Merge lp://staging/~twom/ols-jenkaas/junit-for-sca into lp://staging/ols-jenkaas

I think the indentation is a bit off, and you'll need to allow empty results, otherwise if for some reason the subunit data is not generated, the job will fail, which is annoying.

Example from the docker projects (and this is tested working):

=== modified file jobs/project-docker.yaml
--- jobs/project-docker.yaml 2018-02-28 17:22:12 +0000
+++ jobs/project-docker.yaml 2018-03-13 15:55:57 +0000
@@ -59,6 +59,10 @@
             property-file: build-params.prop
             node-parameters: true
             condition: 'UNSTABLE_OR_BETTER'
+ - junit:
+ results: '{workdir}/reports/*.xml'
+ allow-empty-results: true
+

Also - I see "results" in your job specification, whereas the Makefile outputs to "reports" (which we said is good because it's also already .gitignored). so that'll need an update too.

+1 thanks!

Sadly this didn't work :( and the same thing happened to me so I should have known about this.

Publisher configuration can't be put in a project definition; it needs to be put in the job-template definition.

The problem-ish here is that adding the publisher to the widely-used {name} job-template, will cause every single code merge job to have a junit publisher.

It's not a huge issue though because allow-empty-results: true will simply do nothing if there're no reports for a specific project.

Still, I didn't want to just apply the change without consulting people first, given the work you've done on this.

(Of course, there may be another way of doing this with jenkins-job-builder which I'm missing).

The diff that'll move the junit publisher to the global landing job template is here:

https://pastebin.canonical.com/p/BNksFgNxTG/