Vincent Siveton wrote:
Unfortunately, Huson is always unhappy...
Yep, but this time it left us some nice debug infos [0]: The generated
"options" file specifies "-charset 'ISO-8859-1'" but misses
"docencoding". This is an indication that the IT build is using an
outdated version of the plugin which doesn't contain Hervé's
enhancements. Looking at the build output [1] of the first IT that was
run and filled up the IT repo, one can spot the line
Downloading:
http://localhost:8081/content/groups/public/org/apache/maven/plugins/maven-javadoc-plugin/2.5-SNAPSHOT/maven-javadoc-plugin-2.5-20080726.211140-49.jar
122K downloaded
i.e. it's really getting the last publically deployed snapshot via Nexus
instead of using the snapshot staged in the IT repo created by the
Invoker Plugin.
My current state of belief is that this caused by MINVOKER-43 which I am
about to change to be of type "major bug". The Invoker Plugin 1.2 is
using the repository id "it-repo" when staging the plugin into the IT
repo. This will create "maven-metadata-it-repo.xml" files in the IT
repo. However, when the forked Maven IT builds access the IT repo as a
local repository, Maven will use the id "local" to lookup metadata. I.e.
its looking for "maven-metadata-local.xml" and misses
"maven-metadata-it-repo.xml" which finally makes Maven pull the plugin
from the remote repos because it appears to be non-existent locally.
What I personally can't understand is why the timestamp of the staged
artifact (which is close to now) is effectively ignored and a two-day
old version is used instead. But that seems like another story ;-)
Benjamin
[0]
https://ci.sonatype.org/job/Maven-Plugins/ws/trunk/maven-javadoc-plugin/src/it/MJAVADOC-206/encoding/target/site/apidocs/options/*view*/
[1]
https://ci.sonatype.org/job/Maven-Plugins/ws/trunk/maven-javadoc-plugin/src/it/MJAVADOC-110/build.log/*view*/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
