Passing false looks fine to me. I'm not sure we should do this for the sake of adding type hints. Every deprecation adds to the migration process, we already have a lot of them. I'd prefer updating the docblock and mention that false is legal (and explain the purpose of it of course.)

Okay, I agree, this went too far :)
I will fix the typehint in this PR and will target 3.4, ok? (To avoid creating another one)

Passing false looks fine to me. I'm not sure we should do this for the sake of adding type hints. Every deprecation adds to the migration process, we already have a lot of them.

I don't think anyone will be affected by this. Poeple use the option and are unlikely to have changed the implementation. Otherwise I would also say it's not worth it. But currently there are two ways to do the same thing: passing false and passing a callable that returns false. And passing false directly violates the interface. So either we need to break BC by changing the interface or we need to do this here. So to me, this PR is the better solution.

Then let's finish this! @yceruto , I've just committed what you proposed. Did I get you right?

@yceruto , thanx, done. Also added tests to make sure the deprecation is triggered by default.

This looks clumsy. I'm for updating the type annotation and declare it as callable|false|null, with proper runtime validation that throws a TypeError when appropriate.

This looks clumsy. I'm for updating the type annotation and declare it as callable|false|null, with proper runtime validation that throws a TypeError when appropriate.

Yes, I agree it could be the simpler variant.