This sounds VERY much like the problem Richard Vowles reported awhile back which is solved by ungrading Maven to use Aether 1.12.
The problem lies in having some artifcts downloaded directly, some from a mirror. Maven generates a metadata file which includes a references to the upstream repository id used originally to fetch the artifact, and when you run a project that doesn't match that - the artifact gets reported as not being found - even if it is. ( at least, thats my simple summation ). http://jira.codehaus.org/browse/MNG-5084 was the ticket that Richard raised. Sadly it looks like no one on the Apache team has responded to either Richard or my queries about this problem, tho I did discover that Aether 1.12 actually fixes the problem - unfortunately due to the EPL license change ( see other threads ) I'm not sure when we will see this fixed. Jason van Zyl has made available a distro/packaging of maven with additional fixes ( see his thread about that for details ) , including this Aether release - I wonder if you could try that and see if this solves the problem you're seeing? On 30/07/2011, at 2:22 AM, Mark Struberg wrote: > This test fails since the last commit but I actually think it only ever > worked by accident. It did rely on a commons-logging-1.0.4.pom which got > resolved by a test which did run previously. > But after updating to commons-logging.1.1.1 there is no pre-resolved artifact > available in target/local-repo anymore, thus the javadoc link info cannot get > built and the test fails in line 829: >> assertTrue( optionsContent.contains( "-link >> 'http://commons.apache.org/logging/apidocs'" ) ); >
