CTFE eval_fn_call: use FnAbi to determine argument skipping and compatibility
This makes use of the `FnAbi` type in CTFE/Miri, which `@eddyb` has been saying for years is what we should do.^^ `FnAbi` is used to
- determine which arguments to skip (rather than the previous heuristic of skipping ZST arguments with the Rust ABI)
- impose further restrictions on whether caller and callee are consistent in how a given argument is passed
I was hoping it would also simplify the code, but that is not the case -- the previous type compatibility checks are still required (AFAIK), only the ZST skipping is gone and that took barely any code. We also need some hacks because `FnAbi` assumes a certain way of implementing `caller_location` (by passing extra arguments), but Miri can just read the caller location from the call stack so it doesn't need those arguments. (The fact that every backend has to separately implement support for these arguments seems suboptimal -- looks like this might have been better implemented on the MIR level.) To avoid having to implement those unnecessary arguments in Miri, we just compute *whether* the argument is present on the caller/callee side, but don't actually pass that argument around.
I have no idea if this looks the way `@eddyb` thinks it should look... but it makes Miri's test suite pass. ;)
One of rustc's tests fails unfortunately (`ui/const-generics/issues/issue-67739.rs`), some const generic code that is evaluated too early -- I think that should raise `TooGeneric` but instead it ICEs. My assumption is this is some FnAbi code that has not been properly tested on polymorphic code, but it might also be me calling that FnAbi code the wrong way.
r? `@oli-obk` `@eddyb`
Fixes https://github.com/rust-lang/rust/issues/56166
Miri PR at https://github.com/rust-lang/miri/pull/1928
RawVec: don't recompute capacity after allocating.
Currently it sets the capacity to `ptr.len() / mem::size_of::<T>()`
after any buffer allocation/reallocation. This would be useful if
allocators ever returned a `NonNull<[u8]>` with a size larger than
requested. But this never happens, so it's not useful.
Removing this slightly reduces the size of generated LLVM IR, and
slightly speeds up the hot path of `RawVec` growth.
r? `@ghost`
Remove useless `#[global_allocator]` from rustc and rustdoc.
This was added in #83152, which has several errors in its comments.
This commit also fix up the comments, which are quite wrong and
misleading.
r? `@alexcrichton`
Rollup of 5 pull requests
Successful merges:
- #90625 (Add `UnwindSafe` to `Once`)
- #92121 (disable test with self-referential generator on Miri)
- #92166 (Fixed a small typo in ui test comments)
- #92203 (Store a `DefId` instead of an `AdtDef` in `AggregateKind::Adt`)
- #92231 (Update books)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Update books
## nomicon
1 commits in 49681ea4a9fa81173dbe9ffed74b4d4a35eae9e3..c05c452b36358821bf4122f9c418674edd1d713d
2021-11-24 16:27:28 +0900 to 2021-12-13 15:23:48 +0900
- Update the guidance on uninitialized data with ptr::addr_of_mut (rust-lang/nomicon#325)
## reference
3 commits in 954f3d441ad880737a13e241108f791a4d2a38cd..06f9e61931bcf58b91dfe6c924057e42ce273ee1
2021-11-29 11:11:30 -0800 to 2021-12-17 07:31:40 -0800
- keep consistent for primitive types (rust-lang/reference#1118)
- README.md: link to mdbook docs (rust-lang/reference#1117)
- Say that `...` range patterns are rejected in the 2021 edition (rust-lang/reference#1114)
## book
4 commits in 5f9358faeb1f46e19b8a23a21e79fd7fe150491e..8a0bb3c96e71927b80fa2286d7a5a5f2547c6aa4
2021-12-05 21:33:16 -0500 to 2021-12-22 20:54:27 -0500
- Propagate edits back
- Fix number disagreement. Fixesrust-lang/book#2858.
- Wrap some code in main to make scopes clearer. Fixesrust-lang/book#2830.
