believe it's possible with Bash's builtin functionality.

You can get notification when a child exits:

#!/bin/sh
set -o monitor        # enable script job control
trap 'echo "child died"' CHLD

However there's no apparent way to get the child's exit status in the signal handler.

Getting that child status is usually the job of the wait family of functions in the lower level POSIX APIs. Unfortunately Bash's support for that is limited - you can wait for one specific child process (and get its exit status) or you can wait for all of them, and always get a 0 result.

What it appears impossible to do is the equivalent of waitpid(-1), which blocks until any child process returns.

---

The following code will wait for completion of all calculations and return exit status 1 if any of doCalculations fails.

#!/bin/bash
for i in $(seq 0 9); do
   (doCalculations $i >&2 & wait %1; echo $?) &
done | grep -qv 0 && exit 1

---

This is an expansion on the most-upvoted answer, by @Luca Tettamanti, to make a fully-runnable example.

That answer left me wondering:

What type of variable is n_procs, and what does it contain? What type of variable is procs, and what does it contain? Can someone please update this answer to make it runnable by adding definitions for those variables? I don't understand how.

...and also:

How do you get the return code from the subprocess when it has completed (which is the whole crux of this question)?

Anyway, I figured it out, so here is a fully-runnable example.

Notes:

$! is how to obtain the PID (Process ID) of the last-executed sub-process. Running any command with the & after it, like cmd &, for example, causes it to run in the background as a parallel suprocess with the main process. myarray=() is how to create an array in bash. To learn a tiny bit more about the wait built-in command, see help wait. See also, and especially, the official Bash user manual on Job Control built-ins, such as wait and jobs, here: https://www.gnu.org/software/bash/manual/html_node/Job-Control-Builtins.html#index-wait.

Full, runnable program: wait for all processes to end

multi_process_program.sh (from my eRCaGuy_hello_world repo):

#!/usr/bin/env bash


# This is a special sleep function which returns the number of seconds slept as
# the "error code" or return code" so that we can easily see that we are in
# fact actually obtaining the return code of each process as it finishes.
my_sleep() {
    seconds_to_sleep="$1"
    sleep "$seconds_to_sleep"
    return "$seconds_to_sleep"
}

# Create an array of whatever commands you want to run as subprocesses
procs=()  # bash array
procs+=("my_sleep 5")
procs+=("my_sleep 2")
procs+=("my_sleep 3")
procs+=("my_sleep 4")

num_procs=${#procs[@]}  # number of processes
echo "num_procs = $num_procs"

# run commands as subprocesses and store pids in an array
pids=()  # bash array
for (( i=0; i<"$num_procs"; i++ )); do
    echo "cmd = ${procs[$i]}"
    ${procs[$i]} &  # run the cmd as a subprocess
    # store pid of last subprocess started; see:
    # https://unix.stackexchange.com/a/30371/114401
    pids+=("$!")
    echo "    pid = ${pids[$i]}"
done

# OPTION 1 (comment this option out if using Option 2 below): wait for all pids
for pid in "${pids[@]}"; do
    wait "$pid"
    return_code="$?"
    echo "PID = $pid; return_code = $return_code"
done
echo "All $num_procs processes have ended."

Change the file above to be executable by running chmod +x multi_process_program.sh, then run it like this:

time ./multi_process_program.sh 

Sample output. See how the output of the time command in the call shows it took 5.084sec to run. We were also able to successfully retrieve the return code from each subprocess.

eRCaGuy_hello_world/bash$ time ./multi_process_program.sh num_procs = 4 cmd = my_sleep 5 pid = 21694 cmd = my_sleep 2 pid = 21695 cmd = my_sleep 3 pid = 21697 cmd = my_sleep 4 pid = 21699 PID = 21694; return_code = 5 PID = 21695; return_code = 2 PID = 21697; return_code = 3 PID = 21699; return_code = 4 All 4 processes have ended. PID 21694 is done; return_code = 5; 3 PIDs remaining. PID 21695 is done; return_code = 2; 2 PIDs remaining. PID 21697 is done; return_code = 3; 1 PIDs remaining. PID 21699 is done; return_code = 4; 0 PIDs remaining. real 0m5.084s user 0m0.025s sys 0m0.061s

Going further: determine live when each individual process ends

If you'd like to do some action as each process finishes, and you don't know when they will finish, you can poll in an infinite while loop to see when each process terminates, then do whatever action you want.

