Refactor the way bootstrap invokes `cargo miri`
Instead of basically doing `cargo run --manifest-path=<cargo-miri's manifest> -- miri`, let's invoke the `cargo-miri` binary directly. That means less indirections, and also makes it easier to e.g. run the libcore test suite in Miri. (But there are still other issues with that.)
Also also adjusted Miri's stage numbering so that it is consistent with rustc/rustdoc.
This also makes `./x.py test miri` honor `--no-doc`.
And this fixes https://github.com/rust-lang/rust/issues/123177 by moving where we handle parallel_compiler.
Make source tarball generation more reproducible
This PR performs several changes to source tarball generation (`x dist rustc-src`) in order to make it more reproducible (in light of the recent "xz backdoor"...). I want to follow up on it with making a separate CI workflow for generating the tarball.
After this PR, running this locally produces identical checksums:
```bash
$ ./x dist rustc-src
$ sha256sum build/dist/rustc-1.79.0-src.tar.gz
$ ./x dist rustc-src
$ sha256sum build/dist/rustc-1.79.0-src.tar.gz
```
r? `@Mark-Simulacrum`
copy any file from stage0/lib to stage0-sysroot/lib
With the LLVM 18 upgrade, the name of the LLVM library has been changed to something like `libLLVM.so.18.1-rust-1.78.0-beta`, which `is_dylib` function cannot determine as it only looks whether files are ending with ".so" or not.
This change resolves this problem by no longer doing that ".so" check, as we need all files from the stage0/lib as they are all dependency of rustc anyway.
Fixes#122913
With the LLVM 18 upgrade, the name of the LLVM library has been changed to something like
`libLLVM.so.18.1-rust-1.78.0-beta`, which `is_dylib` function cannot determine as it only
looks whether files are ending with ".so" or not.
This change resolves this problem by no longer doing that ".so" check, as we need all files
from the stage0/lib as they are all dependency of rustc anyway.
Signed-off-by: onur-ozkan <work@onurozkan.dev>
fix(bootstrap/dist): use versioned dirs when vendoring
Currently, if you attempt to run ui tests in a vendored build, you will
see this failure
```
---- [ui] tests/ui/issues/issue-21763.rs stdout ----
diff of stderr:
8 = note: required because it appears within the type `(Rc<()>, Rc<()>)`
9 = note: required for `hashbrown::raw::RawTable<(Rc<()>, Rc<()>)>` to implement `Send`
10 note: required because it appears within the type `hashbrown::map::HashMap<Rc<()>, Rc<()>, RandomState>`
- --> $HASHBROWN_SRC_LOCATION
+ --> /rust/deps/hashbrown/src/map.rs:190:12
12 note: required because it appears within the type `HashMap<Rc<()>, Rc<()>>`
13 --> $SRC_DIR/std/src/collections/hash/map.rs:LL:COL
14 note: required by a bound in `foo`
```
This happens because the code that attempts to remap
`HASHBROWN_SRC_LOCATION` expects it to be under `hashbrown-$version`,
which is the case in a normal cargo registry, but not when vendoring, where
by default crates may not have the version in their directory name.
This change passes `--versioned-dirs` to `cargo vendor` to enforce that
every crate includes the version in the subdir name, which fixes the ui
test and brings `--enable-vendor` builds closer to normal ones.
Fix nix patching for LLVM 18
LLVM 18 now ships `libLLVM*.so.*`, so `.so` is not the sole extension anymore, which breaks the dylib detection. Oops! Adjust it to only search for `.so` somewhere.
fixes#122906
The dead_code lint was previously eroneously missing this dead code.
Since this lint bug has been fixed, the unused field need to be removed
or marked as `#[allow(dead_code)]`.
Given that this struct is deserialized without #[serde(deny_unknown_fields)]
it is ok to simply delete the never read fields.
LLVM 18 now ships `libLLVM*.so.*`, so `.so` is not the sole extension
anymore, which breaks the dylib detection. Oops! Adjust it to only
search for `.so` somewhere.
