Please find attached a patch.
This patch introduces profiles to select which line of Apache Hadoop you wish to build against. By default, it will build against 0.20.205.0.
Flume will not build yet against Hadoop 0.23 until
HADOOP-7937 gets resolved. But this will not require any change on Flume side. Hadoop 0.23 is still in development, hence the -SNAPSHOT version
- HDFSTextFormatter.java has to be updated to work against both Hadoop 0.20.20X and Hadoop 0.23. Text::getBytes() still returns the raw byte array, but now Text is also doing some preallocation. So the byte array would be different from what would be expected
- TestHDFSEventSink.java had at least one of its assert parameters in the wrong order (got, expected) versus (expected, got)
Don't hesitate to let me know if you would like me to rework any part