Community
Participate
Working Groups
in the platform-releng we've always had this master-jetty "buildtime" feature. http://git.eclipse.org/c/platform/eclipse.platform.releng.git/tree/features/master-jetty Its one use, as far as I can tell, is to "collect" things to put into the final repository. I'd always assumed there was nothing else that put these things into the repository. And, we've always "removed" the feature itself from the final repo, leaving in the repo what it put there. But, while investigating some build failures due to (trying) to upgrade the jetty pre-req (bug 406691) I've seen there is also a org.eclipse.equinox.server.jetty feature. http://git.eclipse.org/c/equinox/rt.equinox.bundles.git/tree/features/org.eclipse.equinox.server.jetty The difference between these two features seem inconsequential ... the master-jetty includes org.eclipse.osgi but surely there's plenty of other things to include that! the org.eclipse.equinox.server.jetty includes javax.servlet (pulled from Orbit, no doubt) but the remaining 9 bundles are identical. AND ... org.eclipse.equinox.server.jetty does end-up in our repository (so everything it includes would as well). As far as I can see, we can just "delete" master-jetty feature, and the end result would be the same. Anyone recall any history here? If we don't need it, it is important to remove it, since removing things in a post-build process is bad: a) always bad since then our repo is a little different than a pure CBI build, but b) even worse, I'm learning, because the build-time "comparator" will often fail based on things being removed (but, still needed for "build time compare".
I did a test build without master-jetty. It wasn't a "perfect test" ... since I had built once, then built again with the modification (so there might have been some "old targets" left around not cleaned up) ... but, that test build did just fine, which makes me think it is worth a "real" test build, with the changes made in git repo and doing another test build from scratch. Should be plenty of time to complete before tonight's M7 warm-up I-build. I know we need to "stop changing the build" soon ... but ... under the circumstances (making such a radical change from PDE to CBI so late) I think it is worth the risk ... and could save us from many "rebuilds" later, just because the "fake feature" gets removed.
Commit for aggregator: http://git.eclipse.org/c/platform/eclipse.platform.releng.aggregator.git/commit/?id=2fa9f1a04a41c2325f0f1edcedac10248e204be0 Commit for o.e.releng: http://git.eclipse.org/c/platform/eclipse.platform.releng.git/commit/?id=858cbffd468244565dc480a8f9a295e969254b57
The test build went fine. I compared "number if IUs" in the final repo, and there was the same number, except the new one had one less (master-jetty.feature.jar). I suppose if we are going to remove some master feature group we should have been removing jar too? But ... counting as fixed.
mass change to 'verified', as these bugs are either routine or obviously fixed build breaks.
