Looking at the Clirr violations, it seems that the problem is a change
in the number of parameters to various method in the generated parser
classes:
[ERROR]
org.apache.maven.artifact.repository.metadata.io.xpp3.MetadataXpp3Reader:
In method 'public boolean getBooleanValue(java.lang.String,
java.lang.String, org.codehaus.plexus.util.xml.pull.XmlPullParser)' the
number of arguments has changed
[ERROR]
org.apache.maven.artifact.repository.metadata.io.xpp3.MetadataXpp3Reader:
In method 'public java.util.Date getDateValue(java.lang.String,
java.lang.String, org.codehaus.plexus.util.xml.pull.XmlPullParser)' the
number of arguments has changed
IMO, these methods are meant to be private, and shouldn't show up at all
in Clirr checks. Using these methods on the generated parser classes is
an abuse of the spirit of that class (i.e. parsing a model, not value
coercion), so there shouldn't be any reason for these to need to be
public...unless I'm missing something.
Since you mention adding a flag to Modello in the future (beyond
-alpha-19, I'm guessing) to allow us to set these sorts of methods to
private, I guess I'm confused as to what you're planning to add to
-alpha-19 to fix the above generated compatibility problems.
IMO, it's not a major problem to simply skip Clirr checks on these
generated classes for now...then, in the next release, I expect these
methods won't have changed again, so we could remove the exclusions.
Am I missing something?
-john
Dennis Lundberg wrote:
Brian, I was planning to upgrade the dependency on Modello in the
core, because of a binary incompatibility [1] [2]. I was waiting for
the build to fail in CI before I fixed it, but it doesn't seem to have
failed yet.
The change I was going to make would fix the Modello issue that you
documented in the root pom.xml in [3]. I have tweaked Modello so that
1.0-alpha-19-SNAPSHOT now includes the missing helper methods, even
though they should be private. I'll get stared on getting that version
of Modello released.
On a side note, we should make it possible to tell Modello to make
these helper methods private before we start pushing out 2.1 releases.
[1] https://svn.apache.org/viewvc?view=rev&revision=674674
[2] https://svn.apache.org/viewvc?view=rev&revision=674666
[3] https://svn.apache.org/viewvc?view=rev&revision=675380
Brian E. Fox wrote:
It looks like we've got a good set of issues fixed for 2.0.10 and things
are starting to stabilize. We'll start publishing 2.0.10 RC's as early
as today. I think we worked out a good process with the 2.0.9 release
and we should continue in that direction. The basic principles are:
1) we will stop to fix any regressions between 2.0.9 and 2.0.10
2) Nothing else should be included once we start this process
The initial RCs will be posted to the dev list, but naturally everyone
is welcome to try them. After we get comfortable with the stability of
the RCs here, I'll again involve the user list for broader testing of
the system. This is where we found the majority of the serious issues
last time.
The issues chosen for 2.0.10 were primarily regressions identified over
earlier versions and the next highest voted issues that could be solved
without risking further destabilization and regressions.
Because we expect to have several iterations, the RCs will be tagged as
RCs to avoid us having to rewind all the versions back many times. Once
the final RC has stabilized, we will redo the release a final time so
that the version is correct. This build will be the one voted on and
hopefully ultimately released.
Thanks,
Brian
--
John Casey
Developer, PMC Member - Apache Maven (http://maven.apache.org)
Blog: http://www.ejlife.net/blogs/buildchimp/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
