|
|
|
@@ -0,0 +1,132 @@
|
|
|
|
|
# Rust 1.95 Follow-up Code Update Plan
|
|
|
|
|
|
|
|
|
|
Status: deferred. The Rust 1.95 compatibility bump should stay limited to MSRV/toolchain/docs unless CI exposes a required fix. This plan is for later code cleanup once the team is comfortable requiring Rust 1.95 everywhere.
|
|
|
|
|
|
|
|
|
|
## Baseline
|
|
|
|
|
|
|
|
|
|
- `rustc 1.95.0` and `cargo 1.95.0` build the current workspace without source changes.
|
|
|
|
|
- Verified locally before the compatibility bump:
|
|
|
|
|
- `cargo check --workspace --all-features`
|
|
|
|
|
- `cargo clippy --workspace --all-targets --all-features -- -D warnings`
|
|
|
|
|
- `cargo test --workspace --all-features`
|
|
|
|
|
- `cargo test --workspace --no-default-features`
|
|
|
|
|
- `cargo fmt --all -- --check`
|
|
|
|
|
- No custom JSON target specs, unstable `#[feature]` gates, or `$crate` self-imports were found.
|
|
|
|
|
- The codebase has many feature/platform `cfg` blocks and many small nested `if let` patterns. Those are the main places where Rust 1.95 may improve clarity.
|
|
|
|
|
|
|
|
|
|
## Guardrails
|
|
|
|
|
|
|
|
|
|
- Do not do a repo-wide syntax migration.
|
|
|
|
|
- Change one concern at a time and keep each PR behavior-preserving.
|
|
|
|
|
- Prefer the existing module boundaries. If a file is already large, use the cleanup as a chance to move toward concern-based modules rather than adding more code.
|
|
|
|
|
- Keep public config/session formats unchanged unless a separate feature requires it.
|
|
|
|
|
- After each slice, run at least `cargo fmt --all -- --check`, `cargo clippy --workspace --all-targets --all-features -- -D warnings`, and the narrow affected tests. Before merge, run `./tools/lint-and-test.sh`.
|
|
|
|
|
|
|
|
|
|
## Candidate Updates
|
|
|
|
|
|
|
|
|
|
### 1. Use `cfg_select!` for dense cfg dispatch
|
|
|
|
|
|
|
|
|
|
Rust 1.95 stabilizes `cfg_select!`. Use it only where it removes duplicated function bodies or closely related cfg branches.
|
|
|
|
|
|
|
|
|
|
Good candidates:
|
|
|
|
|
- `src/time_utils.rs`: Unix/non-Unix time conversion branches and target-pointer-width split.
|
|
|
|
|
- `src/file_uri.rs`: Unix/non-Unix path conversion.
|
|
|
|
|
- `src/session/lock.rs`: Unix/non-Unix lock behavior.
|
|
|
|
|
- `src/backend/wayland/state/core/accessors.rs`: tablet/non-tablet state accessors.
|
|
|
|
|
- `src/config/types/mod.rs` and `src/config/core.rs`: tablet-gated config shape.
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Prototype on one small module, likely `src/session/lock.rs` or `src/file_uri.rs`.
|
|
|
|
|
2. Compare readability against the current `#[cfg]` attributes.
|
|
|
|
|
3. Keep module-level cfg attributes when they are clearer than `cfg_select!`.
|
|
|
|
|
4. Run both feature matrices, especially `--all-features` and `--no-default-features`, because tablet and portal/tray gates are easy to regress.
|
|
|
|
|
|
|
|
|
|
### 2. Use `if let` guards in match arms where they flatten control flow
|
|
|
|
|
|
|
|
|
|
Rust 1.95 stabilizes `if let` guards. Prefer this only when it makes the match's decision table easier to read.
|
|
|
|
|
|
|
|
|
|
Good candidates:
|
|
|
|
|
- `build.rs`: nested `if let` logic in git-dir resolution.
|
|
|
|
|
- `src/tray_action.rs`: queued action parsing/cleanup paths.
|
|
|
|
|
- `src/daemon/tray/helpers.rs`: command spawning, runtime selection, and config/session result handling.
|
|
|
|
|
- `src/input/state/core/menus/layout.rs`: nested layout/index/entry checks.
|
|
|
|
|
- `src/input/state/core/menus/commands.rs`: repeated anchor/layout fallback decisions.
