Todd Lipcon wrote:
- Whenever a task will need to touch both Common and one of the components (Mapred/HDFS) should there be two JIRAs or is it sufficient to have just one "HADOOP" JIRA with separate patches uploaded for the two repositories?
Two Jiras, I think. In the long run, such issues should be few. E.g., we should not be changing the FileSystem API incompatibly much.
- If we're to do two separate JIRAs, is the best bet to use JIRA's "linking" feature to show the dependency between them?
+1
- When we have one of these cross-project changes, how are we supposed to do builds?
I talked with Owen & Nigel about this, and we thought that, for now, it might be reasonable to have the mapreduce and hdfs trunk each contain an svn external link to the current common jar. Then folks can commit new versions of hadoop-common-trunk.jar as they commit changes to common's trunk. We'd need to remove or update this svn external link when branching. Thoughts?
Doug
