Improvements to building and CI for mingw/msys
I was getting error messages when trying to follow the build instructions the mingw build for Rust, and managed to track the issue down to an incomparability of Rust's bootstrap program with MSYS2's version of git. Essentially, the problem is that MSYS2's git works in emulated unix-y paths, but bootstrap expects a Windows path. I found a workaround for this by using relative paths instead of absolute paths.
Along with that fix, this PR also updates the build instructions for MinGW to be compatible with modern versions of MSYS2, and some changes to CI to make sure that MSYS2's version of git is tested. In particular, I'm suggesting using the [MSYS2 github action](https://github.com/marketplace/actions/setup-msys2) specially made for this purpose, which is much less hacky than the old approach and gives us more control of what packages are installed. I also cleaned up as many alternate versions of key tools as I could find from PATH, to avoid accidental usage, and cleaned up some abuses of the `CUSTOM_MINGW` environment variable.
This fixes https://github.com/rust-lang/rust/issues/105696 and fixes https://github.com/rust-lang/rust/issues/117567
Make most bootstrap step types !Copy
This makes all bootstrap types except for `Compiler` and `TargetSelection` `!Copy`. This makes it easier to modify them by adding !Copy types in the future, something that `@saethlin` has complained about before, and comes at no cost of code clarity, the impls were completely unused.
Making `Compiler` and `TargetSelection` `!Copy` (which would allow getting rid of interning) is highly nontrivial as they are used and copied **all over the place**. This should hopefully get most of the benefits.
Add newtypes for bool fields/params/return types
Fixed all the cases of this found with some simple searches for `*/ bool` and `bool /*`; probably many more
Windows: Use ProcessPrng for random keys
Windows 10 introduced [`ProcessPrng`](https://learn.microsoft.com/en-us/windows/win32/seccng/processprng) for random number generation. This allows us to replace the overly complicated (and prone to failure) `BCryptGenRandom` with a documented function.
For the tier 3 Windows 7 target, we simply use the older `RtlGenRandom`, which is undocumented. It should be fine even on modern systems (for comparability reasons) as it's just a wrapper for `ProcessPrng`. However, it does require loading an extra intermediary DLL which we can avoid when we know we have Windows 10+.
speed up `x install` by skipping archiving and compression
Performing archiving and compression on `x install` is nothing more than a waste of time and resources. Additionally, for systems like gentoo(which uses `x install`) this should be highly beneficial.
[benchmark report](https://github.com/rust-lang/rust/pull/118724#issuecomment-1848964908)
Resolves#109308
r? Mark-Simulacrum (I think you want to review this, feel free to change it if otherwise.)
This makes all bootstrap types except for `Compiler` and
`TargetSelection` `!Copy`. This makes it easier to modify them by adding
!Copy types in the future and comes at no cost of code clarity, the
impls were completely unused.
compiletest: call cargo-miri directly rather than via 'cargo run'
Fixes https://github.com/rust-lang/miri/issues/3297. Thanks to `@bjorn3` for figuring out the cause of this.
r? `@oli-obk`
Add "cargo miri clean" command
My first reaction when my miri cache was messed up was to attempt run this, which obviously failed. This helps paper over platform differences and such.
Windows miri-script execution egronomics
This allows for Windows users to use miri-script without pain. As working on miri earlier I was doing
`.\miri-script\target\debug\miri-script.exe { install | build | ... }` which wasn't fun.
Ignore less tests in debug builds
Since https://github.com/rust-lang/rust/pull/120594 and https://github.com/rust-lang/rust/pull/120863, nearly all UB-detecting debug assertions get compiled out of code that is monomorphized by a crate built with debug assertions disabled.
Which means that if we default all our codegen tests to `-Cdebug-assertions=no`, most of them work just fine against a sysroot built with debug assertions.
I also tried to explain a bit better why some tests need to be skipped, for those that still need to be skipped.
Rustdoc: include crate name in links for local primitives
Fixes#121106.
