Note: there was an existing code path involving `Interpolated` in
`MetaItem::from_tokens` that was dead. This commit transfers that to the
new form, but puts an `unreachable!` call inside it.
The one notable test change is `tests/ui/macros/trace_faulty_macros.rs`.
This commit removes the complicated `Interpolated` handling in
`expected_expression_found` that results in a longer error message. But
I think the new, shorter message is actually an improvement.
The original complaint was in #71039, when the error message started
with "error: expected expression, found `1 + 1`". That was confusing
because `1 + 1` is an expression. Other than that, the reporter said
"the whole error message is not too bad if you ignore the first line".
Subsequently, extra complexity and wording was added to the error
message. But I don't think the extra wording actually helps all that
much. In particular, it still says of the `1+1` that "this is expected
to be expression". This repeats the problem from the original complaint!
This commit removes the extra complexity, reverting to a simpler error
message. This is primarily because the traversal is a pain without
`Interpolated` tokens. Nonetheless, I think the error message is
*improved*. It now starts with "expected expression, found `pat`
metavariable", which is much clearer and the real problem. It also
doesn't say anything specific about `1+1`, which is good, because the
`1+1` isn't really relevant to the error -- it's the `$e:pat` that's
important.
It mirrors `ExprKind::Binary`, and contains a `BinOpKind`. This makes
`AssocOp` more like `ExprKind`. Note that the variants removed from
`AssocOp` are all named differently to `BinOpToken`, e.g. `Multiply`
instead of `Mul`, so that's an inconsistency removed.
The commit adds `precedence` and `fixity` methods to `BinOpKind`, and
calls them from the corresponding methods in `AssocOp`. This avoids the
need to create an `AssocOp` from a `BinOpKind` in a bunch of places, and
`AssocOp::from_ast_binop` is removed.
`AssocOp::to_ast_binop` is also no longer needed.
Overall things are shorter and nicer.
`AssocOp::AssignOp` contains a `BinOpToken`. `ExprKind::AssignOp`
contains a `BinOpKind`. Given that `AssocOp` is basically a cut-down
version of `ExprKind`, it makes sense to make `AssocOp` more like
`ExprKind`. Especially given that `AssocOp` and `BinOpKind` use semantic
operation names (e.g. `Mul`, `Div`), but `BinOpToken` uses syntactic
names (e.g. `Star`, `Slash`).
This results in more concise code, and removes the need for various
conversions. (Note that the removed functions `hirbinop2assignop` and
`astbinop2assignop` are semantically identical, because `hir::BinOp` is
just a synonum for `ast::BinOp`!)
The only downside to this is that it allows the possibility of some
nonsensical combinations, such as `AssocOp::AssignOp(BinOpKind::Lt)`.
But `ExprKind::AssignOp` already has that problem. The problem can be
fixed for both types in the future with some effort, by introducing an
`AssignOpKind` type.
Use `edition = "2024"` in the compiler (redux)
Most of this is binding mode changes, which I fixed by running `x.py fix`.
Also adds some miscellaneous `unsafe` blocks for new unsafe standard library functions (the setenv ones), and a missing `unsafe extern` block in some enzyme codegen code, and fixes some precise capturing lifetime changes (but only when they led to errors).
cc ``@ehuss`` ``@traviscross``
test building enzyme in CI
1) This PR fixes a significant compile-time regression, by only running the expensive autodiff pipeline, if the users pass the newly introduced Enable value to the `-Zautodiff=` flag. It updates the test(s) accordingly. It gives a nice error if users forget that.
2) It fixes macos support by explicitly linking against the Enzyme build folder. This doesn't cover CI macos yet.
3) It fixes the issue that setting ENZYME_RUNPASS was ignored by enzyme and in fact did not schedule enzyme's opt pass.
4) It also re-enables support for various other values for the autodiff flag, which were ignored since the refactor.
5) I merged some improvements to Enzyme core, which means we do not longer depend on LLVM being build with the Plugin Interface enabled.
6) Unrelated to other fixes, this changes `rustc_autodiff` to `EncodeCrossCrate::Yes`. It is not enough on it's own to enable usage of Enzyme in libraries, but it is for sure a piece of the fixes needed to get this to work.
try-job: x86_64-gnu
r? `@oli-obk`
Tracking:
- https://github.com/rust-lang/rust/issues/124509
Notes about tests:
- tests/ui/parser/macro/trait-object-macro-matcher.rs: the syntax error
is duplicated, because it occurs now when parsing the decl macro
input, and also when parsing the expanded decl macro. But this won't
show up for normal users due to error de-duplication.
- tests/ui/associated-consts/issue-93835.rs: similar, plus there are
some additional errors about this very broken code.
