Oliver Hamlet

This would have to be done in libloot, as the `bad_format_string` exception thrown has `get_pos()` and `get_next()` member functions, which are only useful in a context that also has the...

I've done this in the differentiate-group-edge-types branches of the libloot and loot repositories: the changes break API compatibility for libloot so I'll hold them back for the future v0.19.0 release.

It's both: a group edge is identified as a user group edge if * it involves a group "load after" relationship that's defined in the userlist (this includes involving groups...

Can you try again with the latest `differentiate-group-edge-types` build and share your debug log and userlist?

It's not obvious what's going on, so here it is written out: * A is a master of AA, so AA loads after A * B is a master of...

Figured it out, I wasn't taking into account where the current plugin's group came from, only the groups of its predecessor plugins. The latest build should work as expected.

Done in fd2c5d5bd8ede58363858d85d89c064c5797ee57.

I've noticed this happening with my own application, and also (probably of more concern to Google engineers) to Google Play Music's playlist button. 

That's a valid point, it doesn't handle dropdowns that extend above the top of the list item, but I can't think of a good general solution, since the problem is...

Given the probable complexity of a general solution, I'd be OK managing item z-indexes as an implementer, but is there not a cleaner way to do so without wrapping everything...