1.8.0 requires code changes around functions taking git_commit about libgit2

[ethomson]

Right, I wasn't intending or expecting this change to be quite so troublesome. Indeed, I can imagine that C++ and const correctness would require some changes. I didn't realize that C++ enforced const correctness on extern C but that's little surprise as I'm not a C++ programmer and this is not a C++ library.

In any case, git_commits aren't const. They have an internal state that points into an internal cache, those pointers can change, and they're refcounted which will change. Making this const git_commit ** was a mistake in the first place, which you can see by the unnecessary casts that existed in the examples, etc.

However:

1. There are a surprising number of places that take a const git_commit *. Which surprises me.
2. I wasn't intending for this to be a breaking change. (Even if you are using C++.)

I need to have a deeper look at this, but I wanted to provide the rationale behind the change.

[carrotIndustries]

For reference, I've added a workaround in the meantime: horizon-eda/horizon@af6c626