This change makes links to primitives easier to use when the path of the page where they will be embedded is not known beforehand such as when we generate impls dynamically from the `register_type_impls` method in `main.js`, which is exactly what is happening in #121106.
An example to show the effect of this change: earlier, if the current page in `cx.current` inside `primitive_link_fragment()` was `std::simd::prelude::Simd` the generated path would be `../../primitive.<prim>.html`. Now it would be `../../../std/primitive.<prim>.html` instead.
A side effect of the change is that local primitive links _everywhere_ will now contain the crate name, even outside of the dynamic situation mentioned above. I'm not sure if there are any major downsides of that other than making the links a bit longer. Ideally I wanted to restrict this behaviour change to only the dynamic cases. We could have achieved that by passing an additional bool arg to `primitive_link_fragment()`, but it felt awkward to do so. Any alternative suggestions are welcome.
It makes the link easier to use in cases in which
the path of the page where it will be embedded is not
known beforehand such as when we generate impls
dynamically from `register_type_impls` method in
`main.js`
Earlier for local primitives we would generate a path
that was relative to the current page depth passed in `cx.current`
. e.g if the current page was `std::simd::prelude::Simd` the
generated path would be `../../primitive.<prim>.html` After this
change the path will first take you to the the wesite root and add
the crate name. e.g. for `std::simd::prelude::Simd` the path now
will be `../../../std/primitive.<prim>.html`
Explicitly call `emit_stashed_diagnostics`.
Commit 72b172b in #121206 changed things so that
`emit_stashed_diagnostics` is only called from `run_compiler`. But rustfmt doesn't use `run_compiler`, so it needs to call `emit_stashed_diagnostics` itself to avoid an abort in `DiagCtxtInner::drop` when stashed diagnostics occur.
Fixes#121450.
r? `@oli-obk`
remove `llvm.assertions=true` in compiler profile
Having this set to true disrupts compiler development workflows for people who use `llvm.download-ci-llvm = true` because we don't provide ci-llvm on the `rustc-alt-builds` server. Therefore, it is kept off by default.
cc `@Nilstrieb` `@compiler-errors`
For more context, see https://rust-lang.zulipchat.com/#narrow/stream/242791-t-infra/topic/CI.20LLVM.20for.20aarch64
Move as many tests from tests/ui/numbers-arithmetic to tests/ui/lint as possible
Fixes#119851 , and also consolidates as many individual tests as possible from numbers-arithmetic. I might have moved the tests in too aggressively, so let me know
Commit 72b172b in #121206 changed things so that
`emit_stashed_diagnostics` is only called from `run_compiler`. But
rustfmt doesn't use `run_compiler`, so it needs to call
`emit_stashed_diagnostics` itself to avoid an abort in
`DiagCtxtInner::drop` when stashed diagnostics occur.
Fixes#121450.
bootstrap: don't resolve symlinks for initial_cargo
I have put the following in my `config.toml`:
```toml
# Includes one of the default files in src/bootstrap/defaults
profile = "compiler"
change-id = 121203
[build]
cargo = "/usr/bin/cargo"
rustc = "/usr/bin/rustc"
rustfmt = "/usr/bin/rustfmt"
```
I have rustup installed from Arch's repos, which has all of the above paths be symlinks to `/usr/bin/rustup`. This works just fine with the `argv[0]` trick that rustup uses.
However, `bootstrap` resolves symlinks to check whether `cargo` exists and then uses the resolved path, so it ends up calling `rustup` directly expecting it to behave like `cargo`. Which it doesn't.
This PR removes the canonicalization step, in turn fixing the issue, but sacrificing a pretty error message. However, this exact thing is checked by `x.py` in advance, so I hope it is not a big deal?
Having this set to true disrupts compiler development workflows for people who use `llvm.download-ci-llvm = true`
because we don't provide ci-llvm on the `rustc-alt-builds` server. Therefore, it is kept off by default.
Signed-off-by: onur-ozkan <work@onurozkan.dev>
