[
https://issues.apache.org/jira/browse/MJAVADOC-431?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14943681#comment-14943681
]
Peter lynch commented on MJAVADOC-431:
--------------------------------------
[~michael-o] The javadoc plugin accepts a generic archiver configuration for
the javadoc-jar goal, yet silently prevents users from adding the descriptor as
part of that configuration. This is an abuse of the generic archiver
configuration.
The maven descriptor files are located at a well known path and contains well
known contents, a properties file and pom. The existence of the GA specific
paths in the javadoc-jar archive can help identify the GA of the artifact
without needing the original pom or original jar. There are cases I need to
know the GA, but I do not have the 'main' jar or the pom separately.
There is no easy way to automatically generate a maven descriptor like path or
file and put this into the javadoc jar, without polluting the source tree with
dummy class files which exist only generate packages at certain GA paths.
Adding complexity to a build process to generate such files is overkill and
needless. Even then we still can't put these under META-INF to indicate this is
meta information only. Certain organizations have strict policies around adding
dummy class files with no operational purpose. All that is needed is to fix the
bug where the archiver configuration is not respected when configured by plugin
configuration - a single line to the javadoc plugin configuration - done.
The javadoc and source plugin are the only two core plugins known to do this,
yet there is no good reason documented. I suggest it was one plugin copied the
other. If there was a good reason, I suggest the archiver configuration library
should go to greater lengths to not have a generic option to add the descriptor.
The existing forced, non-overridable blocking of the descriptor is confusing
and not-consistent with standard archiver configuration use in the very least.
The patch preserves backwards compatibility of the default behaviour just in
case we cannot foresee a good reason why it was done this way.
I can accept a plugin chooses defaults for the archiver configuration, but
unless we can cite a solid reason not to respect user configured archiver
configuration, like it breaks well established Maven tooling which would rely
on it only existing except in the main jar ( even then one could argue it is at
user discretion ), I urge this patch to be accepted.
The source plugin has already moved ahead with a patch to fix this. See
MSOURCES-81.
> allow javadoc jar to contain Maven descriptor
> ---------------------------------------------
>
> Key: MJAVADOC-431
> URL: https://issues.apache.org/jira/browse/MJAVADOC-431
> Project: Maven Javadoc Plugin
> Issue Type: Bug
> Affects Versions: 2.10.3
> Reporter: Peter lynch
> Fix For: 2.10.4
>
>
> The javadoc:jar mojo explicitly prevents the Maven descriptor from being
> added to the produced javadoc.jar file.
> https://github.com/apache/maven-plugins/blob/trunk/maven-javadoc-plugin/src/main/java/org/apache/maven/plugin/javadoc/JavadocJar.java#L299-299
> I could not find an explanation or technical reason why this is done.
> Adding the maven descriptor to the javadoc jar can help expose valuable
> information about the build that produced it and should be at the discretion
> of the build process.
> Expected:
> - allow the archiver used to create the javadoc jar to respect the plexus
> archiver configuration if it is configured to include the Maven descriptor.
> - preserve the default behaviour of not including the Maven descriptor, for
> (unknown) backwards compatibility reasons only
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