- The changes to metavariable descriptions in #132629 are now visible in
error message for several tests.
adding autodiff tests
I'd like to get started with upstreaming some tests, even though I'm still waiting for an answer on how to best integrate the enzyme pass. Can we therefore temporarily support the -Z llvm-plugins here without too much effort? And in that case, how would that work? I saw you can do remapping, e.g. `rust-src-base`, but I don't think that will give me the path to libEnzyme.so. Do you have another suggestion?
Other than that this test simply checks that the derivative of `x*x` is `2.0 * x`, which in this case is computed as
`%0 = fadd fast double %x.0.val, %x.0.val`
(I'll add a few more tests and move it to an autodiff folder if we can use the -Z flag)
r? ``@jieyouxu``
Locally at least `-Zllvm-plugins=${PWD}/build/x86_64-unknown-linux-gnu/enzyme/build/Enzyme/libEnzyme-19.so` seems to work if I copy the command I get from x.py test and run it manually. However, running x.py test itself fails.
Tracking:
- https://github.com/rust-lang/rust/issues/124509
Zulip discussion: https://rust-lang.zulipchat.com/#narrow/channel/326414-t-infra.2Fbootstrap/topic/Enzyme.20build.20changes
tree-wide: parallel: Fully removed all `Lrc`, replaced with `Arc`
tree-wide: parallel: Fully removed all `Lrc`, replaced with `Arc`
This is continuation of https://github.com/rust-lang/rust/pull/132282 .
I'm pretty sure I did everything right. In particular, I searched all occurrences of `Lrc` in submodules and made sure that they don't need replacement.
There are other possibilities, through.
We can define `enum Lrc<T> { Rc(Rc<T>), Arc(Arc<T>) }`. Or we can make `Lrc` a union and on every clone we can read from special thread-local variable. Or we can add a generic parameter to `Lrc` and, yes, this parameter will be everywhere across all codebase.
So, if you think we should take some alternative approach, then don't merge this PR. But if it is decided to stick with `Arc`, then, please, merge.
cc "Parallel Rustc Front-end" ( https://github.com/rust-lang/rust/issues/113349 )
r? SparrowLii
`@rustbot` label WG-compiler-parallel
includes post-developed commit: do not suggest internal-only keywords as corrections to parse failures.
includes post-developed commit: removed tabs that creeped in into rustfmt tool source code.
includes post-developed commit, placating rustfmt self dogfooding.
includes post-developed commit: add backquotes to prevent markdown checking from trying to treat an attr as a markdown hyperlink/
includes post-developed commit: fix lowering to keep contracts from being erroneously inherited by nested bodies (like closures).
Rebase Conflicts:
- compiler/rustc_parse/src/parser/diagnostics.rs
- compiler/rustc_parse/src/parser/item.rs
- compiler/rustc_span/src/hygiene.rs
Remove contracts keywords from diagnostic messages
Autodiff Upstreaming - rustc_codegen_ssa, rustc_middle
This PR should not be merged until the rustc_codegen_llvm part is merged.
I will also alter it a little based on what get's shaved off from the cg_llvm PR,
and address some of the feedback I received in the other PR (including cleanups).
I am putting it already up to
1) Discuss with `@jieyouxu` if there is more work needed to add tests to this and
2) Pray that there is someone reviewing who can tell me why some of my autodiff invocations get lost.
Re 1: My test require fat-lto. I also modify the compilation pipeline. So if there are any other llvm-ir tests in the same compilation unit then I will likely break them. Luckily there are two groups who currently have the same fat-lto requirement for their GPU code which I have for my autodiff code and both groups have some plans to enable support for thin-lto. Once either that work pans out, I'll copy it over for this feature. I will also work on not changing the optimization pipeline for functions not differentiated, but that will require some thoughts and engineering, so I think it would be good to be able to run the autodiff tests isolated from the rest for now. Can you guide me here please?
For context, here are some of my tests in the samples folder: https://github.com/EnzymeAD/rustbook
Re 2: This is a pretty serious issue, since it effectively prevents publishing libraries making use of autodiff: https://github.com/EnzymeAD/rust/issues/173. For some reason my dummy code persists till the end, so the code which calls autodiff, deletes the dummy, and inserts the code to compute the derivative never gets executed. To me it looks like the rustc_autodiff attribute just get's dropped, but I don't know WHY? Any help would be super appreciated, as rustc queries look a bit voodoo to me.
Tracking:
- https://github.com/rust-lang/rust/issues/124509
r? `@jieyouxu`
remove support for the (unstable) #[start] attribute
As explained by `@Noratrieb:`
`#[start]` should be deleted. It's nothing but an accidentally leaked implementation detail that's a not very useful mix between "portable" entrypoint logic and bad abstraction.