Simply comment out the "OPTION 1" block of code above, and replace it with this "OPTION 2" block instead:

# OR OPTION 2 (comment out Option 1 above if using Option 2): poll to detect
# when each process terminates, and print out when each process finishes!
while true; do
    for i in "${!pids[@]}"; do
        pid="${pids[$i]}"
        # echo "pid = $pid"  # debugging

        # See if PID is still running; see my answer here:
        # https://stackoverflow.com/a/71134379/4561887
        ps --pid "$pid" > /dev/null
        if [ "$?" -ne 0 ]; then
            # PID doesn't exist anymore, meaning it terminated

            # 1st, read its return code
            wait "$pid"
            return_code="$?"

            # 2nd, remove this PID from the `pids` array by `unset`ting the
            # element at this index; NB: due to how bash arrays work, this does
            # NOT actually remove this element from the array. Rather, it
            # removes its index from the `"${!pids[@]}"` list of indices,
            # adjusts the array count(`"${#pids[@]}"`) accordingly, and it sets
            # the value at this index to either a null value of some sort, or
            # an empty string (I'm not exactly sure).
            unset "pids[$i]"

            num_pids="${#pids[@]}"
            echo "PID $pid is done; return_code = $return_code;" \
                 "$num_pids PIDs remaining."
        fi
    done

    # exit the while loop if the `pids` array is empty
    if [ "${#pids[@]}" -eq 0 ]; then
        break
    fi

    # Do some small sleep here to keep your polling loop from sucking up
    # 100% of one of your CPUs unnecessarily. Sleeping allows other processes
    # to run during this time.
    sleep 0.1
done

Sample run and output of the full program with Option 1 commented out and Option 2 in-use:

eRCaGuy_hello_world/bash$ ./multi_process_program.sh num_procs = 4 cmd = my_sleep 5 pid = 22275 cmd = my_sleep 2 pid = 22276 cmd = my_sleep 3 pid = 22277 cmd = my_sleep 4 pid = 22280 PID 22276 is done; return_code = 2; 3 PIDs remaining. PID 22277 is done; return_code = 3; 2 PIDs remaining. PID 22280 is done; return_code = 4; 1 PIDs remaining. PID 22275 is done; return_code = 5; 0 PIDs remaining.

Each of those PID XXXXX is done lines prints out live right after that process has terminated! Notice that even though the process for sleep 5 (PID 22275 in this case) was run first, it finished last, and we successfully detected each process right after it terminated. We also successfully detected each return code, just like in Option 1.

Other References:

*****+ [VERY HELPFUL] Get exit code of a background process - this answer taught me the key principle that (emphasis added): wait waits until the process with PID is complete (it will block until the process completes, so you might not want to call this until you are sure the process is done), and then returns the exit code of the completed process. In other words, it helped me know that even after the process is complete, you can still call wait on it to get its return code! How to check if a process id (PID) exists my answer Remove an element from a Bash array - note that elements in a bash array aren't actually deleted, they are just "unset". See my comments in the code above for what that means. How to use the command-line executable true to make an infinite while loop in bash: https://www.cyberciti.biz/faq/bash-infinite-loop/

---

If you have bash 4.2 or later available the following might be useful to you. It uses associative arrays to store task names and their "code" as well as task names and their pids. I have also built in a simple rate-limiting method which might come handy if your tasks consume a lot of CPU or I/O time and you want to limit the number of concurrent tasks.

The script launches all tasks in the first loop and consumes the results in the second one.

This is a bit overkill for simple cases but it allows for pretty neat stuff. For example one can store error messages for each task in another associative array and print them after everything has settled down.

#! /bin/bash

main () {
    local -A pids=()
    local -A tasks=([task1]="echo 1"
                    [task2]="echo 2"
                    [task3]="echo 3"
                    [task4]="false"
                    [task5]="echo 5"
                    [task6]="false")
    local max_concurrent_tasks=2

    for key in "${!tasks[@]}"; do
        while [ $(jobs 2>&1 | grep -c Running) -ge "$max_concurrent_tasks" ]; do
            sleep 1 # gnu sleep allows floating point here...
        done
        ${tasks[$key]} &
        pids+=(["$key"]="$!")
    done

    errors=0
    for key in "${!tasks[@]}"; do
        pid=${pids[$key]}
        local cur_ret=0
        if [ -z "$pid" ]; then
            echo "No Job ID known for the $key process" # should never happen
            cur_ret=1
        else
            wait $pid
            cur_ret=$?
        fi
        if [ "$cur_ret" -ne 0 ]; then
            errors=$(($errors + 1))
            echo "$key (${tasks[$key]}) failed."
        fi
    done

    return $errors
}

main

---

Here's my version that works for multiple pids, logs warnings if execution takes too long, and stops the subprocesses if execution takes longer than a given value.