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Start with non-runtime-critical code such as `build.rs`.
|
|
|
|
|
2. Replace only a nested `match`/`if let` cluster when the guard expresses the condition directly.
|
|
|
|
|
3. Avoid clever guards with side effects or long expressions.
|
|
|
|
|
4. Keep tests focused on the touched behavior.
|
|
|
|
|
|
|
|
|
|
### 3. Consider `Vec::push_mut` and `Vec::insert_mut` for immediate mutation
|
|
|
|
|
|
|
|
|
|
Rust 1.95 adds stable mutable references from vector insertion APIs. Use them only where code currently inserts then immediately re-finds or mutates the inserted item.
|
|
|
|
|
|
|
|
|
|
Potential search areas:
|
|
|
|
|
- UI/model builders that construct vectors of toolbar/menu entries.
|
|
|
|
|
- Session/history code around inserted shapes or actions.
|
|
|
|
|
- `src/input/state/core/command_palette/search.rs`, where recent command insertion may benefit only if follow-up mutation appears later.
|
|
|
|
|
- `src/draw/frame/history/frame/apply.rs`, where shape insertion is sensitive; change only with regression tests.
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Search for `push(...);` followed by `last_mut`, `len() - 1`, or `get_mut`.
|
|
|
|
|
2. Search for `insert(...);` followed by `get_mut` or index-based mutation.
|
|
|
|
|
3. Prefer no change if the current code is already clearer or if the inserted value is not mutated.
|
|
|
|
|
4. Add or keep regression coverage for shape ordering/history behavior before touching frame insertion paths.
|
|
|
|
|
|
|
|
|
|
### 4. Revisit atomic state updates
|
|
|
|
|
|
|
|
|
|
Rust 1.95 adds atomic `update` and `try_update`. Current atomic usage appears simple (`fetch_add`, bool flags, small overrides), so this is likely a no-op.
|
|
|
|
|
|
|
|
|
|
Potential files:
|
|
|
|
|
- `src/tray_action.rs`
|
|
|
|
|
- `src/session_override.rs`
|
|
|
|
|
- daemon and backend state that use atomics for lifecycle flags
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Search for manual `compare_exchange` loops. If none exist, skip this slice.
|
|
|
|
|
2. Do not replace simple `store`, `load`, or `fetch_add`; that would reduce clarity.
|
|
|
|
|
3. If a compare/exchange loop is introduced later, prefer `try_update` when the failure path matters.
|
|
|
|
|
|
|
|
|
|
### 5. Audit new Clippy lints without churn
|
|
|
|
|
|
|
|
|
|
Clippy 1.95 adds default lints including `manual_checked_ops`, `manual_take`, and `disallowed_fields`. Current `-D warnings` passed, so there is no required work.
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Run Clippy after dependencies are refreshed in future work.
|
|
|
|
|
2. Treat new warnings as local cleanups, not a style migration.
|
|
|
|
|
3. Prefer `#[expect(..., reason = "...")]` for intentional exceptions.
|
|
|
|
|
|
|
|
|
|
### 6. Keep performance-only APIs measured
|
|
|
|
|
|
|
|
|
|
Rust 1.95 includes `core::hint::cold_path`. Do not add it broadly.
|
|
|
|
|
|
|
|
|
|
Potential areas only after profiling:
|
|
|
|
|
- Rare config parse/reporting error paths.
|
|
|
|
|
- Capture failure/reporting paths.
|
|
|
|
|
- Session corruption fallback paths.
|
|
|
|
|
|
|
|
|
|
Approach:
|
|
|
|
|
1. Add only after a measured hot path shows code layout matters.
|
|
|
|
|
2. Keep the hint near the rare branch, with a short reason.
|
|
|
|
|
3. Avoid using it as documentation for ordinary error handling.
|
|
|
|
|
|
|
|
|
|
## Suggested PR Sequence
|
|
|
|
|
|
|
|
|
|
1. `build.rs` and one small platform module: prove `if let` guards or `cfg_select!` improve clarity.
|
|
|
|
|
2. Tablet cfg cleanup: only if `cfg_select!` meaningfully reduces duplicate tablet/non-tablet accessors.
|
|
|
|
|
3. History/vector insertion cleanup: only if `push_mut` or `insert_mut` removes indexing or lookup without changing ordering.
|
|
|
|
|
4. Atomic audit: likely close as "no change needed" unless compare/exchange loops appear.
|
|
|
|
|
5. Final Clippy pass after dependency updates.
|
|
|
|
|
|
|
|
|
|
## Nix Follow-up
|
|
|
|
|
|
|
|
|
|
The flake currently relies on the pinned `nixpkgs` Rust toolchain. When Nix is available, run:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
nix flake update nixpkgs
|
|
|
|
|
nix develop --command sh -lc 'rustc --version && cargo --version'
|
|
|
|
|
nix build .#wayscriber .#wayscriber-configurator
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
If the updated `nixpkgs` still does not provide Rust 1.95+, add an explicit Rust 1.95 toolchain override in `flake.nix` rather than relying on an older default `pkgs.rustPlatform`.
|
devmobasa