Add support for targets without unwinding in `mir-opt`, and improve `--bless` for it
The main goal of this PR is to add support for targets without unwinding support in the `mir-opt` test suite, by adding the `EMIT_MIR_FOR_EACH_PANIC_STRATEGY` comment. Similarly to 32bit vs 64bit, when that comment is present, blessed output files will have the `.panic-unwind` or `.panic-abort` suffix, and the right one will be chosen depending on the target's panic strategy.
The `EMIT_MIR_FOR_EACH_PANIC_STRATEGY` comment replaced all the `ignore-wasm32` comments in the `mir-opt` test suite, as those comments were added due to `wasm32` being a target without unwinding support. The comment was also added on other tests that were only executed on x86 but were still panic strategy dependent.
The `mir-opt` suite was then blessed, which caused a ton of churn as most of the existing output files had to be renamed and (mostly) duplicated with the abort strategy.
---
After [asking on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/mir-opt.20tests.20and.20panic.3Dabort), the main concern about this change is it'd make blessing the `mir-opt` suite even harder, as you'd need to both bless it with an unwinding target and an aborting target. This exacerbated the current situation, where you'd need to bless it with a 32bit and a 64bit target already.
Because of that, this PR also makes significant enhancements to `--bless` for the `mir-opt` suite, where it will automatically bless the suite four times with different targets, while requiring minimal cross-compilation.
To handle the 32bit vs 64bit blessing, there is now an hardcoded list of target mapping between 32bit and 64bit. The goal of the list is to find a related target that will *probably* work without requiring additional cross-compilation toolchains on the system. If a mapping is found, bootstrap will bless the suite with both targets, otherwise just with the current target.
To handle the panic strategy blessing (abort vs unwind), I had to resort to what I call "synthetic targets". For each of the target we're blessing (so either the current one, or a 32bit and a 64bit depending on the previous paragraph), bootstrap will extract the JSON spec of the target and change it to include `"panic-strategy": "abort"`. It will then build the standard library with this synthetic target, and bless the `mir-opt` suite with it.
As a result of these changes, blessing the `mir-opt` suite will actually bless it two or four times with different targets, ensuring all possible variants are actually blessed.
---
This PR is best reviewed commit-by-commit.
r? `@jyn514`
cc `@saethlin` `@oli-obk`
Rollup of 5 pull requests
Successful merges:
- #112197 (Erase regions even if normalization fails in writeback (in new solver))
- #112495 (fix(resolve): update shadowed_glob more precision)
- #112520 (Fix the overflow issue for transmute_generic_consts)
- #112571 (rustdoc-search: search never type with `!`)
- #112581 ([rustdoc] Fix URL encoding of % sign)
r? `@ghost`
`@rustbot` modify labels: rollup
rustdoc-search: search never type with `!`
This feature extends rustdoc to support the syntax that most users will naturally attempt to use to search for diverging functions. Part of #60485
It's already possible to do this search with `primitive:never`, but that's not what the Rust language itself uses, so nobody will try it if they aren't told or helped along.
Because tiny CGUs make compilation less efficient *and* result in worse
generated code.
We don't do this when the number of CGUs is explicitly given, because
there are times when the requested number is very important, as
described in some comments within the commit. So the commit also
introduces a `CodegenUnits` type that distinguishes between default
values and user-specified values.
This change has a roughly neutral effect on walltimes across the
rustc-perf benchmarks; there are some speedups and some slowdowns. But
it has significant wins for most other metrics on numerous benchmarks,
including instruction counts, cycles, binary size, and max-rss. It also
reduces parallelism, which is good for reducing jobserver competition
when multiple rustc processes are running at the same time. It's smaller
benchmarks that benefit the most; larger benchmarks already have CGUs
that are all larger than the minimum size.
Here are some example before/after CGU sizes for opt builds.
- html5ever
- CGUs: 16, mean size: 1196.1, sizes: [3908, 2992, 1706, 1652, 1572,
1136, 1045, 948, 946, 938, 579, 471, 443, 327, 286, 189]
- CGUs: 4, mean size: 4396.0, sizes: [6706, 3908, 3490, 3480]
- libc
- CGUs: 12, mean size: 35.3, sizes: [163, 93, 58, 53, 37, 8, 2 (x6)]
- CGUs: 1, mean size: 424.0, sizes: [424]
- tt-muncher
- CGUs: 5, mean size: 1819.4, sizes: [8508, 350, 198, 34, 7]
- CGUs: 1, mean size: 9075.0, sizes: [9075]
Note that CGUs of size 100,000+ aren't unusual in larger programs.
