List of mwes that panic

[ammkrn]

These are minimal working examples of inputs after hx that cause a panic, tested on ubuntu and macOS.

j, b

j, w

j, e

j, f, f

d, f, f

d, f, n

i, a, esc, b, w

i, a, esc, b, e

and someone else brought up d, d.

I tried making the calls to slice.char(_) safe in the varioushelix-core::movement functions, but that caused off by ones in coords_at_pos and something in the compositor, I'm not sure if it's because of bad indexing elsewhere, or if my "fixes" aren't obeying rules about the anchor/head points for Range. Others have also noted that it seems like you can move down one more line than you would be able to in IE kakoune, but that doesn't seem to explain the d, d or b cases.

Just poking around, there seems to be a lot of interaction with ropey via partial functions which may not be ideal in the long run. It's a bummer that ropey doesn't seem to offer total versions of the indexing functions.

Cool project by the way, I wish you the best of luck.

[archseer]

I think I addressed some of these in the latest push, it's mostly of by one errors or missing bounds checks -- I agree that having a safe indexing wrapper would be useful as well.

[ammkrn]

Thanks. It looks like the remaining ones are d, d, i, a, esc, b, w and i, a, esc, b, e. If the ropey maintainers are interested in having an api extension that doesn't panic, I'll see about adding one.

[archseer]

After f4560cb just dd left :)

[ammkrn]

Right on. The ropey maintainer(s) agreed to accept the api extension, I'm waiting for them to review the initial implementation. I'll @ mention you if/when it gets pulled into master.

[heliostatic]

Fantastic! Just came to report this (with backtrace) but delighted to see it will be fixed upstream.

[cessen]

Re: non-panicking version of Ropey's functions

@ammkrn and I have been discussing this in cessen/ropey#39. @archseer: if you could join the discussion over there, as the primary author of Helix, I'd really appreciate it!

[jbaa]

I believe the issue with dd was fixed with #90.

[ammkrn]

@jbaa Seems like it to me. Since the ropey api discussion is happening there, I'll go ahead and close this. Thanks!