Various issues with Key-Value Observing implementation about libs-base HOT 4 OPEN

I think the determination of how to proceed with this should be made by @rfm. Although I am okay with the integration of MIT code, we would need to note that in our documentation. I think it's important for us to be as compatible as possible with the Apple implementation. Whether that means using the code you mention or fixing our own implementation.

from libs-base.

Hi Frederik,

I think the most important part that you may contribute are test cases. Just open up a merge request with plenty of not supported or failing test cases. That way we see which funtionality is missing or wrong. Of course addressing performance issues is a bit harder but these are a second step.

from libs-base.

Hi Frederik,

I think the most important part that you may contribute are test cases. Just open up a merge request with plenty of not supported or failing test cases. That way we see which funtionality is missing or wrong. Of course addressing performance issues is a bit harder but these are a second step.

100% agree.

from libs-base.

Thanks for the offer. It sounds like it coudl be very useful,
I agree that we really want ObjC code rather than C++, both because multiple languages in a codebase are a pain for maintainability, and because ObjC++ is also an issue for portability. Looking at the code would be good even if it's hard to translate.

As Fred says, what we most want are testcases, because I have absolutely no code using KVO (and I think that's similar for most other developers), so we need help understanding how it's intended to behave.

Greg has provided code to expose contexts (the remove method), which looks reasonable to me, but without testcases we don't know if it's really correct.

From what you say, that leaves support for dependent keys missing, but I don't really know what's needed for that (what the system is supposed to do), so help there would be very useful.

from libs-base.