This feature extends rustdoc to support the syntax that most users will
naturally attempt to use to search for diverging functions.
Part of #60485
It's already possible to do this search with `primitive:never`, but
that's not what the Rust language itself uses, so nobody will try it if
they aren't told or helped along.
Rollup of 4 pull requests
Successful merges:
- #112302 (Suggest using `ptr::null_mut` when user provided `ptr::null` to a function expecting `ptr::null_mut`)
- #112416 (Fix debug ICE for extern type with where clauses)
- #112527 (Add windows_sys type definitions for ARM32 manually)
- #112546 (new solver: extend assert to other aliases)
r? `@ghost`
`@rustbot` modify labels: rollup
implement stdout streaming in `render_tests::Renderer`
This way, we can show the test dot characters on the console immediately, without having to wait for the entire line to finish.
cc `@GuillaumeGomez`
It might happen that a synthetic target name does not match one of the
hardcoded ones in std's build script, causing std to fail to build. This
commit changes the std build script avoid including the restricted-std
feature unconditionally when a synthetic target is being built.
Some targets are added to these hashmaps at runtime, and are not present
during dry runs. To avoid errors, this commit changes all the related
functions to always return empty strings/paths during dry runs.
This has real differences in the effective debuginfo: in particular, it omits the module-level information and breaks perf.
Allow passing `line-tables-only` directly in config.toml instead.
rustdoc: Add `item_template` macro
Closes#112021
This change removes the use of `self.borrows()` in Askama templates, removes code duplication from `item_and_mut_cx()`, and improved readability by eliminating the prefix `item_template_` when calling from the template.
References:
- Discussion issue: https://github.com/rust-lang/rust/issues/112021
- `ItemTemplate` PR: https://github.com/rust-lang/rust/pull/111946
r? `@GuillaumeGomez`
Rollup of 7 pull requests
Successful merges:
- #112475 (Fix issue for module name when surround the struct literal with parentheses)
- #112477 (Give more helpful progress messages in `Assemble`)
- #112484 (Fix ntdll linkage issues on Windows UWP platforms)
- #112492 (Migrate GUI colors test to original CSS color format)
- #112493 (iat selection: normalize self ty & completely erase bound vars)
- #112497 (abs_sub: fix typo 0[-:][+.]0)
- #112498 (Update links to Rust Reference in diagnostic)
r? `@ghost`
`@rustbot` modify labels: rollup
Give more helpful progress messages in `Assemble`
Before (download-rustc):
```
# no output
```
After (download-rustc):
```
Creating a sysroot for stage2 compiler (use `rustup toolchain link 'name' build/host/stage2`)
```
Before (compiling from source):
```
Building compiler artifacts (stage0 -> stage1, x86_64-unknown-linux-gnu)
Assembling stage1 compiler
Build stage1 library artifacts (x86_64-unknown-linux-gnu -> i686-unknown-linux-gnu)
Building compiler artifacts (stage0:x86_64-unknown-linux-gnu -> stage1:i686-unknown-linux-gnu)
Assembling stage1 compiler (i686-unknown-linux-gnu)
```
After (compiling from source):
```
Building compiler artifacts (stage0 -> stage1, x86_64-unknown-linux-gnu)
Creating a sysroot for stage1 compiler (use `rustup toolchain link 'name' build/host/stage1`)
Build stage1 library artifacts (x86_64-unknown-linux-gnu)
Building compiler artifacts (stage0:x86_64-unknown-linux-gnu -> stage1:i686-unknown-linux-gnu)
Creating a sysroot for stage1 compiler (i686-unknown-linux-gnu) (use `rustup toolchain link 'name' build/i686-unknown-linux-gnu/stage1`)
```
cc https://rust-lang.zulipchat.com/#narrow/stream/122651-general/topic/Unable.20to.20compile.20rustc.20MSVC, https://discord.com/channels/273534239310479360/957720175619215380/1116867245499498506
Don't compile rustc to self-test compiletest
This was changed from stage 0 to 1 in https://github.com/rust-lang/rust/pull/108905, but I'm not sure why. Change it to `top_stage` instead to allow people to choose the stage.
This should save quite a bit of time in the `mingw-check` builder, which explicitly runs `x test --stage 0 compiletest`.
