docs(hid): stop linking a cfg-gated type so rustdoc passes on Windows

[yuzi-co]

crates/openlogi-hid/src/transport.rs's module doc links AsyncHidChannel::supports_short_long_hidpp, but the type is declared #[cfg(not(target_os = "windows"))] (transport.rs:435). On Windows the link has no target, lib.rs:11 denies rustdoc::broken_intra_doc_links, and the whole crate's docs fail to build:

error: unresolved link to `AsyncHidChannel::supports_short_long_hidpp`
 --> crates\openlogi-hid\src\transport.rs:7
  |
7 | //! straight from [`AsyncHidChannel::supports_short_long_hidpp`]: USB / receiver
  |                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ no item named `AsyncHidChannel` in scope
  |
note: the lint level is defined here
 --> crates\openlogi-hid\src\lib.rs:11
  |
11 | #![deny(rustdoc::broken_intra_doc_links)]

error: could not document `openlogi-hid`

Why CI is green anyway

ci.yml's rustdoc (hid crates) job runs on ubuntu-latest only, where the type exists and the link resolves. The latest master run (822c6e41) reports that job as success, and it is not wrong — the break is invisible from Linux.

Why it still matters

The cost lands on contributors rather than CI. AGENTS.md makes the four-command local gate a hard stop:

Never git push until the final tree has passed the full local gate. … Run all four on the commit you are about to push
… Exit non-zero on any of those → fix, re-run the whole set, then push. Do not push "to see if CI likes it."

On a Windows checkout of unmodified master, step four exits 101. Following the documented process to the letter is impossible, so the honest options are to skip the mandated step or to stop. That is a bad first experience for exactly the platform the README calls the young port.

It is also the failure mode AGENTS.md itself warns about at length under "Platform / cfg-gated code (macOS-green is a trap)" — same class, just with Linux as the green platform this time.

Fix

Name the type in plain backticks rather than linking it. That is already how node_ledger.rs refers to AsyncHidChannel (lines 5 and 37) — this module doc was the lone outlier — and a short note inline records why, so the link does not get "helpfully" restored later.

No behaviour change; docs text only.

Verification

Full gate from AGENTS.md, run on Windows 11 x86_64 (rustc 1.97.1) on the commit being pushed:

cargo fmt --all -- --check                                    exit=0
cargo clippy --workspace --all-targets -- -D warnings         exit=0
cargo test --workspace                                        exit=0
RUSTDOCFLAGS="-D warnings" cargo doc -p openlogi-hid \
  -p openlogi-hidpp -p openlogi-hidpp-derive --no-deps \
  --document-private-items                                    exit=0

Before the change, that fourth command exits 101 on the same machine against clean master; I confirmed that by stashing the diff and re-running.

[greptile-apps]

Greptile Summary

This PR fixes Windows rustdoc generation by replacing an intra-doc link to a non-Windows-only type with plain code formatting.

• Documents why AsyncHidChannel must remain unlinked on Windows.
• Aligns the reference with the existing convention in node_ledger.

Confidence Score: 5/5

The PR appears safe to merge because it makes a documentation-only correction that avoids the cfg-gated Windows rustdoc failure.

The changed reference no longer requires rustdoc to resolve a type excluded on Windows, while the surrounding documentation and runtime code remain unchanged.

Important Files Changed

Filename	Overview
crates/openlogi-hid/src/transport.rs	Replaces a platform-invalid rustdoc link with plain code formatting and records the cfg-related rationale; no issues found.

_{Reviews (1): Last reviewed commit: "docs(hid): stop linking a cfg-gated type..." | Re-trigger Greptile}

[yuzi-co]

Closing — master fixed this independently, and better.

This PR escaped the link to plain backticks so deny(broken_intra_doc_links) would stop failing the crate's docs on Windows. Master reached the same outcome without giving up the link.

Two things landed since this branch was cut:

• 9126146 moved the file: crates/openlogi-hid/src/transport.rs → crates/openlogi-hid/src/channel/transport.rs.
• The type was renamed AsyncHidChannel → RawHidChannel, and the doc comment now reads [hidpp::channel::RawHidChannel::supports_short_long_hidpp] — a fully-qualified path into hidpp, where the type is not cfg-gated. It resolves on every platform, so the link stays clickable instead of becoming inert backticks.

Verified on master at 55a6ff2, on Windows:

RUSTDOCFLAGS="-D warnings" cargo doc -p openlogi-hid --no-deps --document-private-items   exit=0

Rebasing this would have meant reapplying a workaround to a problem that no longer exists, against a file that has moved.

The same bug is still live in another crate

crates/openlogi-permissions/src/lib.rs:17 links PermissionStatus::Unknown, and that enum is #[cfg(any(target_os = "macos", target_os = "linux"))] — so on Windows the type does not exist and the link cannot resolve. It fails the whole crate's docs there:

error: unresolved link to `PermissionStatus::Unknown`
  --> crates\openlogi-permissions\src\lib.rs:17:27

That reproduces on unmodified master, and CI never catches it because the rustdoc job runs on Linux. Same failure mode this PR was written for, one crate over. Noting it here so it is on record; not fixing it in this PR.