As part of my build process, I am running a git commit as an execute shell step. However, if there are no changes in the workspace, Jenkins is failing the build. This is because git is returning an error code when there are no changes to commit. I'd like to either abort the build, or just mark it as unstable if this is the case. Any ideas?
15 Answers
To stop further execution when command fails:
command || exit 0
To continue execution when command fails:
command || true
9 Comments
Nir Alfasi
You don't need the
|| exit 0 in the first case, if command returns false the execution will stop. That said, the second option is very helpful!
Quolonel Questions
@alfasin You don't understand the problem. OP doesn't want the Jenkins build to fail; ergo we must
exit 0 because any non-zero exit code will fail the build.Nir Alfasi
I see, in that case I would change the wording from: "To stop further execution when command fails:" to: "To stop further execution when command fails and mark Jenkins job as successful:".
noamik
@alfasin While I agree that Quolonel Questions snappy remark was unprofessional, he was right in what he said. "exit 0" will NOT mark the job successful. It will just mark the current build step successful. The job can still fail on one of the next build steps.
Nir Alfasi
@noamik you're right! I was dealing with jobs with only one step of "execute shell" hence the misunderstanding. Thanks for clarifying!
|
Jenkins is executing shell build steps using /bin/sh -xe by default. -x means to print every command executed. -e means to exit with failure if any of the commands in the script failed.
So I think what happened in your case is your git command exit with 1, and because of the default -e param, the shell picks up the non-0 exit code, ignores the rest of the script and marks the step as a failure. We can confirm this if you can post your build step script here.
If that's the case, you can try to put #!/bin/sh so that the script will be executed without option; or do a set +e or anything similar on top of the build step to override this behavior.
Edited: Another thing to note is that, if the last command in your shell script returns non-0 code, the whole build step will still be marked as fail even with this setup. In this case, you can simply put a true command at the end to avoid that.
1 Comment
Ray Oei
This actually solved my problem with Jenkins failing even when trying to suppress errors with
2>/dev/null. Thanks!If there is nothing to push git returns exit status 1. Execute shell build step is marked as failed respectively. You can use OR statement || (double pipe).
git commit -m 'some messasge' || echo 'Commit failed. There is probably nothing to commit.'
That means, execute second argument if first failed (returned exit status > 0). Second command always returns 0. When there is nothing to push (exit status 1 -> execute second command) echo will return 0 and build step continues.
To mark build as unstable you can use post-build step Jenkins Text Finder. It can go through console output, match pattern (your echo) and mark build as unstable.
Comments
There is another smooth way to tell Jenkins not to fail. You can isolate your commit in a build step and set the shell to not fail:
set +e
git commit -m "Bla."
set -e
1 Comment
jcasner
Make sure to add
set -e after the command that you want to run irrespective of exit code. Otherwise, you may end up executing commands you don't intend to. I wanted to handle the error myself, so I did something like: ` set +e commit -m "bla" EXIT_CODE="${?}" set -e # handle exit code logic `This answer is correct, but it doesn't specify the || exit 0 or || true goes inside the shell command. Here's a more complete example:
sh "adb uninstall com.example.app || true"
The above will work, but the following will fail:
sh "adb uninstall com.example.app" || true
Perhaps it's obvious to others, but I wasted a lot of time before I realized this.
Comments
https://jenkins.io/doc/pipeline/steps/workflow-durable-task-step/#sh-shell-script
if you include a returnStatus: true property, then the shell return is ignored.
1 Comment
AndresVia
this is the real answer, a full example would look like:
def rc = sh(script: "git commit", returnStatus: true) ; echo "git commit return code was: ${rc} moving on"I was able to get this working using the answer found here:
How to git commit nothing without an error?
git diff --quiet --exit-code --cached || git commit -m 'bla'
2 Comments
Slav
What the above does is: "Do
git diff command, and if that fails, do git commit command. Basically, it only does the commit, if the git diff found something to commit. However @jwernerny answer was correct that you should be able to add exit 0 as the last statement to any script to make Jenkins treat it as success. I can think of one scenario where this would fail if you were doing Linux shell step, but in Batch this should always work.
