rustdoc: Use own logic to print #[repr(..)] attributes in JSON output.

[obi1kenobi]

Switch away from Debug-like representation of #[repr(..)] attributes, and start using rustdoc's own logic for pretty-printing #[repr(..)] in rustdoc JSON.

Part of addressing #137645 but not a complete solution for it.

r? @aDotInTheVoid

[rustbot]

rustdoc-json-types is a public (although nightly-only) API. If possible, consider changing src/librustdoc/json/conversions.rs; otherwise, make sure you bump the FORMAT_VERSION constant.

cc @CraftSpider, @aDotInTheVoid, @Enselic, @obi1kenobi

Some changes occurred in tests/rustdoc-json

cc @aDotInTheVoid

[aDotInTheVoid]

This looks really good! Thanks for the very thorough tests, it's appreciated.

Could you update the docs on Item::attribute? After that and rebasing to bump the format version again, this should be good to merge.

[aDotInTheVoid]

NIT: It'd be nice to cite where these rules come from: https://doc.rust-lang.org/nomicon/other-reprs.html#reprtransparent

[aDotInTheVoid]

Another PR currently in the queue is making this change: #138109 (comment)

Once that lands, you should rebase ontop of that, and then bump again.

[obi1kenobi]

Happy to do it, though I'm currently traveling for another week so it won't be soon. I'd be thrilled to have this merged ASAP though, so if you'd like to make the necessary changes and then merge, go for it. I don't care whose name is on the commit :)

…

On Tue, Mar 11, 2025, 4:55 PM Alona Enraght-Moony ***@***.***> wrote: ***@***.**** commented on this pull request. This looks really good! Thanks for the very thorough tests, it's appreciated. Could you update the docs on Item::attribute? After that and rebasing to bump the format version again, this should be good to merge. ------------------------------ In tests/rustdoc-json/attrs/repr_transparent.rs <#138018 (comment)>: > @@ -0,0 +1,35 @@ +#![no_std] + +// Rustdoc JSON *only* includes `#[repr(transparent)]` +// if the transparency is public API: +// - if a non-1-ZST field exists, it has to be public +// - otherwise, all fields are 1-ZST and at least one of them is public NIT: It'd be nice to cite where these rules come from: https://doc.rust-lang.org/nomicon/other-reprs.html#reprtransparent ------------------------------ In src/rustdoc-json-types/lib.rs <#138018 (comment)>: > @@ -30,7 +30,7 @@ pub type FxHashMap<K, V> = HashMap<K, V>; // re-export for use in src/librustdoc /// This integer is incremented with every breaking change to the API, /// and is returned along with the JSON blob as [`Crate::format_version`]. /// Consuming code should assert that this value matches the format version(s) that it supports. -pub const FORMAT_VERSION: u32 = 40; +pub const FORMAT_VERSION: u32 = 41; Another PR currently in the queue is making this change: #138109 (comment) <#138109 (comment)> Once that lands, you should rebase ontop of that, and then bump again. — Reply to this email directly, view it on GitHub <#138018 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAR5MSQMCSMYZWTJIBFBHC32T5SVVAVCNFSM6AAAAABYKTDHVCVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDMNZWGE4TONJSG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[bors]

☔ The latest upstream changes (presumably #138548) made this pull request unmergeable. Please resolve the merge conflicts.

[aDotInTheVoid]

I've done the test changes in only in #138569. I'll look at the content changes in a bit (if you don't get around to it first).

Good thing about doing tests before impl change: It tells you more about what's being changed. In this case, we go from always showing #[repr(transparent)], to only showing #[repr(transparent)] when it's a part of the public API. Still not sure that's a the right choice.