[ODLPARENT-75] Feature repositories should aggregate (as opposed to redefine) single features. Created: 03/Mar/17  Updated: 24/Jan/18  Resolved: 27/Apr/17

Status: Resolved
Project: odlparent
Component/s: General
Affects Version/s: None
Fix Version/s: None

Type: Bug
Reporter: Vratko Polak Assignee: Stephen Kitt
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Operating System: All
Platform: All

External issue ID: 7900

 Description   

This change request wants each single feature descriptor to define single feature while referencing dependency descriptors, but feature repository descriptor to define zero features and referencing single feature descriptors. This implies SingleFeatureTest will be run for every feature-related pom (but ending up testing nothing for the new feature repositories).

Basically, Karaf 3 feature-index [0] (from Integration/Distribution) is an example of what the generated feature repository should look like.

The main motivation is to speed up SingleFeatureTest and avoid duplicate definitions.

The fix probably requires changes to karaf-feature-plugin.

[0] https://nexus.opendaylight.org/content/repositories/opendaylight.release/org/opendaylight/integration/features-integration-index/0.5.2-Boron-SR2/features-integration-index-0.5.2-Boron-SR2-features.xml

 Comments   
Comment by Stephen Kitt [ 06/Mar/17 ]

Does this actually speed up SingleFeatureTest ?

Comment by Vratko Polak [ 07/Mar/17 ]

> Does this actually speed up SingleFeatureTest?

For common repos, time spent testing should be the same, surefire overhead would be bigger.
But SFT on integration-index will become no-op, speeding up every distribution-check run.
Overall, time to verify will be shorter.

Comment by Stephen Kitt [ 07/Mar/17 ]

(In reply to Vratko Polák from comment #2)
> > Does this actually speed up SingleFeatureTest?
>
> For common repos, time spent testing should be the same, surefire overhead
> would be bigger.
> But SFT on integration-index will become no-op, speeding up every
> distribution-check run.
> Overall, time to verify will be shorter.

Ah yes, the integration-index gain is interesting.

Comment by Stephen Kitt [ 08/Mar/17 ]

https://git.opendaylight.org/gerrit/53016

Comment by Vratko Polak [ 24/Apr/17 ]

The fix [1] is merged, but only to master (Nitrogen).

Looking at distribution:features-index [2], we really need [1] cherry-picked to Carbon. Distribution merge carbon already takes ~74 minutes [3], and that that is for just some offset 0 feature repos.

We may test things more on Nitrogen, but I am not sure how stable are the features there.

[1] https://git.opendaylight.org/gerrit/53016
[2] https://nexus.opendaylight.org/content/repositories/opendaylight.snapshot/org/opendaylight/integration/features-index/0.6.0-SNAPSHOT/features-index-0.6.0-20170424.163643-635-features.xml
[3] https://jenkins.opendaylight.org/releng/view/distribution/job/distribution-merge-carbon/1630/

Generated at Wed Feb 07 20:27:33 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.