@legal-discuss, brief recap: In Spark's test source code and release, there are some JAR files which exist to test handling of JAR files. Example: TestSerDe.jar in https://github.com/apache/spark/tree/master/sql/hive/src/test/resources/data/files
Justin raises the legitimate question: these don't belong in a source release, do they? My operating theory had been that they are more like binary blobs w.r.t. Spark, like a test JPEG or data file, and are not the compiled version of any test code in Spark. They need to exist in order to run the tests from a source release. So it's not quite a case of shipping compiled Spark code in a source release. I can imagine three opinions: 1) It's OK. 2) It's OK, but you need to include the source code to even those test JAR files somewhere 3) It's not fine, and the toolchain has to separately build these from source first automatically I found https://markmail.org/thread/nf3lsdy5m3c3ovbr on legal-discuss previously, which seems to incline towards 2. I'm also inclined towards 2, as 3 is probably relatively tricky in practice even though that's a nice-to-have. I'd welcome opinions on this one. Sean On Sat, Jun 23, 2018 at 7:34 PM Justin Mclean <jus...@classsoftware.com> wrote: > > It's not test code; test code would indeed have to be distributed as > source as well. They are binary blobs, if you like, needed by test code, > that happen to be JARs here and not JPEGs or .docx files or something. > These help test handling of JAR files. > > Which IMO is still not allowed in a source release, but as I said it would > be best for you to check on legal discuss. > >
