Use pinned flow_parser

[cometkim]

Experimented with using the upstream parser to catch up with updates and reduce maintenance costs.

Some errors from the Ounit tests are fixed by #7381.

There are some unintended changes in JS artifacts; I really don't get how the JS parser causes them. This needs investigation.

[cknitt]

Just tried this locally (after rebasing on latest master).
Did not see any unintended changes in JS artifacts so far, but did get compile errors in tests.

[cometkim]

After fixing the part @cknitt pointed out, this works properly.

I would support this change as long as it doesn't cause serious problems in performance or binary size. We don't want to maintain the whole JS parser logic anyway.

The only problem is that it requires OCaml compiler 5.2 or higher, but since it doesn't actually require a higher version of OCaml (yet), we can just fork the package and fix the range.

[cometkim]

Windows again... I have no idea why opam pin is failing on Windows :/

[cometkim]

Opened an issue: ocaml/opam#6472

[cknitt]

Until this is fixed in OPAM, can we vendor the new version?

[cometkim]

Not needed, I fixed all issues in our fork facebook/flow@main...rescript-lang:flow:rescript

[cometkim]

This increases the binary size ~1MiB per platform. But once #6183 is done, it would be a minor range.

It's a somewhat expected impact since the flow_parser includes a lot of additional syntax. Since we only use a small portion of its functionality, we might have a lighter dedicated parser later.

[cknitt]

Not needed, I fixed all issues in our fork facebook/[email protected]:flow:rescript

Wow, nice work! 🎉

Could warrant a CHANGELOG entry. 😉

[cknitt]

Ok, will merge now and update the CHANGELOG while preparing the next alpha release.