Community
Participate
Working Groups
canonical: org.eclipse.update.feature,org.eclipse.equinox.core.sdk,3.6.1.R36x_v20101007-7sF99FsLFSyAVFKfV4c7B9BF Difference found for canonical: org.eclipse.update.feature,org.eclipse.equinox.core.sdk,3.6.1.R36x_v20101007-7sF99FsLFSyAVFKfV4c7B9BF between file:/builds/transfer/files/updates/3.6.x/M20101103-0800 and file:/builds/M201011100800/org.eclipse.releng.eclipsebuilder/../src/reposource/ The feature has a different number of entries/ packed: org.eclipse.update.feature,org.eclipse.equinox.core.sdk,3.6.1.R36x_v20101007-7sF99FsLFSyAVFKfV4c7B9BF Difference found for packed: org.eclipse.update.feature,org.eclipse.equinox.core.sdk,3.6.1.R36x_v20101007-7sF99FsLFSyAVFKfV4c7B9BF between file:/builds/transfer/files/updates/3.6.x/M20101103-0800 and file:/builds/M201011100800/org.eclipse.releng.eclipsebuilder/../src/reposource/ The feature has a different number of entries/
I assume this is because we added a new fragment for the launcher?
There actually appears to be a bug in the jar comparator and this is a false positive. The bug comes from the multiple entries in the feature.xml for the same bundle with different platform filters. (macosx is included 3 times for ppc, x86, and x86_64). These entries get added to a set and the equals for a FeatureEntry does not account for the platform filter so we "lose" 2 of the entries. Kim, can we just leave the version alone or should we still retag?
I could just exclude this feature from the comparator when we mirror the repo.
*** Bug 344915 has been marked as a duplicate of this bug. ***
(In reply to comment #2) > There actually appears to be a bug in the jar comparator and this is a false > positive. Who owns the comparator? It would be good to look at this.
Andrew owns the comparators though iirc Olivier did provide the jar comparator.
Created attachment 195439 [details] patch I'm not comfortable changing the equals/hashCode for FeatureEntry at this point in the release because it is widely used throughout PDE/Build and p2.publisher. Instead I created a wrapping object that will consult the os/ws/arch for comparison purposes. A similar change was required in PDE/Build's FeatureGenerator for testing.
Patch needs some review.
Thank you DJ.
The comparator test no longer failed in I20110514-0800.
Unfortunately it failed again in I20110517-0800: http://download.eclipse.org/eclipse/downloads/drops/I20110517-0800/buildlogs/comparatorlog.txt
The releng build appears to be running on org.eclipse.equinox.p2.repository.tools_2.0.100.v20110419-0147.jar the fix is in v20110510 I believe Kim will need to update the org.eclipse.releng.basebuilder to RC1 for the test to start passing without retagging because the test is just parsing log files created by the releng builder and not actually doing the comparison itself.
(In reply to comment #10) > The comparator test no longer failed in I20110514-0800. So this was just luck :-(
Yes, I'm doing that today. I was away from work for the past two days.
Looks like this works now. Thanks!
Verified in I20110523-0800.