Community
Participate
Working Groups
Currently, we in Orbit only use the "previous release" in our p2 mirroring comparator to avoid changing bundles that haven't changed since previous release. While this satisfies the main requirement, it would be better to also use previous I builds (say, by using a comparator composite) so avoid churn I-build to I-build for "new" bundles.
This was fixed for I-builds some time ago. But, I did not account for M-builds. (This is probably part of the reason for bug 487833). For the primary file involved, see "getLatestIBuild.sh". I should "pass in" buildType to that file, whereas now it is hard coded to look for I-builds. Note: at times, such as when an M-branch is first created, it may take "copying" an I build or even released build to the right location ... or, provide an "override" value.
I've changed the filename to getLatestBuild.sh and pass in "$buildType". There is a possibility $buildType is not inherited at that point in the scripts, so I put a "fail unless=builtType" in the calling program "buildutilities.xml".
I should also mention I tagged the builder version and updated the value in maps build.cfg file -- in both HEAD and in Mars_maintenance. Trying an M-build first.
Verified in the previous M-build on build machine (M20160215191726) the reference repository was incorrectly an I-build. [orbit-set1] mirrorArtifacts: [orbit-set1] [echo] reference repository: /home/data/httpd/download.eclipse.org/tools/orbit/downloads/drops/I20160209143415/repository but on a re-build test on the build machine (M20160221192158) it is correct. [orbit-set1] mirrorArtifacts: [orbit-set1] [echo] reference repository: /home/data/httpd/download.eclipse.org/tools/orbit/downloads/drops/M20151221205849/repository So this bug is 'fixed'. Not sure of implications for "correctness" of repository. Am running the 'verifydir.sh' script on it now to make sure everything at least verifies.