Oliver Hamlet

I've come up with an approach that I think is conceptually simpler and results in fewer edges being added. The algorithm is: 1. Create a graph of the groups, with...

I've pushed an implementation of the approach described above, it's in the `reimpl-group-edges` branch. Unfortunately, since writing it I've realised that the lack of special handling for the default group...

I've pushed a few more commits that stop sharing the colour map and implement the two-stage default group handling. It turns out that sharing the colour map was a mostly...

I came up with a few more cases that the new code couldn't handle well, so I've stripped out the check to see if a plugin could add all its...

Unfortunately the new logic is much slower, here's some benchmarks: ## Group plugin sorting performance All tests with 1619 plugins and a 77 kB userlist that adds 9 groups and...

Some initial thoughts on this: Trying to load plugins as close as possible to their masters seems like a generally good idea, it's a more specific expression of the "load...

This isn't an issue with LOOT: if `AyleidCitadel.esp` has `Hearthfires.esm` instead of `HearthFires.esm`, that's an issue with that plugin (except not really, I think it's unreasonable to expect case correctness...

No, emulating filesystem case-insensitivity at an application level sounds like a recipe for all sorts of weird edge cases to me. It's bad enough that libloot already has to attempt...

For future reference, loot/loot#1095 is also about case-sensitive paths on Linux and offers a couple of solutions.

Aside from data read from game files not matching the case of filenames in the filesystem, libloot also hardcodes the following file extensions and uses them to construct paths that...