Second this issue.

In most (if not all) of these cases, there's some reason not to use the latest stable release. However, it's a fair point, for most of them we should probably use specific git commits or create new releases as appropriate.

Creating releases for some of them might make no sense at all IMO because for example Kivent modules can have separate fixes and if there was a bug in one of the, I doubt there would be a separate release for the whole Kivent (I might be wrong though).

Then there are packages which have never even had a release (smaller, from Kivy org mostly) so we would need to create some release process for them (hopefully even with tests and CI where applicable). The change to specific commits would make it faster for p4a builds to become more reproducible this way though.

Is there any reason p4a can't use something like entry_points built into setuptools to allow version specific recipes to be installed independently of p4a?

Using this, each recipe could be in it's own repo and you could install a specific version with pip.

pip install p4a-<recipe>==<version>

Adding this should just be an update to https://github.com/kivy/python-for-android/blob/master/pythonforandroid/recipe.py#L620 , to check for packages installed with a given entry point (for example a fancy p4a_recipe) entry point.

I can submit a PR if this is an acceptable solution.

P.S. An entry_point for custom bootstraps (ex p4a_bootstrap) could be used in the same way.

Closing as not a priority for these specific recipes, but something that has been addressed in several cases where possible by sticking to a specific commit.