Xiawei Zhang
@Ben Jenkins is executing shell build steps using
/bin/sh -xe by default as mentioned here (in the middle). So you can try to put #!/bin/bash or do a set +e on top of the build step to override this behavior, which will continue the rest of the step even one command inside exit with non-0 codeJenkins determines the success/failure of a step by the return value of the step. For the case of a shell, it should be the return of the last value. For both Windows CMD and (POSIX) Bash shells, you should be able to set the return value manually by using exit 0 as the last command.
4 Comments
Ben
this doesn't seem to work for an 'execute windows bat' that has 2 lines: git commit -m "message" exit 0
jwernerny
@Ben I use
exit 0 with "execute windows batch command" in multiple builds on my Windows Jenkins install, and it works as expected. Something else must be going on. Could you post the relevant part of the console log?Ben
are you using it with git commit -m "blah" in your first step? I tried creating a bat script on the machine manually, and put an echo and an exit 0 after the git command. Neither of the other commands are run when there is nothing to commit...
Steven the Easily Amused
See answer from @xiawei. The default behavior of Jenkins it to execute a shell with
#!/bin/sh -xv which results in stopping the script if any error is encountered.On the (more general) question in title - to prevent Jenkins from failing you can prevent it from seeing exit code 1. Example for ping:
bash -c "ping 1.2.3.9999 -c 1; exit 0"
And now you can e.g. get output of ping:
output=`bash -c "ping 1.2.3.9999 -c 1; exit 0"`
Of course instead of ping ... You can use any command(s) - including git commit.
Comments
You can use the Text-finder Plugin. It will allow you to check the output console for an expression of your choice then mark the build as Unstable.
1 Comment
Ben
this looked promising, but for some reason it kept failing the build.
For multiple shell commands, I ignores the failures by adding:
set +e
commands
true
1 Comment
Raúl Salinas-Monteagudo
I discourage unsetting -e in general. If you want to ignore some specific command's return value, you can add "|| true" or something more meaningful returning true, such as: stop-service.sh || echo Service was already down
If you put this commands into shell block:
false
true
your build will be marked as fail ( at least 1 non-zero exit code ), so you can add (set +e) to ignore it:
set +e
false
true
will not fail. However, this will fail even with the (set +e) in place:
set +e
false
because the last shell command must exit with 0.
Comments
The following works for mercurial by only committing if there are changes. So the build only fails if the commit fails.
hg id | grep "+" || exit 0
hg commit -m "scheduled commit"
Comments
Another one answer with some tips, can be helpful for somebody:
remember to separate your commands with the following rule:
command1 && command2 - means, that command2 will be executed, only if command1 success
command1 ; command2 - means, that command 2 will be executed despite on result of command1
for example:
String run_tests = sh(script: "set +e && cd ~/development/tests/ && gmake test ;set -e;echo 0 ", returnStdout: true).trim()
println run_tests
will be executed successfully with set -e and echo 0 commands if gmake test failed (your tests failed), while the following code snipped:
String run_tests = sh(script: "set +e && cd ~/development/tests/ && gmake test && set -e && echo 0 ", returnStdout: true).trim()
println run_tests
a bit wrong and commands set -e and echo 0 in&& gmake test && set -e && echo 0 will be skipped, with the println run_tests statement, because failed gmake test will abort the jenkins build. As workaround you can switch to returnStatus:true, but then you will miss the output from your command.
2 Comments
Alexander Stohr
context is everything - and if you don't want to capture console output, then you can even make that sh() function return (instead of failing for a non-0) the status code. sometimes you want both, then you might want to pipe the command outputs to a file - or hope for some simple solution to do so with the status code. ;-)
Sysanin
@AlexanderStohr true fact! Can't disagree
For Windows node can be done the following trick:
command_that_fails || set errorlevel=0
Yes, I see #linux and #unix tags, but Jenkins also runs on Windows and there ignoring errors can also be very useful :)
shstep to continue on failure usereturnStatus: true(the exit code is returned if you need to evaluate it).