function WaitForTaskCompletion {
    local pids="${1}" # pids to wait for, separated by semi-colon
    local soft_max_time="${2}" # If execution takes longer than $soft_max_time seconds, will log a warning, unless $soft_max_time equals 0.
    local hard_max_time="${3}" # If execution takes longer than $hard_max_time seconds, will stop execution, unless $hard_max_time equals 0.
    local caller_name="${4}" # Who called this function
    local exit_on_error="${5:-false}" # Should the function exit program on subprocess errors       

    Logger "${FUNCNAME[0]} called by [$caller_name]."

    local soft_alert=0 # Does a soft alert need to be triggered, if yes, send an alert once 
    local log_ttime=0 # local time instance for comparaison

    local seconds_begin=$SECONDS # Seconds since the beginning of the script
    local exec_time=0 # Seconds since the beginning of this function

    local retval=0 # return value of monitored pid process
    local errorcount=0 # Number of pids that finished with errors

    local pidCount # number of given pids

    IFS=';' read -a pidsArray <<< "$pids"
    pidCount=${#pidsArray[@]}

    while [ ${#pidsArray[@]} -gt 0 ]; do
        newPidsArray=()
        for pid in "${pidsArray[@]}"; do
            if kill -0 $pid > /dev/null 2>&1; then
                newPidsArray+=($pid)
            else
                wait $pid
                result=$?
                if [ $result -ne 0 ]; then
                    errorcount=$((errorcount+1))
                    Logger "${FUNCNAME[0]} called by [$caller_name] finished monitoring [$pid] with exitcode [$result]."
                fi
            fi
        done

        ## Log a standby message every hour
        exec_time=$(($SECONDS - $seconds_begin))
        if [ $((($exec_time + 1) % 3600)) -eq 0 ]; then
            if [ $log_ttime -ne $exec_time ]; then
                log_ttime=$exec_time
                Logger "Current tasks still running with pids [${pidsArray[@]}]."
            fi
        fi

        if [ $exec_time -gt $soft_max_time ]; then
            if [ $soft_alert -eq 0 ] && [ $soft_max_time -ne 0 ]; then
                Logger "Max soft execution time exceeded for task [$caller_name] with pids [${pidsArray[@]}]."
                soft_alert=1
                SendAlert

            fi
            if [ $exec_time -gt $hard_max_time ] && [ $hard_max_time -ne 0 ]; then
                Logger "Max hard execution time exceeded for task [$caller_name] with pids [${pidsArray[@]}]. Stopping task execution."
                kill -SIGTERM $pid
                if [ $? == 0 ]; then
                    Logger "Task stopped successfully"
                else
                    errrorcount=$((errorcount+1))
                fi
            fi
        fi

        pidsArray=("${newPidsArray[@]}")
        sleep 1
    done

    Logger "${FUNCNAME[0]} ended for [$caller_name] using [$pidCount] subprocesses with [$errorcount] errors."
    if [ $exit_on_error == true ] && [ $errorcount -gt 0 ]; then
        Logger "Stopping execution."
        exit 1337
    else
        return $errorcount
    fi
}

# Just a plain stupid logging function to be replaced by yours
function Logger {
    local value="${1}"

    echo $value
}

Example, wait for all three processes to finish, log a warning if execution takes loger than 5 seconds, stop all processes if execution takes longer than 120 seconds. Don't exit program on failures.

function something {

    sleep 10 &
    pids="$!"
    sleep 12 &
    pids="$pids;$!"
    sleep 9 &
    pids="$pids;$!"

    WaitForTaskCompletion $pids 5 120 ${FUNCNAME[0]} false
}
# Launch the function
someting
    

---

Wait for all jobs and return the exit code of the last failing job. Unlike solutions above, this does not require pid saving, or modifying inner loops of scripts. Just bg away, and wait.

function wait_ex {
    # this waits for all jobs and returns the exit code of the last failing job
    ecode=0
    while true; do
        [ -z "$(jobs)" ] && break
        wait -n
        err="$?"
        [ "$err" != "0" ] && ecode="$err"
    done
    return $ecode
}

EDIT: Fixed the bug where this could be fooled by a script that ran a command that didn't exist.