I think the way the stable user-facing entrypoint should work (and works today on stable) is pretty simple:
- `std`-using cross-platform programs should use `fn main()`. the compiler, together with `std`, will then ensure that code ends up at `main` (by having a platform-specific entrypoint that gets directed through `lang_start` in `std` to `main` - but that's just an implementation detail)
- `no_std` platform-specific programs should use `#![no_main]` and define their own platform-specific entrypoint symbol with `#[no_mangle]`, like `main`, `_start`, `WinMain` or `my_embedded_platform_wants_to_start_here`. most of them only support a single platform anyways, and need cfg for the different platform's ways of passing arguments or other things *anyways*
`#[start]` is in a super weird position of being neither of those two. It tries to pretend that it's cross-platform, but its signature is a total lie. Those arguments are just stubbed out to zero on ~~Windows~~ wasm, for example. It also only handles the platform-specific entrypoints for a few platforms that are supported by `std`, like Windows or Unix-likes. `my_embedded_platform_wants_to_start_here` can't use it, and neither could a libc-less Linux program.
So we have an attribute that only works in some cases anyways, that has a signature that's a total lie (and a signature that, as I might want to add, has changed recently, and that I definitely would not be comfortable giving *any* stability guarantees on), and where there's a pretty easy way to get things working without it in the first place.
Note that this feature has **not** been RFCed in the first place.
*This comment was posted [in May](https://github.com/rust-lang/rust/issues/29633#issuecomment-2088596042) and so far nobody spoke up in that issue with a usecase that would require keeping the attribute.*
Closes https://github.com/rust-lang/rust/issues/29633
try-job: x86_64-gnu-nopt
try-job: x86_64-msvc-1
try-job: x86_64-msvc-2
try-job: test-various
Account for C string literals and `format_args` in `HiddenUnicodeCodepoints` lint
This is stacked on #134955, and either that can land first or both of them can land together here. I split this out because this is a bit more involved of an impl.
Fixes#94945
Precedence improvements: closures and jumps
This PR fixes some cases where rustc's pretty printers would redundantly parenthesize expressions that didn't need it.
<table>
<tr><th>Before</th><th>After</th></tr>
<tr><td><code>return (|x: i32| x)</code></td><td><code>return |x: i32| x</code></td></tr>
<tr><td><code>(|| -> &mut () { std::process::abort() }).clone()</code></td><td><code>|| -> &mut () { std::process::abort() }.clone()</code></td></tr>
<tr><td><code>(continue) + 1</code></td><td><code>continue + 1</code></td></tr>
</table>
Tested by `echo "fn main() { let _ = $AFTER; }" | rustc -Zunpretty=expanded /dev/stdin`.
The pretty-printer aims to render the syntax tree as it actually exists in rustc, as faithfully as possible, in Rust syntax. It can insert parentheses where forced by Rust's grammar in order to preserve the meaning of a macro-generated syntax tree, for example in the case of `a * $rhs` where $rhs is `b + c`. But for any expression parsed from source code, without a macro involved, there should never be a reason for inserting additional parentheses not present in the original.
For closures and jumps (return, break, continue, yield, do yeet, become) the unneeded parentheses came from the precedence of some of these expressions being misidentified. In the same order as the table above:
- Jumps and closures are supposed to have equal precedence. The [Rust Reference](https://doc.rust-lang.org/1.83.0/reference/expressions.html#expression-precedence) says so, and in Syn they do. There is no Rust syntax that would require making a precedence distinction between jumps and closures. But in rustc these were previously 2 distinct levels with the closure being lower, hence the parentheses around a closure inside a jump (but not a jump inside a closure).
- When a closure is written with an explicit return type, the grammar [requires](https://doc.rust-lang.org/1.83.0/reference/expressions/closure-expr.html) that the closure body consists of exactly one block expression, not any other arbitrary expression as usual for closures. Parsing of the closure body does not continue after the block expression. So in `|| { 0 }.clone()` the clone is inside the closure body and applies to `{ 0 }`, whereas in `|| -> _ { 0 }.clone()` the clone is outside and applies to the closure as a whole.
- Continue never needs parentheses. It was previously marked as having the lowest possible precedence but it should have been the highest, next to paths and loops and function calls, not next to jumps.
Overhaul keyword handling
The compiler's list of keywords has some problems.
- It contains several items that aren't keywords.
- The order isn't quite right in a couple of places.
- Some of the names of predicates relating to keywords are confusing.
- rustdoc and rustfmt have their own (incorrect) versions of the keyword list.
- `AllKeywords` is unnecessarily complex.
r? ```@jieyouxu```
