bfredl
bfredl
yeah, I think a "range" callback should always be able to defer the next call from the same provider to an explicit later position if it knows there is no...
it is getting late here but taking a good look at the details will be a prio for my upcoming work week:)
Sorry I forgor. Let's get this merged and we can iterate on the discussed things (deprecations, fine-grained skipping) in a follow up. Would you mind a final rebase? :)
I haven't checked the reproduce myself yet, but would you mind trying this patch? https://github.com/neovim/neovim/pull/34412
> It is currently impossible to replace lines/text without completely affecting extmarks. If I understand correctly, it behaves as if lines/text are deleted -> extmarks adjusted -> lines/text is placed....
> The UI multigrid protocol does not include enough information to render the UI like the TUI. This is expected behavior as the entire point of the multigrid _mode_ is...
ok, perhaps put in a better way: anything which we expect to happen in multigrid mode should not be a part of the compositor. These are two mutually exclusive modes...
Multigrid UI:s are absolutely supposed do their own position calculations and have done so just fine for 6 years. Not sure what the point is to write retcon fiction about...
merged. I'm going to some refactors as a follow-up PR.
`--headless` kinda implies that an UI is _not_ expected (cf `--embed` with `--embed --headless`) so if we support this specific pattern a different flag would be preferable.