We are in the process of migrating from a custom in-house build system to TeamCity. I have a build configuration using the Ant Runner to run a target that in turn runs a large number of other targets (wrapping up the entire build process) including building, unit testing and running Clover. The build has a large amount of historical information that clover uses for checking that our code coverage maintains a certain level.
In our old build system, the clover report generation is very fast, taking ~2 minutes to complete the 'current' report. When running the same target within TeamCity, this has been taking over 50 minutes. I have yet to determine what is different between the two. The builds are running on the same hardware, and utilizing the same clover data. I was able to get the TeamCity configuration to complete the clover report in 18 minutes by setting the cache attribute to 1g and bumping up the number of threads for clover to 6. That is still almost 10x slower though.
I noticed that the classpath is slightly different for the TeamCity build. It has: /TeamCityAgent4/plugins/antPlugin/lib/buildServerAntRuntime.jar:/TeamCityAgent4/lib/runtime-util.jar injected into it (I'm assuming this is for TC to introspect the ant process better). Is it possible that either of these two jars is impacting the clover report generation somehow? I did not find any open issues relating to report generation, so it doesn't seem this is a common problem. Any comments will be appreciated.