Thanks a lot Martin for the analysis I'll downgrade javadoc plugin version to perform release :)
On Wed, 11 Dec 2019 at 05:06, Martin Desruisseaux < [email protected]> wrote: > Hello Maxim > > Le 10/12/2019 à 18:30, Maxim Solodovnik a écrit : > > Here is the pom https://github.com/apache/openmeetings/blob/master/pom.xml > > The command is `mvn clean install -PallModules,rc -DskipTests=true` > > I cloned the project and tested on my machine. Downgrading > maven-javadoc-plugin from 3.1.1 to 3.0.1 (by editing line 42 in the root > pom.xml) seems to fix the problem - no need to exclude Jakarta anymore > neither. I opened one module javadoc in Firefox and it seems fine. > > The reason why a tried to downgrade maven-javadoc-plugin is because I > looked at the openmeetings-util/target/apidocs/options file created by > maven-javadoc-plugin:3.1.1 and found options there a bit surprising. It > contained an --module-path option with a very long list of JAR files > (presumably all dependencies), even if many of those files where not Jigsaw > modules at all. By contrast, older maven-javadoc-plugin versions used the > traditional --class-path option. I presume that --module-path enforces > the Jigsaw checks, which include verifying that there is no package > collision, while --class-path continues to have the behavior it always > had (to not check for collision). Using --module-path makes senses for > dependencies that are Jigsaw modules, since the developers of that > dependency have been forced by javac to avoid package collision. But I > find surprising that maven-javadoc-plugin puts everything under > --module-path including dependencies that are known to not be ready yet > for Jigsaw. > > Another surprising behavior is that maven-javadoc-plugin:3.1.1 also > created a --patch-module option, "patching" openmeetings-util with > apparently all direct dependencies (i.e. asking Jigsaw to treat all > dependencies as if they were part of the openmeetings-util module > itself). It is not clear to me why maven-javadoc-plugin does that. Maybe > it results from the work done for supporting Maven aggregated javadoc ( > https://issues.apache.org/jira/browse/MJAVADOC-449). > > Martin > > > -- WBR Maxim aka solomax