- Respond to ch5 nostarch edits
## rustc-dev-guide
9 commits in a374e7d8bb6b79de45b92295d06b4ac0ef35bc09..9bf0028b557798ddd07a6f652e4d0c635d3d6620
2021-12-03 09:26:47 -0800 to 2021-12-20 21:53:57 +0900
- remove rustfix item in test intro (rust-lang/rustc-dev-guide#1277)
- Move date-check comment to fix Markdown syntax
- Update humor docs for special-casing ferris emoji
- Fix some broken links (rust-lang/rustc-dev-guide#1274)
- Update rustdoc internals
- Update HIR chapter to use `HirId` instead of `NodeId`
- Fix some broken links
- Update src/getting-started.md
- Improve documentation on r?
Store a `DefId` instead of an `AdtDef` in `AggregateKind::Adt`
The `AggregateKind` enum ends up in the final mir `Body`. Currently,
any changes to `AdtDef` (regardless of how significant they are)
will legitimately cause the overall result of `optimized_mir` to change,
invalidating any codegen re-use involving that mir.
This will get worse once we start hashing the `Span` inside `FieldDef`
(which is itself contained in `AdtDef`).
To try to reduce these kinds of invalidations, this commit changes
`AggregateKind::Adt` to store just the `DefId`, instead of the full
`AdtDef`. This allows the result of `optimized_mir` to be unchanged
if the `AdtDef` changes in a way that doesn't actually affect any
of the MIR we build.
disable test with self-referential generator on Miri
Running the libcore test suite in Miri currently fails due to the known incompatibility of self-referential generators with Miri's aliasing checks (https://github.com/rust-lang/unsafe-code-guidelines/issues/148). So let's disable that test in Miri for now.
Backport LLVM changes to disable deferred inlining
Fixes#91128
I was thinking of how to best add the test case from the issue, and I think rust perf infrastructure would probably be the best place for something like it.
Update chalk to 0.75.0
- Compute flags in `intern_ty`
- Remove `tracing-serde` from `PERMITTED_DEPENDENCIES`
- Bump `tracing-tree` to 0.2.0
- Bump `tracing-subscriber` to 0.3.3
Use panic() instead of panic!() in some places in core.
See https://github.com/rust-lang/rust/pull/92068 and https://github.com/rust-lang/rust/pull/92140.
This avoids the `panic!()` macro in a few potentially hot paths. This becomes more relevant when switching `core` to Rust 2021, as it'll avoid format_args!() and save some compilation time. (It doesn't make a huge difference, but still.) (Also the errors in const panic become slightly nicer.)
Quote bat script command line
Fixes#91991
[`CreateProcessW`](https://docs.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-createprocessw#parameters) should only be used to run exe files but it does have some (undocumented) special handling for files with `.bat` and `.cmd` extensions. Essentially those magic extensions will cause the parameters to be automatically rewritten. Example pseudo Rust code (note that `CreateProcess` starts with an optional application name followed by the application arguments):
```rust
// These arguments...
CreateProcess(None, `@"foo.bat` "hello world""`@,` ...);
// ...are rewritten as
CreateProcess(Some(r"C:\Windows\System32\cmd.exe"), `@""foo.bat` "hello world"""`@,` ...);
```
However, when setting the first parameter (the application name) as we now do, it will omit the extra level of quotes around the arguments:
```rust
// These arguments...
CreateProcess(Some("foo.bat"), `@"foo.bat` "hello world""`@,` ...);
// ...are rewritten as
CreateProcess(Some(r"C:\Windows\System32\cmd.exe"), `@"foo.bat` "hello world""`@,` ...);
```
This means the arguments won't be passed to the script as intended.
Note that running batch files this way is undocumented but people have relied on this so we probably shouldn't break it.
Don't emit shared files when scraping examples from dependencies in Rustdoc
This PR fixes#91605. The issue is that `Context::init` gets called when scraping dependencies. By default, just calling `init` calls into `write_shared` and `build_index` which register the scraped crate into a list that later gets used for the Rustdoc sidebar. The fix is to ensure that `write_shared` is not called when scraping.
r? `@jyn514`
Change Backtrace::enabled atomic from SeqCst to Relaxed
This atomic is not synchronizing anything outside of its own value, so we don't need the `Acquire`/`Release` guarantee that all memory operations prior to the store are visible after the subsequent load, nor the `SeqCst` guarantee of all threads seeing all of the sequentially consistent operations in the same order.
Using `Relaxed` reduces the overhead of `Backtrace::capture()` in the case that backtraces are not enabled.
