Dear Loyal Readers,
Everyone uses Jenkins. It is like Pringles on any project; once you start using it, who doesn’t want a nice framework/web UI for scheduling and polling tasks!?
My first use of Jenkins was a simple CI build where every job simply executed the classic Makefile:
make make test make install
But as we started using Jenkins for more and more complicated tasks, and especially for push-button deployments, our execution shell block got longer and longer. There are some cool plugins to handle more complicated jobs- This one saves all config changes internally, and this one syncs the config changes to a git repo.
But this is Jenkins, so why not just wrap up our shell commands into a single script and put it in a repository? That way it will be tracked in source control and not in the Jenkins config.xml.
We ran into one problem. We lost my favorite Jenkins feature: every line error checking. Jenkins will fail at the FIRST error. In Jenkins, an execute-shell block like this:
echo "Starting" false echo "Finishing"
would fail:
+ echo Starting Starting + false Build step 'Execute shell' marked build as failure Finished: FAILURE
But once I put that all in a single script, we lost the error checking:
+ ./runme.sh Starting Build We should never get here Finished: SUCCESS
For deployments, this is an important feature we don’t want to lose. If any preceding command fails, we need to stop and fail immediately. We now have to write a bunch of ugly bash error checking to replace the whole execute block with “./build_me.sh” …or do we?
#!/bin/bash -e
I was so happy to find this bash flag: -e
This shebang line option tells bash to return non-zero at the first sign of trouble. Now all my Jenkins scripts die as soon as there is a problem 🙂
There is only one step left that all ancient bash nerds know: -x
The “-x” flag is bash’s debug mode that _echo_s every command. Now my scripts can be saved from that extra crappy bash error checking and those extra _echo_s I would have to add.