---

Just store the results out of the shell, e.g. in a file.

#!/bin/bash
tmp=/tmp/results

: > $tmp  #clean the file

for i in `seq 0 9`; do
  (doCalculations $i; echo $i:$?>>$tmp)&
done      #iterate

wait      #wait until all ready

sort $tmp | grep -v ':0'  #... handle as required

---

I've had a go at this and combined all the best parts from the other examples here. This script will execute the checkpids function when any background process exits, and output the exit status without resorting to polling.

#!/bin/bash

set -o monitor

sleep 2 &
sleep 4 && exit 1 &
sleep 6 &

pids=`jobs -p`

checkpids() {
    for pid in $pids; do
        if kill -0 $pid 2>/dev/null; then
            echo $pid is still alive.
        elif wait $pid; then
            echo $pid exited with zero exit status.
        else
            echo $pid exited with non-zero exit status.
        fi
    done
    echo
}

trap checkpids CHLD

wait

---

#!/bin/bash
set -m
for i in `seq 0 9`; do
  doCalculations $i &
done
while fg; do true; done

set -m allows you to use fg & bg in a script

fg, in addition to putting the last process in the foreground, has the same exit status as the process it foregrounds

while fg will stop looping when any fg exits with a non-zero exit status

unfortunately this won't handle the case when a process in the background exits with a non-zero exit status. (the loop won't terminate immediately. it will wait for the previous processes to complete.)

---

I've just been modifying a script to background and parallelise a process.

I did some experimenting (on Solaris with both bash and ksh) and discovered that 'wait' outputs the exit status if it's not zero , or a list of jobs that return non-zero exit when no PID argument is provided. E.g.

Bash:

$ sleep 20 && exit 1 &
$ sleep 10 && exit 2 &
$ wait
[1]-  Exit 2                  sleep 20 && exit 2
[2]+  Exit 1                  sleep 10 && exit 1

Ksh:

$ sleep 20 && exit 1 &
$ sleep 10 && exit 2 &
$ wait
[1]+  Done(2)                  sleep 20 && exit 2
[2]+  Done(1)                  sleep 10 && exit 1

This output is written to stderr, so a simple solution to the OPs example could be:

#!/bin/bash

trap "rm -f /tmp/x.$$" EXIT

for i in `seq 0 9`; do
  doCalculations $i &
done

wait 2> /tmp/x.$$
if [ `wc -l /tmp/x.$$` -gt 0 ] ; then
  exit 1
fi

While this:

wait 2> >(wc -l)

will also return a count but without the tmp file. This might also be used this way, for example:

wait 2> >(if [ `wc -l` -gt 0 ] ; then echo "ERROR"; fi)

But this isn't very much more useful than the tmp file IMO. I couldn't find a useful way to avoid the tmp file whilst also avoiding running the "wait" in a subshell, which wont work at all.

---

I needed this, but the target process wasn't a child of current shell, in which case wait $PID doesn't work. I did find the following alternative instead:

while [ -e /proc/$PID ]; do sleep 0.1 ; done

That relies on the presence of procfs, which may not be available (Mac doesn't provide it for example). So for portability, you could use this instead:

while ps -p $PID >/dev/null ; do sleep 0.1 ; done

---

There are already a lot of answers here, but I am surprised no one seems to have suggested using arrays... So here's what I did - this might be useful to some in the future.

n=10 # run 10 jobs
c=0
PIDS=()

while true

    my_function_or_command &
    PID=$!
    echo "Launched job as PID=$PID"
    PIDS+=($PID)

    (( c+=1 ))

    # required to prevent any exit due to error
    # caused by additional commands run which you
    # may add when modifying this example
    true

do

    if (( c < n ))
    then
        continue
    else
        break
    fi
done 


# collect launched jobs

for pid in "${PIDS[@]}"
do
    wait $pid || echo "failed job PID=$pid"
done

---

This works, should be just as a good if not better than @HoverHell's answer!

