Refactor the `MirPass for QualifyAndPromoteConstants`
This is an accumulation of drive-by commits while working on `Vec::new` as a stable `const fn`.
The PR is probably easiest read commit-by-commit.
r? @oli-obk
cc @eddyb @ecstatic-morse -- your two PRs https://github.com/rust-lang/rust/pull/63812 and https://github.com/rust-lang/rust/pull/63860 respectively will conflict with this a tiny bit but it should be trivial to reintegrate your changes atop of this.
Rollup of 4 pull requests
Successful merges:
- #62205 (Add Iterator comparison methods that take a comparison function)
- #64152 (Use backtrace formatting from the backtrace crate)
- #64265 (resolve: Mark more erroneous imports as used)
- #64267 (rustdoc: fix diagnostic with mixed code block styles)
Failed merges:
r? @ghost
rustdoc: fix diagnostic with mixed code block styles
This fixes a relatively obscure issue where the diagnostic (emitted [here](ef54f57c5b/src/librustdoc/passes/check_code_block_syntax.rs (L69))) would get confused since the "is_fenced" flag wasn't reset properly.
Add Iterator comparison methods that take a comparison function
This PR adds `Iterator::{cmp_by, partial_cmp_by, eq_by, ne_by, lt_by, le_by, gt_by, ge_by}`. We already have `Iterator::{cmp, partial_cmp, ...}` which are less general (but not any simpler) than the ones I'm proposing here.
I'm submitting this PR now because #61505 has been merged, so this change should not have a noticeable effect on the `Iterator` docs page size.
The diff is quite messy, here's what I changed:
- The logic of `cmp` / `partial_cmp` / `eq` is moved to `cmp_by` / `partial_cmp_by` / `eq_by` respectively, changing `x.cmp(&y)` to `cmp(&x, &y)` in the `cmp` method where `cmp` is the given comparison function (and similar for `partial_cmp_by` and `eq_by`).
- `ne_by` / `lt_by` / `le_by` / `gt_by` / `ge_by` are each implemented in terms of one of the three methods above.
- The existing comparison methods are each forwarded to their `_by` counterpart, passing one of `Ord::cmp` / `PartialOrd::partial_cmp` / `PartialEq::eq` as the comparison function.
The corresponding `_by_key` methods aren't included because they're not as fundamental as the `_by` methods and can easily be implemented in terms of them. Is that reasonable, or would adding the `_by_key` methods be desirable for the sake of completeness?
I didn't add any tests – I couldn't think of any that weren't already covered by our existing tests. Let me know if there's a particular test that would be useful to add.
Fix regex replacement in theme detection
Fixes#64061.
This is sadly a lot of bad luck: after making the changes and re-build the docs, I just forgot to force reload the page. Hence having the old (working) version with two replacements instead of the failing regex. Sorry again about that...
cc @fenhl
r? @Mark-Simulacrum
Rustdoc: formatting to buffers
This should be reviewed commit-by-commit.
I've not attempted to fully flesh out what the end state of this PR could look like yet as I wanted to get it up for some early feedback (I already think this has some wins, too, so we could land it as-is).
The primary idea with `Buffer` is that it internally tracks whether we're printing to HTML or text, and the goal is that eventually instead of branch on `fmt.alternate()` anywhere, we'd call a helper like `buf.nbsp()` which would either return ` ` or ` ` depending on the target we're printing to. Obviously, that's not included in this PR, in part because it was already getting quite big.
This means that callers can pass in a closure like
`|buf| some_function(..., &mut buf)` and pass in arbitrary arguments to
that function without complicating the trait definition. We also keep
the impl for str and String, since it's useful to be able to just pass
in "" or format!("{}"...) results in some cases.
This changes Print's definition to take self, instead of &self, because
otherwise FnOnce cannot be called directly. We could instead take FnMut
or even Fn, but that seems like it'd merely complicate matters -- most
of the time, the FnOnce does not constrain us at all anyway. If it does,
a custom Print impl for &'_ SomeStruct is not all that painful.
Rollup of 5 pull requests
Successful merges:
- #64052 (Rename test locals to work around LLDB bug)
- #64066 (Support "soft" feature-gating using a lint)
- #64177 (resolve: Do not afraid to set current module to enums and traits)
- #64229 (Reduce span to function name in unreachable calls)
- #64255 (Add methods for converting `bool` to `Option<T>`)
Failed merges:
r? @ghost
Reduce span to function name in unreachable calls
As title suggests, this might close#64103. Refer to the updated tests for expected output.
There is potential to further improve usability. In particular, is it favourable that the exact diverging expression/statement be pointed out (not only in this case, but for all unreachable code)? Certainly that would deserve another issue, but I'm interested in the opinions.
resolve: Do not afraid to set current module to enums and traits
After cfbb60bf6d it's ok.
This is likely required for https://github.com/rust-lang/rust/pull/63468 to work correctly, because that PR starts resolving attributes on enum variants.
r? @matthewjasper @c410-f3r
Rename test locals to work around LLDB bug
LLDB's expression parser can't unambiguously resolve local variables in
some cases, as described in #47938. Work around this by using names that
don't shadow direct submodules of `core`.
Closes#64050.
