Allow custom default address spaces and parse `p-` specifications in the datalayout string
Some targets, such as CHERI, use as default an address space different from the "normal" default address space `0` (in the case of CHERI, [200 is used](https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-877.pdf)). Currently, `rustc` does not allow to specify custom address spaces and does not take into consideration [`p-` specifications in the datalayout string](https://llvm.org/docs/LangRef.html#langref-datalayout).
This patch tries to mitigate these problems by allowing targets to define a custom default address space (while keeping the default value to address space `0`) and adding the code to parse the `p-` specifications in `rustc_abi`. The main changes are that `TargetDataLayout` now uses functions to refer to pointer-related informations, instead of having specific fields for the size and alignment of pointers in the default address space; furthermore, the two `pointer_size` and `pointer_align` fields in `TargetDataLayout` are replaced with an `FxHashMap` that holds info for all the possible address spaces, as parsed by the `p-` specifications.
The potential performance drawbacks of not having ad-hoc fields for the default address space will be tested in this PR's CI run.
r? workingjubilee
Disable download-rustc for library profile
The feature currently completely breaks `x test` (rust-lang/rust#142505), core functionality of working on the standard library. Therefore it should be disabled by default until that problem is fixed. Having to wait a bit longer for a check build is nothing compared to completely mysterious build errors when testing.
compiler: Deduplicate `must_emit_unwind_tables()` comments
There is one comment at a call site and one comment in the function definition that are mostly saying the same thing. Fold the call site comment into the function definition comment to reduce duplication.
There are actually some inaccuracies in the comments but let's deduplicate before we address the inaccuracies.
mbe: Refactors and function extractions in `compile_declarative_macro`
These refactors help pave the way for parsing attribute rules.
Best reviewed commit-by-commit.
- **mbe: Simplify compile_declarative_macro by factoring out some variables**
- **mbe: Factor out a helper to check an LHS**
- **mbe: Factor out a helper to check for unexpected EOF in definition**
- **mbe: Clarify comments about error handling in `compile_declarative_macro`**
Dont resolve instance of root in `mir_callgraph_cyclic`
`Instance::try_resolve` on a default trait body method will always fail, since it's still possible to further substitute. This leads to a cycle, since in `tests/mir-opt/inline_default_trait_body.rs`, both `Trait::a` and `Trait::b` need to consider the other to be cyclical, but since we couldn't resolve a body, we'd just consider *nothing* to be cyclical.
The root instance we care about when computing `mir_callgraph_cyclic` is trivial to compute (it's just `InstanceKind::Item`), so just replace it with a call to `Instance::new_raw`.
r? `@cjgillot` `@oli-obk`
Fixesrust-lang/rust#143534
interpret: rename StackPopCleanup
The name `StackPopCleanup` stopped making sense a long time ago IMO -- in the common case, it has nothing to do with "cleanup", and everything with where the program should jump next. If we didn't have unwinding this would be just the return block, but given that we do have unwinding I figured maybe "continuation" would be a good name. This comes up in [continuation-passing style](https://en.wikipedia.org/wiki/Continuation-passing_style) and refers to where the program will *continue* when a function is done. So from a PL perspective it is the most fitting term I think -- but it may be too jargony.
r? `@oli-obk` what do you think?
Move `stable_mir` back to its own crate
We've finished the refactoring, so it's time to move `stable_mir` back to its own crate.
This PR leaves an empty `rustc_internal` module with a `#[deprecated]` attribute in `rustc_smir` to let users know we just moved it to `stable_mir`.
Fix short linker error output
This PR does 2 things:
- It removes the braces when there's a single element. This is required since brace expansion (at least in bash and zsh) only triggers if there's at least 2 elements.
- It removes the extra `.rlib` suffixes of the elements. See https://github.com/rust-lang/rust/pull/135707#discussion_r2185212393 for context.
Running `cargo +stage1 build` on the following program:
```rust
unsafe extern "C" {
fn foo() -> libc::c_int;
}
fn main() {
let x = unsafe { foo() } as u32;
// println!("{}", data_encoding::BASE64.encode(&x.to_le_bytes()));
}
```
Gives the following diff before and after the PR:
```diff
-/tmp/foo/target/debug/deps/{liblibc-faf416f178830595.rlib}.rlib
+/tmp/foo/target/debug/deps/liblibc-faf416f178830595.rlib
```
Running on the same program with the additional dependency, we get the following diff:
```diff
-/tmp/foo/target/debug/deps/{liblibc-faf416f178830595.rlib,libdata_encoding-84bb5aadfa9e8839.rlib}.rlib
+/tmp/foo/target/debug/deps/{liblibc-faf416f178830595,libdata_encoding-84bb5aadfa9e8839}.rlib
```
Get rid of build-powerpc64le-toolchain.sh
The dist-powerpc64le-linux-musl runner never actually used the toolchain that the script produced, it instead used the one from crosstool-ng.
