ObserverOfTime
ObserverOfTime
> You actually broke the ABI in the past. Which is what sparked this (long overdue) change. What's the problem here?
> Between 0.22.2 and 0.22.3 the ABI changed. To be precise, the structure `TSTreeCursor` changed. You're _very_ late to [the party](https://github.com/tree-sitter/tree-sitter/issues/3296#issuecomment-2054409905). > As this was an incompatible interface change, the...
> We _can_ fail on breaking ABI changes _if_ the soname version isn't bumped in the same commit. What if there are multiple ABI breaking commits within a release?
No such thing. :)
Windows has the cmd scripts for some of them (which I'm not touching).
Might as well keep this open until then.
> I would prefer to keep the management commands local to `npm` (or ideally `tree-sitter`) and leave the `Makefile` as a simple build harness Well, `make test` is still valuable...
I think it makes sense to expose the supertype and/or subtypes in the API (first C and then Rust).
If this *must* be implemented with patching then it should be a separate package (e.g. `web-tree-sitter-vercel`). If instead it can be implemented with an optional module parameter and some conditions,...
> * create directories if not existing (i.e., `ts b -o foo/bar.so`) > * call `ts g` if `parser.c` doesn't exist Those are beyond my capabilities. > * output parity...