[WIP][JENKINS-26583] Env-inject suppress variables contributed by extension points

[olivergondza]

JENKINS-26583

[jenkinsadmin]

Thank you for a pull request! Please check this document for how the Jenkins project handles pull requests

[lanwen]

please don't hide validation process in "mistic guest" this is antipattern of test writing. Use hamcrest matchers :) they cool

[olivergondza]

@lanwen, I have addressed your concerns. Using hamcrest for validation* would require bumping the inherited version.

*) there are no Matcher#describeMismatch( or TypesafeMatcher#describeMismatchSafely) so it would actually made defect localization weaker.

[lanwen]

@olivergondza thanks for quick reply!
Think junit version can be bumped to 4.12 without any doubt. This version is stable and used in production testing half a year. Test should use all features of new version of this lib and hamcrest lib

[oleg-nenashev]

I would be also happy to see it merged

[olivergondza]

The fix does not work is it should, I did not managed to untangle the injection logic IIRC.

[dummy-lanwen-bot]

SonarQube analysis reported 10 issues:

• 10 major

Watch the comments in this conversation to review them.
Note: the following issues could not be reported as comments because they are located on lines that are not displayed in this pull request:

• Add a message to this assertion.
• Add a message to this assertion.
• Add a message to this assertion.
• Add a message to this assertion.
• Add a message to this assertion.
• Add a message to this assertion.
• Add a message to this assertion.

[jukkasi]

+1

We use EnvInject plugin and Delivery Pipeline plugin, which uses Build Name Setter Plugin which sets $BUILD_DISPLAY_NAME.

When $BUILD_DISPLAY_NAME is used in HipChat plugin message template, the value is not correct, uninstalling EnvInject plugin solves the problem.

[oleg-nenashev]

Deleting garbage comments from @dummy-lanwen-bot

[oleg-nenashev]

+1 for integrating it into EnvInject-2

[DanielRedOak]

I think we're seeing this problem in combination with the rvm plugin. GEM_PATH ends up with pre-rvm init values so it doesnt work. The job isnt using any envinject features, but it is still causing issues.

[oleg-nenashev]

I will check if I can integrate it towards the 2.1 release

[fiznool]

The 2.1 release came and went without this getting merged... any plans on getting this in to a future release? It's blocking my use of the nodejs plugin 😞

[oleg-nenashev]

@fiznool There is a plan for it, but it is not trivial to fix it without breaking changes in other usa-cases. Right now I do not have so much spare time to work on EnvInject (the plugin is waiting for adoption), and effectively this change priority is being defined by the internal priorities of the company I work for. These priorities shift sometimes, hence I cannot commit on the ETA.

If somebody proposes a fix, I will be happy to review it.

[fiznool]

Thanks for the update. I know how hard it is to maintain OSS. Hope you find some time to carry out the necessary work at some point soon!

[oleg-nenashev]

Created #124, which really fixes some bits

[oleg-nenashev]

Closing in favor of #125, which offers a partially working fix covering JENKINS-26583.
2 tests created by @olivergondza still fails, but I think they can be fixed later since the fix logic is quite different (see #124).