The dist-powerpc64le-linux-gnu runner did use it, from what I can tell mainly to get a glibc 2.17 version with ppc64le support backported. Since crosstool-ng has the necessary patches, we can just use crosstool-ng to get an appropriate toolchain. While at it, use kernel 3.10 headers since that's the version documented in platform support for this target.
try-job: dist-powerpc64le-linux-gnu
try-job: dist-powerpc64le-linux-musl
doc(std): clarify `NonZero<T>` usage limitation in doc comment
Closesrust-lang/rust#142966
This PR clarifies `NonZero<T>` usage limitation in doc comment and fixes a typo.
r? libs
Only work-steal in the main loop for rustc_thread_pool
This PR is a replica of <https://github.com/rust-lang/rustc-rayon/pull/12> that only retained work-steal in the main loop for rustc_thread_pool.
r? `@oli-obk`
cc `@SparrowLii` `@Zoxc` `@cuviper`
Updates rust-lang/rust#113349
The feature currently completely breaks `x test`, core functionality of working on the standard library.
Therefore it should be disabled by default until that problem is fixed.
Having to wait a bit longer for a check build is nothing compared to completely mysterious build errors when testing.
There is one comment at a call site and one comment in the function
definition that are mostly saying the same thing. Fold the call site
comment into the function definition comment to reduce duplication.
There are actually some inaccuracies in the comments but let's
deduplicate before we address the inaccuracies.
lib: more eagerly return `self.len()` from `ceil_char_boundary`
There is no reason to go through the complicated branch as it would
always return `self.len()` in this case. Also helps debug code somewhat
and I guess might make optimizations easier (although I haven't really a
sample to demonstrate this.)
ref. https://github.com/rust-lang/rust/issues/93743
Suggested by `@chrisduerr`
compiler: rename BareFn to FnPtr
At some point "BareFn" was the chosen name for a "bare" function, without the niceties of `~fn`, `&fn`, or a few other ways of writing a function type. However, at some point the syntax for a "bare function" and any other function diverged even more. We started calling them what they are: function pointers, denoted by their own syntax.
However, we never changed the *internal* name for these, as this divergence was very gradual. Personally, I have repeatedly searched for "FnPtr" and gotten confused until I find the name is BareFn, only to forget this until the next time, since I don't routinely interact with the higher-level AST and HIR. But even tools that interact with these internal types only touch on them in a few places, making a migration easy enough. Let's use a more intuitive and obvious name, as this 12+ year old name has little to do with current Rust.
Renamed retain_mut to retain on LinkedList as mentioned in the ACP
This is for proposal: https://github.com/rust-lang/libs-team/issues/250
The original check-in (https://github.com/rust-lang/rust/pull/114136) contained both methods **retain** and **retain_mut**, which does not conform to https://github.com/rust-lang/libs-team/issues/250#issuecomment-1766822671.
I updated the retain documentation to specify **&mut e**, removed the **retain** method and renamed **retain_mut** to **retain** to conform to the request.
The pull request doesn't really contain much that is new, just removes the unwanted method to meet the requirements.
I've run the tests "library/alloc" on the code and no issues.
Hopefully I'm not stepping on the original author's toes. I just ran across a need for the method and wondered why it was unstable.
mbe: Change `unused_macro_rules` to a `DenseBitSet`
Now that it only contains indexes, and no other information, a bitset provides a more compact and simpler representation.
This builds on <https://github.com/rust-lang/rust/pull/143416>. Only the last commit is new.
Align attr fixes
- Remove references to the superseded `repr(align)` syntax
- Allow the attribute on fn items in `extern` blocks
- Test attribute in combination with `async fn` and `dyn`
r? workingjubilee
Tracking issue: https://github.com/rust-lang/rust/issues/82232
`@rustbot` label A-attributes F-fn_align T-compiler
There is no reason to go through the complicated branch as it would
always return `self.len()` in this case. Also helps debug code somewhat
and I guess might make optimizations easier (although I haven't really a
sample to demonstrate this.)
ref. #93743
Suggested by @chrisduerr
Fix some comments and related types and locals where it is obvious, e.g.
- bare_fn -> fn_ptr
- LifetimeBinderKind::BareFnType -> LifetimeBinderKind::FnPtrType
Co-authored-by: León Orell Valerian Liehr <me@fmease.dev>
