This rule originated in 7669f04fb0, to
override the default, limited line-width that makes sense for prose, but
doesn't make sense for code (which typically uses hard-wrapped lines):
7669f04fb0/src/librustdoc/html/static/main.css (L153)
This line width limiter isn't applied to the `<div class="content">` node
any more. It's been moved to a separate wrapper `<div>` that used to be
called `main-inner` (in 135281ed15) but is
now called `width-limiter` (since
d7528e2157).
rustdoc: remove no-op CSS `.block { padding: 0 }`
This rule was changed in 8fb1250aba from the original version that had a non-zero padding. It's not needed, because it's not overriding anything that would've given `.block` a padding.
Extend CSS check to CSS variables
This PR is a bit big because the first commit is a rewrite of the CSS parser to something a bit simpler which still allows to get easily access to CSS properties name.
The other two are about adding tests and adding the CSS variables check.
This check was missing because we are relying more and more on CSS variables rather than CSS selectors in themes.
r? `@notriddle`
rustdoc mobile: move notable traits to return type
These were originally on the left, but were moved to the return type in c90fb7185a. The CSS rule for mobile did not get updated at the time, so updating it now.
r? `@notriddle`
rustdoc: remove no-op CSS `h1-4 { color: --main-color }`
Headers already inherit the font color they need from their parents.
This rule dates back to earlier versions of the rustdoc theme, where headers and body had different text colors.
68c15be8b5/src/librustdoc/html/static/main.css (L72-L98)
Nowadays, since the two have exactly the same color (specified by the `--main-color` variable), this rule does nothing.
Replace `check_missing_items.py` with `jsondoclint`
[zulip discussion](https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/check_missing_items.2Epy.20Replacement.2E)
check_missing_items.py was a python script that checked rustdoc json output to make sure all the Id's referenced existed in the JSON index. This PR replaces that with a rust binary (`jsondoclint`) that does the same thing.
### Motivation
1. Easier to change when `rustdoc-json-types` changes, as `jsondoclint` uses the types directly.
2. Better Errors:
- Multiple Errors can be emited for a single crate
- Errors can say where in JSON they occored
```
2:2889:408 not in index or paths, but refered to at '.index."2:2888:104".inner.items[0]'
2:2890:410 not in index or paths, but refered to at '.index."2:2888:104".inner.items[1]'
```
3. Catches more bugs.
- Because matches are exaustive, all posible variants considered for enums
- All Id's checked
- Has already found #101770, #101199 and #100973
- Id type is also checked, so the Id's in a structs fields can only be field items.
4. Allows the possibility of running from `rustdoc::json`, which we should do in a crator run at some point.
cc ``@CraftSpider``
r? ``@GuillaumeGomez``
Constify impl Fn* &(mut) Fn*
Tracking Issue: [101803](https://github.com/rust-lang/rust/issues/101803)
Feature gate: `#![feature(const_fn_trait_ref_impls)]`
This feature allows using references to Fn* Items as Fn* Items themself in a const context.
Prefer explict closure sig types over expected ones
fixes#100800
Previously we only checked that given closure arguments are equal to expected closure arguments, but now we choose the given closure arguments for the signature that is used when type checking the closure body, and keep the other signature for the type of the closure as seen outside of it.
Fix `#[link kind="raw-dylib"]` to respect `#[link_name]`
Issue Details:
When using `#[link kind="raw-dylib"]` (#58713), the Rust compiler ignored any `#[link_name]` attributes when generating the import library and so the resulting binary would fail to link due to missing symbols.
Fix Details:
Use the name from `#[link_name]` if present when generating the `raw-dylib` import library, otherwise default back to the actual symbol name.
Rustdoc-Json: Fix Type docs.
Primitive doesn't include Array/Slice/Tuple, as they are their own variants.
ResolvedPath doesn't include Traits, as they appear in the DynTrait variant.
This rule was changed in 8fb1250aba from the
original version that had a non-zero padding. It's not needed, because
it's not overriding anything that would've given `.block` a padding.
Normalize struct field types in `confirm_builtin_unsize_candidate`
Fixes#75899
---
edited to move the normalization into `confirm_builtin_unsize_candidate` instead of the coercion code.
On later stages, the feature is already stable.
Result of running:
rg -l "feature.let_else" compiler/ src/librustdoc/ library/ | xargs sed -s -i "s#\\[feature.let_else#\\[cfg_attr\\(bootstrap, feature\\(let_else\\)#"
