Skip to main content
aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorThomas Watson2020-01-13 15:53:33 +0000
committerThomas Watson2020-01-14 22:32:15 +0000
commitb61cf1d96b3e7ef8da2e506f0393d91e422dbcfd (patch)
tree7bbfc8eb49ce1693969e2609cfdfa691556861c5 /launcher-binary-parent
parent4d9b3a851bc877af780577fa4f096976399c121b (diff)
downloadrt.equinox.framework-b61cf1d96b3e7ef8da2e506f0393d91e422dbcfd.tar.gz
rt.equinox.framework-b61cf1d96b3e7ef8da2e506f0393d91e422dbcfd.tar.xz
rt.equinox.framework-b61cf1d96b3e7ef8da2e506f0393d91e422dbcfd.zip
Bug 559118 - Add testcase and fix reexport uses constraint checkI20200116-1800I20200116-0930I20200116-0330I20200116-0310I20200115-1800
This is a testcase that reproduces the issue found in bug 558895. When an exporter of a split package requires multiple bundles that also export the split package and that bundle does a reexport on one of the parts it can cause resolution issues for uses constraints. The actual problem is with reexport itself. The reexporting bundle does not have to also export the package to cause the issue. The problem is that each part of the package pulled in from the require-reexport is checked in isolation with the using bundles wire to the same package name. This is incorrect because that is only a subset of the actual used package from the perspective of the exported package that is using the split package. The fix is to record all the package parts for the split used package from the wiring of the bundle exporting the package that uses the split package. That way during the compatibility check we can accurately use the set of sources for the split package that the exporting bundle is using. Change-Id: I5d6194adabc7c04fe990d663ad1dd6bb77f2ac39 Signed-off-by: Thomas Watson <tjwatson@us.ibm.com>
Diffstat (limited to 'launcher-binary-parent')
0 files changed, 0 insertions, 0 deletions

Back to the top