At maven dev's request, added a maven jira issue for this: https://issues.apache.org/jira/browse/MNG-6503
On 11/1/2018 11:28 AM, Marshall Schor wrote: > found out that the trivial test case does work in older maven v 3.3.9. > > fails in 3.5.4 and 3.6.0. > > -Marshall > > On 11/1/2018 9:35 AM, Marshall Schor wrote: >> I made a trivial test case and reproduced this. >> >> I posted this test case with a question to the maven-users list. >> >> -Marshall >> >> On 10/31/2018 5:16 PM, Marshall Schor wrote: >>> Something strange broke. >>> >>> If you remove from the repo the snapshot of the uimaj-parent pom, and then >>> try >>> something >>> >>> cd [your dir where the top-level uimaj directory is] (the top level of >>> the >>> uimaj 2.10.3-SNAPSHOT checkout, for example), and then do >>> >>> mvn dependency:tree (just some maven thing to run without changing >>> anything) >>> >>> it fails, saying it can't find the uimaj-parent pom, even using the >>> relative-path. >>> The top level finds it via the relative-path ref, but the other modules >>> don't >>> (get 2 failures: for the aggregate-uima, and uimaj-json projects) >>> >>> I probably have done something stupid. >>> >>> I can make it work if I run mvn install on the parent-pom, and then go back >>> and >>> redo this - it then finds the uimaj-parent pom reliably in the local .m2. >>> >>> This makes "releasing" difficult: >>> -- I have to insure the local .m2 has the build for uimaj-parent of >>> 2.10.3-SNAPSHOT before running release:prepare >>> -- I have to run a mvn install on the 2.10.4-SNAPSHOT version of the >>> uimaj-parent after release:prepare, and before release:perform. >>> >>> Any clues to where I messed things up? >>> >>> -Marshall >>> >>>