Currently, if you attempt to run ui tests in a vendored build, you will
see this failure
```
---- [ui] tests/ui/issues/issue-21763.rs stdout ----
diff of stderr:
8 = note: required because it appears within the type `(Rc<()>, Rc<()>)`
9 = note: required for `hashbrown::raw::RawTable<(Rc<()>, Rc<()>)>` to implement `Send`
10 note: required because it appears within the type `hashbrown::map::HashMap<Rc<()>, Rc<()>, RandomState>`
- --> $HASHBROWN_SRC_LOCATION
+ --> /rust/deps/hashbrown/src/map.rs:190:12
12 note: required because it appears within the type `HashMap<Rc<()>, Rc<()>>`
13 --> $SRC_DIR/std/src/collections/hash/map.rs:LL:COL
14 note: required by a bound in `foo`
```
This happend because the code that attempts to remap
`HASHBROWN_SRC_LOCATION` expects it to be under `hashbrown-$version`,
which is the case in a normal cargo registry, but not when vendor, where
by default crates may not have the version in their directory name.
This change passes `--versioned-dirs` to `cargo vendor` to enforce that
every crate includes the version in the subdir name, which fixes the ui
test and brings `--enable-vendor` builds closer to normal ones.
This is implemented with the freshly-released Wasmtime 19 and should
prevent beta breakage from wasm tests that was observed and fixed
in #122640 again.
various clippy fixes
We need to keep the order of the given clippy lint rules before passing them.
Since clap doesn't offer any useful interface for this purpose out of the box,
we have to handle it manually.
Additionally, this PR makes `-D` rules work as expected. Previously, lint rules were limited to `-W`. By enabling `-D`, clippy began to complain numerous lines in the tree, all of which have been resolved in this PR as well.
Fixes#121481
cc `@matthiaskrgr`
Update the minimum external LLVM to 17
With this change, we'll have stable support for LLVM 17 and 18.
For reference, the previous increase to LLVM 16 was #117947.
Enable frame pointers for the standard library
There's been a few past experiments for enabling frame pointers for all our artifacts. I don't think that frame pointers in the distributed compiler are nearly as useful as frame pointers in the standard library. Our users are much more likely to be profiling apps written in Rust than they are profiling the Rust compiler.
So yeah it would be cool to have frame pointers in the compiler, but much more of the value is having them on the precompiled standard library. That's what this PR does.
Previously, when passing lint rules manually using `x clippy ..`, ignored lints would
override manual ones. This change corrects the order by passing ignored lints after the
manual ones.
Signed-off-by: onur-ozkan <work@onurozkan.dev>
We need to keep the order of the given clippy lint rules before passing them.
Since clap doesn't offer any useful interface for this purpose out of the box,
we have to handle it manually.
Signed-off-by: onur-ozkan <work@onurozkan.dev>
run change tracker even when config parse fails
Please note that we are currently validating the build configuration on two entry points (e.g., profile validation is handled on the python side), and change tracker system is handled on the rust side. Once #94829 is completed (scheduled for 2024), we will be able to handle this more effectively.
Fixes#121756
mir-opt: always run tests for the current target
Currently, `./x.py test tests/mir-opt` runs only the tests for the current target, and `./x.py test tests/mir-opt --bless` runs tests for a representative set of targets. That representative set does not include the current target however, which means `--bless` can succeed when tests fail without it. This PR ensures we run the current target always.
Fixes https://github.com/rust-lang/rust/issues/122292
cc ``@RalfJung``
bootstrap: Don't name things copy that are not copies
The bootstrap copy methods don't actually copy, they just hard link. Simply lying about it being "copying" can be very confusing! (ask me how I know!).
I'm not sure whether the name I chose is the ideal name, but it's definitely better than before.
prevent notifying the same changes more than once
Prevents re-reporting of previously notified changes by using the .last-warned-change-id value for change detection.
Resolves#122344
