On Mon, Mar 31, 2025, 21:01 James Addison <j...@jp-hosting.net> wrote:
> ... > I think a path forward could be to patch the matplotlib package in > Debian so that the presence of a dist-packages/gi directory -- perhaps > only created implicitly for the overrides -- is insufficient for the > Python import of gi to be successful. > > If we can write such a patch in a way that would be suitable for other > distributions (I do not know how they ship override files), then > perhaps it could also be sent upstream for their consideration. > Thinking about this more: your code to distinguish between gi-imported-but-empty and gi-imported successfully could be reusable to do exactly this. We could patch that logic into the backend_gtk* import blocks, and raise an ImportError if we find that the module has no file. >
