without any serious benchmarking, they seem to offer comparable speedup. trivial apps will save a second or two of jvm startup, but larger ones will see that gain erased by memory lost to leaked classloaders.
- copy nailgun-server-0.9.1.jar and maven-nailgun-1.0-SNAPSHOT.jar into maven's /lib/ext/
- in maven's /bin/, make copies of
mvn
andm2.conf
(mvn-ng-server
andm2-ng.conf
) - in
mvn-ng-server
change"-Dclassworlds.conf=${M2_HOME}/bin/m2.conf"
to"-Dclassworlds.conf=${M2_HOME}/bin/m2-ng.conf"
- in
m2-ng.conf
changeorg.apache.maven.cli.MavenCli
tocom.github.nigelzor.maven.nailgun.Server
- start the server with
mvn-ng-server localhost:2113
- whenever you would call
mvn foo
, now useng com.github.nigelzor.maven.nailgun.Client foo