2

My script looks like: 

trap "$(pwd)/dd-destroy.sh $l-env;$(pwd)/dd-drop.sh $l-db;exit" INT QUIT TERM EXIT

./dd-all.sh $l $l-db $l-env || exit 1

app_ip=$(./dd-status.sh ip $l-env|grep docker-app|awk '{print $3}')
url=http://$app_ip:8080/app/

wget -O /dev/null $url || (echo "access $url failed" && false) || exit 1

Now I want this trap launch when error occurs, those 'exit 1' (I can change them there). But when nothing happens, like dd-all.sh worked fine, when wget came fine. I do not want to call for destroy.

Can I accomplish this with trap? It's not my script, I need to change it a little. Maybe trap is a bad call here?

|improve this question
  • 1
    trap specifies which signals the dd-destroy.sh... command should be invoked on. I don't know of any way for that command to test what the error code was when getting an EXIT signal. I'd recommend creating a function that has your dd-destroy.sh ... stuff in it; have the trap line invoke that function but omit the signal EXIT; and replace your exit 1 with { that_function; exit 1; }. – dubiousjim Oct 1 '15 at 14:52
  • Why are you listing EXIT in the trap specification if you don't want the trap to run on exit? – tripleee Oct 2 '15 at 10:45
  • @tripleee the OP wants the trap to run on normal (non-signalled) exits but only when the exit code is > 0. – dubiousjim Oct 4 '15 at 1:24
3

Simply change the start of your trap to check for the return code of the last command (which will be that of the exit, or the command at the end of the script) for zero. I.e. 

trap '[ $? = 0 ] && exit;'"..."

where "..." is your current string, which should follow with no intervening spaces. An exit within the trap will not trap again.

If your trap is also handling signals, eg SIGINT (control-C), then the trap is called once for the SIGINT, then once more for the exit. You can probably check if trap is being called from SIGINT as the $? will be 128+n where n is the signal, so 130 for SIGINT. Remember $? will be changed as soon as you give another command, so perhaps start with rc=$?; then use $rc in your tests.

|improve this answer
  • Huh, I was skeptical, but I just confirmed you are right. Tested on a v3 bash as well as a BusyBox sh (variant of ash), works as you say in both cases. – dubiousjim Oct 1 '15 at 20:27
  • Great thx, but there's still a little problem that I do not understand. 'trap '[ $? = 0 ] && exit;'"echo 'DUPA'" INT QUIT TERM EXIT </br> sleep 2 when I hit ctrl+c, i see 'dupa' printed twice. Why's that? Is trap handled twice then? When? – Sarseth Oct 2 '15 at 6:55
  • Yes, trap is called once for the sigint, then once for the exit. You can probably check if trap is being called from sigint as the $? will be 128+n where n is the signal, so 130 for sigint. Remember $? will be changed as soon as you give another command, so perhaps start with rc=$?; then use $rc. – meuh Oct 2 '15 at 7:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.