#!/usr/bin/env bash

set -m # allow for job control
EXIT_CODE=0;  # exit code of overall script

function foo() {
     echo "CHLD exit code is $1"
     echo "CHLD pid is $2"
     echo $(jobs -l)

     for job in `jobs -p`; do
         echo "PID => ${job}"
         wait ${job} ||  echo "At least one test failed with exit code => $?" ; EXIT_CODE=1
     done
}

trap 'foo $? $$' CHLD

DIRN=$(dirname "$0");

commands=(
    "{ echo "foo" && exit 4; }"
    "{ echo "bar" && exit 3; }"
    "{ echo "baz" && exit 5; }"
)

clen=`expr "${#commands[@]}" - 1` # get length of commands - 1

for i in `seq 0 "$clen"`; do
    (echo "${commands[$i]}" | bash) &   # run the command via bash in subshell
    echo "$i ith command has been issued as a background job"
done

# wait for all to finish
wait;

echo "EXIT_CODE => $EXIT_CODE"
exit "$EXIT_CODE"

# end

and of course, I have immortalized this script, in an NPM project which allows you to run bash commands in parallel, useful for testing:

https://github.com/ORESoftware/generic-subshell

---

Exactly for this purpose I wrote a bash function called :for.

Note: :for not only preserves and returns the exit code of the failing function, but also terminates all parallel running instance. Which might not be needed in this case.

#!/usr/bin/env bash

# Wait for pids to terminate. If one pid exits with
# a non zero exit code, send the TERM signal to all
# processes and retain that exit code
#
# usage:
# :wait 123 32
function :wait(){
    local pids=("$@")
    [ ${#pids} -eq 0 ] && return $?

    trap 'kill -INT "${pids[@]}" &>/dev/null || true; trap - INT' INT
    trap 'kill -TERM "${pids[@]}" &>/dev/null || true; trap - RETURN TERM' RETURN TERM

    for pid in "${pids[@]}"; do
        wait "${pid}" || return $?
    done

    trap - INT RETURN TERM
}

# Run a function in parallel for each argument.
# Stop all instances if one exits with a non zero
# exit code
#
# usage:
# :for func 1 2 3
#
# env:
# FOR_PARALLEL: Max functions running in parallel
function :for(){
    local f="${1}" && shift

    local i=0
    local pids=()
    for arg in "$@"; do
        ( ${f} "${arg}" ) &
        pids+=("$!")
        if [ ! -z ${FOR_PARALLEL+x} ]; then
            (( i=(i+1)%${FOR_PARALLEL} ))
            if (( i==0 )) ;then
                :wait "${pids[@]}" || return $?
                pids=()
            fi
        fi
    done && [ ${#pids} -eq 0 ] || :wait "${pids[@]}" || return $?
}

usage

for.sh:

#!/usr/bin/env bash
set -e

# import :for from gist: https://gist.github.com/Enteee/c8c11d46a95568be4d331ba58a702b62#file-for
# if you don't like curl imports, source the actual file here.
source <(curl -Ls https://gist.githubusercontent.com/Enteee/c8c11d46a95568be4d331ba58a702b62/raw/)

msg="You should see this three times"

:(){
  i="${1}" && shift

  echo "${msg}"

  sleep 1
  if   [ "$i" == "1" ]; then sleep 1
  elif [ "$i" == "2" ]; then false
  elif [ "$i" == "3" ]; then
    sleep 3
    echo "You should never see this"
  fi
} && :for : 1 2 3 || exit $?

echo "You should never see this"

$ ./for.sh; echo $?
You should see this three times
You should see this three times
You should see this three times
1

References

[1]: blog

[2]: gist

---

set -e
fail () {
    touch .failure
}
expect () {
    wait
    if [ -f .failure ]; then
        rm -f .failure
        exit 1
    fi
}

sleep 2 || fail &
sleep 2 && false || fail &
sleep 2 || fail
expect

The set -e at top makes your script stop on failure.

expect will return 1 if any subjob failed.

---

There can be a case where the process is complete before waiting for the process. If we trigger wait for a process that is already finished, it will trigger an error like pid is not a child of this shell. To avoid such cases, the following function can be used to find whether the process is complete or not:

isProcessComplete(){
PID=$1
while [ -e /proc/$PID ]
do
    echo "Process: $PID is still running"
    sleep 5
done
echo "Process $PID has finished"
}

---

I used this recently (thanks to Alnitak):

#!/bin/bash
# activate child monitoring
set -o monitor

# locking subprocess
(while true; do sleep 0.001; done) &
pid=$!

# count, and kill when all done
c=0
function kill_on_count() {
    # you could kill on whatever criterion you wish for
    # I just counted to simulate bash's wait with no args
    [ $c -eq 9 ] && kill $pid
    c=$((c+1))
    echo -n '.' # async feedback (but you don't know which one)
}
trap "kill_on_count" CHLD

function save_status() {
    local i=$1;
    local rc=$2;
    # do whatever, and here you know which one stopped
    # but remember, you're called from a subshell
    # so vars have their values at fork time
}

