On 11/21/06, David R Robison <[EMAIL PROTECTED]> wrote:
I'm attaching the log from the "ant -d". The following, from the log, looks interesting: [ivy:retrieve] using default-chain to resolve [ xmlBlaster | xmlBlaster | 1.4 ] [ivy:retrieve] default-chain: no latest strategy defined: using default [ivy:retrieve] orci: no namespace defined: using system [ivy:retrieve] pre 1.3 ivy file: using exactOrRegexp as default matcher [ivy:retrieve] found ivy file in cache for [ xmlBlaster | xmlBlaster | 1.4 ] (resolved by orci): C:\Workspaces\DataGateway\DataGateway-VDOT-OnCall-Task36\ivy\ivy-cache\xmlBlaster\xmlBlaster\ivy- 1.4.xml [ivy:retrieve] orci: found revision in cache: [ xmlBlaster | xmlBlaster | 1.4 ] (resolved by orci): but it's a default one, maybe we can find a better one [ivy:retrieve] orci: no latest strategy defined: using default [ivy:retrieve] trying http://www.openroadsconsulting.com/maven/xmlBlaster/jars/xmlBlaster-1.4.jar [ivy:retrieve] orci: no ivy file found for [ xmlBlaster | xmlBlaster | 1.4 ]: using default data [ivy:retrieve] tried no ivy pattern => no attempt to find module descriptor file for [ xmlBlaster | xmlBlaster | 1.4 ] [ivy:retrieve] checking [ xmlBlaster | xmlBlaster | 1.4 ][default] from orci against null [ivy:retrieve] module revision kept as first found: [ xmlBlaster | xmlBlaster | 1.4 ][default] from orci [ivy:retrieve] ibiblio: no namespace defined: using system [ivy:retrieve] pre 1.3 ivy file: using exactOrRegexp as default matcher [ivy:retrieve] found ivy file in cache for [ xmlBlaster | xmlBlaster | 1.4 ] (resolved by orci): C:\Workspaces\DataGateway\DataGateway-VDOT-OnCall-Task36\ivy\ivy-cache\xmlBlaster\xmlBlaster\ivy- 1.4.xml [ivy:retrieve] found module in cache but with a different resolver: discarding: [ xmlBlaster | xmlBlaster | 1.4 ] It finds it in the cache but thinks there may be a better one so it keeps looking. It eventually scans through all the configured repositories and ends up using the one in the cache. Does anyone know why it would not just use the one in the cache? Does this help any?
OK, the problem is due to a modification in Ivy 1.4 which has a bad side effect in your case. When Ivy find a default module descriptor (i.e. a module descriptor generated because there is no ivy file for a module), then in a chain it checks if there is no other element in the chain with a real ivy file. This is the cause of your performance problem I think, so the only workaround I see is to add an ivy file even for simple modules with only one artifact. I'd also suggest submitting an improvement request in jira, but jira is currently in read only mode until it is migrated to apache. Xavier Thanks, David
