Re: [orbit-dev] How can Orbit bundles change when their classifiers do n

[Roland Grunberg <rgrunber@xxxxxxxxxx>]

On Mon, 2017-12-18 at 17:11 +0100, Gunnar Wagenknecht wrote:
> > On Dec 18, 2017, at 16:12, Andreas Sewe <andreas.sewe@xxxxxxxxxxxxxx> wrote:
> > How can this be?
> 
> It depends on the nature of the change. The qualifier is derived from Git. If there are no changes in Git the qualifier should stay the same.
> 
> However, rebuilding a bundle could cause different bits being generated. Therefore the p2 mirror task should be used when publishing bundles into an existing repo (likely a composite). It has a comparator that only publishes bundles not already present in the composite. The comparator is able to ignore difference caused by re-builds.
> 
> I'm not sure if we use the comparator for delta publishing. I believe Orbit is just publishing the latest repo as is.

For the javax.xml_1.3.4.v201005080400.jar , I see a difference occuring
between Luna SR2 and Mars. However when looking at the actual changes
there, they are completely different from the 0800 -> 0808 changes in
the eclipse-java-oxygen-2 archives. In this case I'm not sure where
that new jar is coming from as it doesn't match anything released from
Orbit.

As for the org.apache.lucene.queryparser_6.1.0.v20161115-1612.jar, the
timestamps seem to indicate the build was still using a version from
Oxygen M4, and has now updated to Oxygen.0. 

The changes seem to be a combination of the pom.properties timestamp
comment always being updated on rebuild, and a change that occured in
our build plugins (likely ebr-maven-plugin) which affected how the
manifest was generated.

This should be the next thing I to look at, in order to have stronger
guarantee of identical content.


Cheers,
-- 
Roland Grunberg