Note that this also fixes a latent bug that depended on running `x build compiler` before `x doc compiler`, as well as a couple cleanups related to symlinks (which made the latent bug easier to find).
cc `@pietroalbini`
rustdoc: re-elide cross-crate default trait-object lifetime bounds
Hide trait-object lifetime bounds (re-exported from an external crate) if they coincide with [their default](https://doc.rust-lang.org/reference/lifetime-elision.html#default-trait-object-lifetimes).
Partially addresses #44306. Follow-up to #103885. [Zulip discussion](https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/clean_middle_ty.3A.20I.20need.20to.20add.20a.20parameter/near/307143097).
Most notably, if `std` exported something from `core` containing a type like `Box<dyn Fn()>`, then it would now be rendered as `Box<dyn Fn(), Global>` instead of `Box<dyn Fn() + 'static, Global>` (hiding `+ 'static` as it is the default in this case). Showing `Global` here is a separate issue, #80379, which is on my agenda.
Note that I am not really fond of the fact that I had to add a parameter to such a widely used function (30+ call sites) to address such a niche bug.
CC `@GuillaumeGomez`
Requesting a review from a compiler contributor or team member as recommended on Zulip.
r? compiler
---
`@rustbot` label T-compiler T-rustdoc A-cross-crate-reexports
Make the build process more beginner friendly:
- Include information explaining that the stage2 toolchain should be
used (and not the stage1 toolchain) due to the `download-rustc`
setting.
- Display a message when the user runs `x setup tools` explaining that
they should use the stage2 toolchain.
bootstrap: Disallow `--exclude test::std`
Use the top-level Kind to determine whether Steps are excluded.
Previously, this would use the `Kind` passed to `--exclude` (and not do any filtering at all if no kind was passed).
That meant that `x test linkchecker --exclude std` would fail - you had to explicitly say `--exclude test::std`.
Change bootstrap to use the top-level Kind instead, which does the right thing automatically.
Note that this breaks things like `x test --exclude doc::std`, but I'm not sure why you'd ever want to do that.
There's a lot of churn here, but the 1-line change in the first commit is the actual behavior change, the rest is just cleanup.
Fixes https://github.com/rust-lang/rust/issues/103201. Note that this effectively reverts most of https://github.com/rust-lang/rust/pull/91965.
cc `@pietroalbini`
Uplift `clippy::cmp_nan` lint
This PR aims at uplifting the `clippy::cmp_nan` lint into rustc.
## `invalid_nan_comparisons`
~~(deny-by-default)~~ (warn-by-default)
The `invalid_nan_comparisons` lint checks comparison with `f32::NAN` or `f64::NAN` as one of the operand.
### Example
```rust,compile_fail
let a = 2.3f32;
if a == f32::NAN {}
```
### Explanation
NaN does not compare meaningfully to anything – not even itself – so those comparisons are always false.
-----
Mostly followed the instructions for uplifting a clippy lint described here: https://github.com/rust-lang/rust/pull/99696#pullrequestreview-1134072751
`@rustbot` label: +I-lang-nominated
r? compiler
Update cargo
12 commits in b0fa79679e717cd077b7fc0fa4166f47107f1ba9..49b6d9e179a91cf7645142541c9563443f64bf2b
2023-06-03 14:19:48 +0000 to 2023-06-09 17:21:19 +0000
- docs: doc comments for all registry kinds (rust-lang/cargo#12247)
- chore: Migrate print-ban from test to clippy (rust-lang/cargo#12246)
- fix: fetch nested git submodules (rust-lang/cargo#12244)
- refactor: registry source cleanup (rust-lang/cargo#12240)
- test: loose overly matches for git cli output (rust-lang/cargo#12241)
- fix: disable multiplexing on macOS for some versions of curl (rust-lang/cargo#12234)
- docs: doc comments for registry source and index (rust-lang/cargo#12239)
- doc: point to nightly cargo doc (rust-lang/cargo#12237)
- Upgrade to `gix` v0.45 for multi-round pack negotiations. (rust-lang/cargo#12236)
- refactor: git source cleanup (rust-lang/cargo#12197)
- Add message on reusing previous temporary path on failed cargo installs (rust-lang/cargo#12231)
- doc: the first line should be a simple sentence instead of a heading (rust-lang/cargo#12228)
r? `@ghost`
Change format of rustdoc-js tests by putting query and correction directly alongside the expected values
As I was working on fixing merge conflicts in #108537, I faced quite a big issue when trying to update the `rustdoc-js*` tests. To make it much simpler, this PR moves the `query` and `correction` directly alongside the expected data so now we know what is the query that is being run without needing to add comments or going back to the top of the file.
r? ```@notriddle```
