In message <201005271258.o4rcwkd7006...@d06av02.portsmouth.uk.ibm.com>, Mark Hindess writes: > > In message <201005271025.o4rappbb016...@d06av03.portsmouth.uk.ibm.com>, > Mark Hindess writes: > > > > [snip] > > > > However, the federated build for -Dhy.select=true currently fails > > because, for example, the "serialver -show" implementation requires > > some modules that aren't included in select. > > > > I think we should consider how/what we want to subset from the jdktools > > component? I think we need at least jdwp. Some of the tools - jar > > for example - are thin wrappers around implementations that are in > > the select scope so I think it makes sense to include them too. > > > > What do other people think? > > > > While discussion continues, I intend to create a first approximation > > using the subset of all of jdktools that builds as I'd like to get > > select builds working on Apache Hudson. > > I think I've fixed things so that the federated build should complete > with -Dhy.select=true now. I'll look at setting up Hudson builds > shortly. > > Something else we should do before we make milestone releases is > change the "java -version" output to make it obvious that select > is being used.
I've created a select build on hudson (java5 only at the moment): http://hudson.zones.apache.org/hudson/view/Harmony/job/Harmony-select-1.5-head-linux-x86_64/ The resulting artefacts are ~18% smaller. It occurs to me that if we move the font dependency handling to the awt module then we'd reduce the jre size by almost another 4% from the non-select size. Regards, Mark.