## Benchmark
```rust
#![feature(backtrace)]
use std::backtrace::Backtrace;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::thread;
use std::time::Instant;
fn main() {
let begin = Instant::now();
let mut threads = Vec::new();
for _ in 0..64 {
threads.push(thread::spawn(|| {
for _ in 0..10_000_000 {
let _ = Backtrace::capture();
static LOL: AtomicUsize = AtomicUsize::new(0);
LOL.store(1, Ordering::Release);
}
}));
}
for thread in threads {
let _ = thread.join();
}
println!("{:?}", begin.elapsed());
}
```
**Before:** 6.73 seconds
**After:** 5.18 seconds
Add some JSDoc comments to rustdoc JS
This follows the Closure Compiler dialect of JSDoc, so we can use it to do some basic type checking. We don't plan to compile with Closure Compiler, just use it to check types. See https://github.com/google/closure-compiler/wiki/ for details.
To try checking the annotations, run:
```
npm i -g google-closure-compiler
google-closure-compiler -W VERBOSE build/x86_64-unknown-linux-gnu/doc/{search-index1.59.0.js,crates1.59.0.js} src/librustdoc/html/static/js/{search.js,main.js,storage.js} --externs src/librustdoc/html/static/js/externs.js >/dev/null
```
You'll see some warnings that "String continuations are not recommended". I'm not addressing those right now.
[Discussed on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/doc.20format.20for.20JS/near/265209466).
r? `@GuillaumeGomez`
Fix duplicate derive clone suggestion
closes https://github.com/rust-lang/rust/issues/91492
The addition of:
```rust
derives.sort();
derives.dedup();
```
is what actually solves the problem.
The rest is just cleanup.
I want to improve the diagnostic message to provide the suggestion as a proper diff but ran into some problems, so I'll attempt that again in a follow up PR.
Allow reverse iteration of lowercase'd/uppercase'd chars
The PR implements `DoubleEndedIterator` trait for `ToLowercase` and `ToUppercase`.
This enables reverse iteration of lowercase/uppercase variants of character sequences.
One of use cases: determining whether a char sequence is a suffix of another one.
Example:
```rust
fn endswith_ignore_case(s1: &str, s2: &str) -> bool {
for eob in s1
.chars()
.flat_map(|c| c.to_lowercase())
.rev()
.zip_longest(s2.chars().flat_map(|c| c.to_lowercase()).rev())
{
match eob {
EitherOrBoth::Both(c1, c2) => {
if c1 != c2 {
return false;
}
}
EitherOrBoth::Left(_) => return true,
EitherOrBoth::Right(_) => return false,
}
}
true
}
```
This follows the Closure Compiler dialect of JSDoc, so we
can use it to do some basic type checking. We don't plan to
compile with Closure Compiler, just use it to check types. See
https://github.com/google/closure-compiler/wiki/ for details.
The `AggregateKind` enum ends up in the final mir `Body`. Currently,
any changes to `AdtDef` (regardless of how significant they are)
will legitimately cause the overall result of `optimized_mir` to change,
invalidating any codegen re-use involving that mir.
This will get worse once we start hashing the `Span` inside `FieldDef`
(which is itself contained in `AdtDef`).
To try to reduce these kinds of invalidations, this commit changes
`AggregateKind::Adt` to store just the `DefId`, instead of the full
`AdtDef`. This allows the result of `optimized_mir` to be unchanged
if the `AdtDef` changes in a way that doesn't actually affect any
of the MIR we build.
Fix failing tests
Fixes tests that have been failing in CI.
I hope it's ok but I've temporarily commented out some of the Windows resolver tests. I actually have a bigger fix for that test code as part of another PR. I could separate it out and patch as necessary but I'd prefer not to rush into that if possible.
r? `@Mark-Simulacrum`
Currently it sets the capacity to `ptr.len() / mem::size_of::<T>()`
after any buffer allocation/reallocation. This would be useful if
allocators ever returned a `NonNull<[u8]>` with a size larger than
requested. But this never happens, so it's not useful.
Removing this slightly reduces the size of generated LLVM IR, and
slightly speeds up the hot path of `RawVec` growth.
