On 02/12/15 13:14, Diego M. Rodriguez wrote:
While the package does work without those components (setup.py falls back to using the Python implementation of the algorithms, and tests.py includes tests cases that do not depend on the .csv files), I'm wondering what should be done about the issue: is there a way to append them gracefully to the Debian package, or would it be preferred to leave them out entirely in order to reflect the official upstream tarball status? Just in case, I have contacted upstream and both submodules repositories would have the same license and copyright as the main repository.
You can have a look at the ArrayFire package that I personally maintain, which also relies on submodules.
You basically have 2 options:- Use the upstream git repository and gbp with the --git-submodules option (which is the solution I use with ArrayFire).
- Provide a custom get-orig-source target which will be responsible for checking out the upstream repository, including submodules, and repack the content to a clean tarball. Then, you can use the standard approach with gbp import-orig.
Regardless of the method used, please consider documenting the process of generating and committing the upstream tarball inside the packaging repository, typically in the README.Debian file.
Let me know if you have further questions, Ghis