# care must be taken not to spawn more than one child per loop
# e.g don't use `seq 0 9` here!
for i in {0..9}; do
    (doCalculations $i; save_status $i $?) &
done

# wait for locking subprocess to be killed
wait $pid
echo

From there one can easily extrapolate, and have a trigger (touch a file, send a signal) and change the counting criteria (count files touched, or whatever) to respond to that trigger. Or if you just want 'any' non zero rc, just kill the lock from save_status.

---

Trapping CHLD signal may not work because you can lose some signals if they arrived simultaneously.

#!/bin/bash

trap 'rm -f $tmpfile' EXIT

tmpfile=$(mktemp)

doCalculations() {
    echo start job $i...
    sleep $((RANDOM % 5)) 
    echo ...end job $i
    exit $((RANDOM % 10))
}

number_of_jobs=10

for i in $( seq 1 $number_of_jobs )
do
    ( trap "echo job$i : exit value : \$? >> $tmpfile" EXIT; doCalculations ) &
done

wait 

i=0
while read res; do
    echo "$res"
    let i++
done < "$tmpfile"

echo $i jobs done !!!

---

solution to wait for several subprocesses and to exit when any one of them exits with non-zero status code is by using 'wait -n'

#!/bin/bash
wait_for_pids()
{
    for (( i = 1; i <= $#; i++ )) do
        wait -n $@
        status=$?
        echo "received status: "$status
        if [ $status -ne 0 ] && [ $status -ne 127 ]; then
            exit 1
        fi
    done
}

sleep_for_10()
{
    sleep 10
    exit 10
}

sleep_for_20()
{
    sleep 20
}

sleep_for_10 &
pid1=$!

sleep_for_20 &
pid2=$!

wait_for_pids $pid2 $pid1

status code '127' is for non-existing process which means the child might have exited.

---

I almost fell into the trap of using jobs -p to collect PIDs, which does not work if the child has already exited, as shown in the script below. The solution I picked was simply calling wait -n N times, where N is the number of children I have, which I happen to know deterministically.

#!/usr/bin/env bash

sleeper() {
    echo "Sleeper $1"
    sleep $2
    echo "Exiting $1"
    return $3
}

start_sleepers() {
    sleeper 1 1 0 &
    sleeper 2 2 $1 &
    sleeper 3 5 0 &
    sleeper 4 6 0 &
    sleep 4
}

echo "Using jobs"
start_sleepers 1

pids=( $(jobs -p) )

echo "PIDS: ${pids[*]}"

for pid in "${pids[@]}"; do
    wait "$pid"
    echo "Exit code $?"
done

echo "Clearing other children"
wait -n; echo "Exit code $?"
wait -n; echo "Exit code $?"

echo "Waiting for N processes"
start_sleepers 2

for ignored in $(seq 1 4); do
    wait -n
    echo "Exit code $?"
done

Output:

Using jobs
Sleeper 1
Sleeper 2
Sleeper 3
Sleeper 4
Exiting 1
Exiting 2
PIDS: 56496 56497
Exiting 3
Exit code 0
Exiting 4
Exit code 0
Clearing other children
Exit code 0
Exit code 1
Waiting for N processes
Sleeper 1
Sleeper 2
Sleeper 3
Sleeper 4
Exiting 1
Exiting 2
Exit code 0
Exit code 2
Exiting 3
Exit code 0
Exiting 4
Exit code 0

---

Starting with Bash 5.1, there is a nice new way of waiting for and handling the results of multiple background jobs thanks to the introduction of wait -p:

#!/usr/bin/env bash

# Spawn background jobs
for ((i=0; i < 10; i++)); do
    secs=$((RANDOM % 10)); code=$((RANDOM % 256))
    (sleep ${secs}; exit ${code}) &
    echo "Started background job (pid: $!, sleep: ${secs}, code: ${code})"
done

# Wait for background jobs, print individual results, determine overall result
result=0
while true; do
    wait -n -p pid; code=$?
    [[ -z "${pid}" ]] && break
    echo "Background job ${pid} finished with code ${code}"
    (( ${code} != 0 )) && result=1
done

# Return overall result
exit ${result}

---