[new release] ppx_deriving (5.3.0)

[NathanReb]

Type-driven code generation for OCaml

• Project page: https://github.com/ocaml-ppx/ppx_deriving
• Documentation: https://ocaml-ppx.github.io/ppx_deriving/

CHANGES:

• Port standard plugins to ppxlib registration and attributes
  Port standard plugins to ppxlib registration and attributes ocaml-ppx/ppx_deriving#263
  (Simmo Saan)

• Optimize forwarding in eq and ord plugins
  Optimize forwarding in eq and ord plugins ocaml-ppx/ppx_deriving#252
  (Simmo Saan)

• Delegate quoter to ppxlib
  Port standard plugins to ppxlib registration and attributes ocaml-ppx/ppx_deriving#263
  (Simmo Saan)

• Introduce Ppx_deriving_runtime.Stdlib with OCaml >= 4.07.
  This module already exists in OCaml < 4.07 but was missing otherwise.
  Add OCaml 5.00 support to the tests and update documentation ocaml-ppx/ppx_deriving#258
  (Kate Deplaix)

[sim642]

As I feared, ocaml-ppx/ppx_deriving#263 leads to a lot of revdep breakage, despite the effort to maintain compatibility.

So far the big one appears to be ppx_let which declares the map extension and ppx_deriving.5.3.0 which declares the derive.map extension. I wonder how these things have co-existed so far and what could be done about it. [%map: ...] from ppx_deriving doesn't appear to be used on sherlocode (not sure what the use case for it would be), so maybe it could be removed.

I can look through all the failures soon.

[NathanReb]

It appears the main problem is the with_path argument to ppx_deriving.show thing, I'll look into fixing that.

[NathanReb]

The conflicting extension thing should be fixable with prefixes, I'm a bit surprised it's causing trouble at that level.

Is it because ppx_let is recording it without a prefix? If so I'll open a PR on ppx_let. That won't solve everything I guess as there could be conflicts from the user's perspective, i.e. some code might have to invoke the extension with a prefix but I guess that's okay.

Either way we'll have to cancel this release for now but we should be able to come up with a couple fixes that should make it usable. Probably worth releasing it as a 6.0 instead if it does break some stuff.

[sim642]

It appears the main problem is the with_path argument to ppx_deriving.show thing, I'll look into fixing that.

ppx_deriving has tests for it though, which makes this odd.

[sim642]

The issue with with_path is when it's used for deriving in signatures. In my PR, intf_generator for show doesn't take arguments, only impl_generator does.
I guess we should just allow the argument also in signatures, even though it has no effect there.

[NathanReb]

Yeah we probably should as it apparently used to define it there before. I'll work on a fix!

In the meantime I opened janestreet/ppx_let#14!

[NathanReb]

Let's close this for now, we'll re-release once we have fixed everything!