Yup, the compute class is different because the original one was doing done some funny things around the presence of examples source code and didn't work without them. Hence, I made the change to make sure that that cluster can be started, etc. Keep in mind, that runtime has been significantly changes recently (e.g. around 0.8.1) and you can see quite a bit of differences coming from it along.
Contributing needed changes back to the open source will be a great idea indeed, especially considering that past 0.9.0 there likely to be much less fluctuations in the runtime environment. Cos On Tue, Jan 21, 2014 at 04:54PM, Mark Grover wrote: > Hi all, > I was poking around Bigtop's integration with Apache Spark and came across > our compute-classpath.sh (see > https://github.com/apache/bigtop/blob/master/bigtop-packages/src/common/spark/compute-classpath.sh) > and noted it is significantly different from the compute-classpath.sh > bundled upstream ( > https://github.com/apache/incubator-spark/blob/branch-0.8/bin/compute-classpath.sh > ). > > Does anyone (maybe, Cos?) know what the rationale was having a separate > copy of compute-classpath.sh in Bigtop? Also, if you have any ideas on how > we could contribute this upstream so we can decrease our divergence here, > I'd love to talk more. > > Thanks in advance! > Mark
