[
https://issues.apache.org/jira/browse/JENA-1830?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17025025#comment-17025025
]
Andy Seaborne edited comment on JENA-1830 at 1/28/20 10:36 AM:
---------------------------------------------------------------
{{travis_retry}} is reasonable if there are only a few jdks. I'll watch to see
the causes of failures; from a quick scan, it looks like it is often the same
as Jenkins - not contacting github - but also having maven setup replicated to
the build node properly.
was (Author: andy.seaborne):
{{travis_retry}} is reasonable if there are only a few jdks. I'll watch to see
the causes of failures; from a quick scan, it looks like it is often the same
as Jenkins - not contacting github as well as not having maven setup properly
available.
> travis.yml
> ----------
>
> Key: JENA-1830
> URL: https://issues.apache.org/jira/browse/JENA-1830
> Project: Apache Jena
> Issue Type: Improvement
> Reporter: Andy Seaborne
> Priority: Minor
>
> Jena has several JVM versions in the Travis setup. In addition, the Jenkins
> server has full build and test runs.
> I use Travis (free option) to verify development changes but the build
> process is not perfect. A complete cycle is 5 travis builds and if any one
> fails, the job is flagged in error. The errors are not Jena related - they
> are build environment issues reflected as maven build problems.
> In the last 50 jobs without development failures, I got 24 green, 22 failed
> one or two jobs and 4 build system fails. So I get 40% false reports of code
> failures.
> The individual fails were distributed as:
> Java8 : 3
> Java9 : 3
> Java10: 4
> Java11: 2
> Java12: 10
> The problem is that with 5 fairly reliable jobs, the probablity of a "failed"
> build is increased. This is semi-interactive feedback, unlike Jenkins which
> is a daily set of builds.
> I propose having LTS (8,11, and soon 14).
> Jenkins will provide the more complete testing.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
