move from proc-macro-error2 to manyhow (fix RUSTSEC-2026-0173)

[FBranca]

See #332

proc-macro-error2 is no longer maintained, manyhow seems to be a good remplacement.

[divarvel]

Thanks for the update! Could you please:

• sign the eclipse eca? (Instructions in the contributing.md file)
• show the new output for compilation errors
• provide an overview of the new crate dependencies? (Just to make sure the licenses are ok)

[FBranca]

Here is sample of output error on a biscuit! proc macro :

datalog parsing error: ParseErrors { errors: [ParseError { input: "bad_input\n ", message: None }] }

The manyhow crate introduce only one new transitive dependency : smallvec using a MIT / Apache v2 license.

├── manyhow v0.11.4
│   ├── manyhow-macros v0.11.4 (proc-macro)
│   │   ├── proc-macro-utils v0.10.0
│   │   │   ├── proc-macro2 v1.0.106 (*)
│   │   │   ├── quote v1.0.45 (*)
│   │   │   └── smallvec v1.15.1
│   │   ├── proc-macro2 v1.0.106 (*)
│   │   └── quote v1.0.45 (*)
│   ├── proc-macro2 v1.0.106 (*)
│   ├── quote v1.0.45 (*)
│   └── syn v2.0.117 (*) 

I'm struggling to get the ECDA recognized by github

[divarvel]

Could you provide a diff for errors, especially with a semantic error like an unbound variable?

[avandecreme]

Could you provide a diff for errors, especially with a semantic error like an unbound variable?

Just a suggestion, but I found trybuild really nice for this: https://docs.rs/trybuild/latest/trybuild/ (you can look at https://github.com/avandecreme/enum_convert to see an example usage)

[FBranca]

I retried linking my github account to eclipse this morning and this time it worked like a charm on the first try.

Following the advice of @avandecreme I also added some tests using trybuild, but this comes with many dependencies :

└── trybuild v1.0.116
    ├── glob v0.3.3
    ├── serde v1.0.228
    │   └── serde_core v1.0.228
    ├── serde_derive v1.0.228 (proc-macro)
    │   ├── proc-macro2 v1.0.106
    │   │   └── unicode-ident v1.0.24
    │   ├── quote v1.0.45
    │   │   └── proc-macro2 v1.0.106 (*)
    │   └── syn v2.0.117
    │       ├── proc-macro2 v1.0.106 (*)
    │       ├── quote v1.0.45 (*)
    │       └── unicode-ident v1.0.24
    ├── serde_json v1.0.150
    │   ├── itoa v1.0.18
    │   ├── memchr v2.8.1
    │   ├── serde_core v1.0.228
    │   └── zmij v1.0.21
    ├── target-triple v1.0.0
    ├── termcolor v1.4.1
    └── toml v1.1.2+spec-1.1.0
        ├── serde_core v1.0.228
        ├── serde_spanned v1.1.1
        │   └── serde_core v1.0.228
        ├── toml_datetime v1.1.1+spec-1.1.0
        │   └── serde_core v1.0.228
        ├── toml_parser v1.1.2+spec-1.1.0
        │   └── winnow v1.0.3
        ├── toml_writer v1.1.1+spec-1.1.0
        └── winnow v1.0.3

and a diff between a cargo license before and after adding the trybuild crate is even more impressive
cargo-license-diff-after-trybuild.txt

Of course there're all dev-dependencies.

For now I pushed this on another branch that you can see here : https://github.com/FBranca/biscuit-rust/tree/proc-macro-error2-to-manyhow-with-trybuild

This shows no change in error messages on tested error cases.

For an unbound variable :

error: datalog parsing error: ParseErrors { errors: [ParseError { input: "\n        can_view($file, $unused) <- right($file, \"read\")", message: Some("the rule contains variables that are not bound by predicates in the rule's body: $unused") }] }
  --> tests/error_message/authorizer_macro.rs:26:13
   |
26 |       let _ = authorizer!(r#"
   |  _____________^
27 | |         can_view($file, $unused) <- right($file, "read");
28 | |         allow if file($f), operation($op), right($f, $op);
29 | |     "#);
   | |_______^
   |
   = note: this error originates in the macro `authorizer` (in Nightly builds, run with -Z macro-backtrace for more info)

Tell me if the added dependencies are OK. If so I'll include the commit in this merge request.