I've now traced "> By looking at the http trace more I realised they were showing an older 1.4 version of apache ivy was getting picked up from my Eclipse classpath. Fully removing these so that it picked up v2.1 correctly made it go back to the original speed.
Thanks, Andrew On Wed, 04 Nov 2009 11:45:00 +0000, Andrew Ormerod wrote: > Hi, > > We've just upgraded our application to use ivy 2.0 (from ivy 1.x). As part > of it we're seeing that the time taken to resolve versions is much larger > (overall ant task goes from ~30 seconds to approximately two and half > minutes). > > From enabling the triggers logging & looking at the http/tcp trace traffic > it seems that the slowness is caused by the version matching that Ivy 2 is > trying to do. We have our modules setup to match on versions like 3.2.+ or > 4.2.+ but I can see that http requests are happening on versions that could > never match (e.g. 0.10.5). If I change to have exact matches the > performance returns to the ~30 seconds region. (In each case no actually > new binaries / revisions are downloaded as they are in the local cache) > > When this resolve behaviour is combined with the repository on the other > end of a WAN & large numbers of published builds it seems that this takes a > long time and significantly slows things down. > > Has this been seen before? (From googling, I > found a couple of "close but > not quite" matches). > Any suggestions on how to change this (ideally in config)? > > Thanks, > > Andrew
