[Release] Address spurious Apache RAT failures in source release script

[asfimport]

Something about the untarring / retarring process is offending Apache RAT and it's reporting a number of unapproved licenses incorrectly

https://gist.github.com/wesm/490dc50c494cc914ae28b4fd897a73eb

Reporter: Wes McKinney / @wesm
Assignee: Kouhei Sutou / @kou

PRs and other links:

• GitHub Pull Request #4018
• GitHub Pull Request #4028

_{Note: This issue was originally created as ARROW-3578. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
Seems this could be caused by #2417

[asfimport]

Krisztian Szucs / @kszucs:
Steps to reproduce?

[asfimport]

Wes McKinney / @wesm:
Run the source release script. It fails here: https://github.com/apache/arrow/blob/master/dev/release/02-source.sh#L114

[asfimport]

Uwe Korn / @xhochy:
Are these symlinks in the tar? At least cmake_modules were a symlink once that created these problems.

[asfimport]

Wes McKinney / @wesm:
I don't think so

[asfimport]

Krisztian Szucs / @kszucs:
Caused by the prepare script: https://github.com/apache/arrow/blob/master/dev/release/00-prepare.sh#L147-L171

On the release branch the rat exclude file is updated, and the 02-source.sh uses the updated exclude file, not the one which belongs to the release.

 

Here is the boundary: https://github.com/apache/arrow/blob/master/dev/release/00-prepare.sh#L141

So We need to run RAT with the exclude file within the archive.

[asfimport]

Uwe Korn / @xhochy:
These are things that are symlinked to in the archive. It probably depends on what comes first.

[asfimport]

Kouhei Sutou / @kou:
Issue resolved by pull request 4028
#4028