parser: not insert dummy field in struct
Fixes#114636
This PR eliminates the dummy field, initially introduced in #113999, thereby enabling unrestricted use of `ident.unwrap()`. A side effect of this action is that we can only report the error of the first macro invocation field within the struct node.
An alternative solution might be giving a virtual name to the macro, but it appears more complex.(https://github.com/rust-lang/rust/issues/114636#issuecomment-1670228715). Furthermore, if you think https://github.com/rust-lang/rust/issues/114636#issuecomment-1670228715 is a better solution, feel free to close this PR.
Make SIGSEGV handler emit nicer backtraces
This annotates the code heavily with comments to explain what is going on, for the benefit of other compiler contributors. The backtrace also emits appropriate comments to clarify, to a programmer who may not know why a bunch of file paths and hexadecimal blather was just dumped into stderr, what is going on. Finally, it detects cycles and uses their regularity to avoid repeating a bunch of text. The previous backtraces we were emitting was extremely unfriendly, potentially confusing, and often alarming, and this makes things almost "nice".
We can't necessarily make them much nicer than this, because a signal handler must use "signal-safe" functions. This precludes conveniences like dynamic allocations. Fortunately, Rust's stdlib has allocation-free formatting, but it may hinder integrating this error with our localization middleware, as I wasn't able to clearly ascertain, at a glance, whether there was a zero-alloc path through it.
r? `@Nilstrieb`
Adapt table sizes to the contents
This is an implementation of https://github.com/rust-lang/compiler-team/issues/666
The objective of this PR is to permit the rmeta format to accommodate larger crates that need offsets larger than a `u32` can store without compromising performance for crates that do not need such range. The second commit is a number of tiny optimization opportunities I noticed while looking at perf recordings of the first commit.
The rmeta tables need to have fixed-size elements to permit lazy random access. But the size only needs to be fixed _per table_, not per element type. This PR adds another `usize` to the table header which indicates the table element size. As each element of a table is set, we keep track of the widest encoded table value, then don't bother encoding all the unused trailing bytes on each value. When decoding table elements, we copy them to a full-width array if they are not already full-width.
`LazyArray` needs some special treatment. Most other values that are encoded in tables are indexes or offsets, and those tend to be small so we get to drop a lot of zero bytes off the end. But `LazyArray` encodes _two_ small values in a fixed-width table element: A position of the table and the length of the table. The treatment described above could trim zero bytes off the table length, but any nonzero length shields the position bytes from the optimization. To improve this, we interleave the bytes of position and length. This change is responsible for about half of the crate metadata win on many crates.
Fixes https://github.com/rust-lang/rust/issues/112934 (probably)
Fixes https://github.com/rust-lang/rust/issues/103607
Always add LC_BUILD_VERSION for metadata object files
As of Xcode 15 Apple's linker has become a bit more strict about the warnings it produces. One of those new warnings requires all valid Mach-O object files in an archive to have a LC_BUILD_VERSION load command:
```
ld: warning: no platform load command found in 'ARCHIVE[arm64][2106](lib.rmeta)', assuming: iOS-simulator
```
This was already being done for Mac Catalyst so this change expands this logic to include it for all Apple platforms. I filed this behavior change as FB12546320 and was told it was the new intentional behavior.
Tweaks and improvements on SMIR around generics_of and predicates_of
r? `@oli-obk`
This allows an API like the following ...
```rust
let trait_decls = stable_mir::all_trait_decls().iter().map(|trait_def| {
let trait_decl = stable_mir::trait_decl(trait_def);
let generics = trait_decl.generics_of();
let predicates = trait_decl.predicates_of().predicates;
```
I didn't like that much `trait_def.trait_decl()` which is it possible but adding a method to a def_id that loads up a whole trait definition looks backwards to me.
Remove conditional use of `Sharded` from query state
`Sharded` is already a zero cost abstraction, so it shouldn't affect the performance of the single thread compiler if LLVM does its job.
r? `@cjgillot`
Use `preserve_mostcc` for `extern "rust-cold"`
As experimentation in #115242 has shown looks better than `coldcc`. Notably, clang exposes `preserve_most` (https://clang.llvm.org/docs/AttributeReference.html#preserve-most) but not `cold`, so this change should put us on a better-supported path.
And *don't* use a different convention for cold on Windows, because that actually ends up making things worse. (See comment in the code.)
cc tracking issue #97544
miri ABI compatibility check: accept u32 and i32
If only the sign differs, then surely these types are compatible. (We do still check that `arg_ext` is the same, just in case.)
Also I made it so that the ABI check must *imply* that size and alignment are the same, but it doesn't actively check that itself. With how crazy ABI constraints get, having equal size and align really shouldn't be used as a signal for anything I think...
Make RPITITs capture all in-scope lifetimes
Much like #114616, this implements the lang team decision from this T-lang meeting on [opaque captures strategy moving forward](https://hackmd.io/sFaSIMJOQcuwCdnUvCxtuQ?view). This will be RFC'd soon, but given that RPITITs are a nightly feature, this shouldn't necessarily be blocked on that.
We unconditionally capture all lifetimes in RPITITs -- impl is not as simple as #114616, since we still need to duplicate RPIT lifetimes to make sure we reify any late-bound lifetimes in scope.
Closes#112194
don't use SnapshotVec in Graph implementation, as it looks unused; use Vec instead
`Graph` don't use `SnapshotVec` methods, so use simple `Vec` instead?
More precisely detect cycle errors from type_of on opaque
Not sure if this still needs work. Just putting it up for initial impressions, since it seems that a few people are frustrated with the increased error verbosity due to #113320.
Essentially we introduce a new sub-query for `type_of` specifically for opaques which returns a value that is able to distinguish "has errors" from "due to cycle recovery".
Fixes#115188
r? `@oli-obk`
codegen_llvm/llvm_type: avoid matching on the Rust type
This `match` is highly suspicious. Looking at `scalar_llvm_type_at` I think it makes no difference. But if it were to make a difference that would be a huge problem, since it doesn't look through `repr(transparent)`!
Cc `@eddyb` `@bjorn3`
Avoid duplicate `large_assignments` lints
By checking for overlapping spans.
This PR does the "reduce noisiness" task in #83518.
r? `@oli-obk` who added E-mentor and E-help-wanted and wrote the initial code.
(The fix itself is in dc82736677. The two commits before that are just small refactorings.)
Rollup of 6 pull requests
Successful merges:
- #109660 (Document that SystemTime does not count leap seconds)
- #114238 (Fix implementation of `Duration::checked_div`)
- #114512 (std/tests: disable ancillary tests on freebsd since the feature itsel…)
- #114919 (style-guide: Add guidance for defining formatting for specific macros)
- #115278 (tell people what to do when removing an error code)
- #115280 (avoid triple-backtrace due to panic-during-cleanup)
r? `@ghost`
`@rustbot` modify labels: rollup
tell people what to do when removing an error code
Currently tidy and CI send developers on a wild goose chase:
- you edit the code
- CI/tidy tells you that an error code is gone, so you remove it from the list
- CI/tidy tells you that the markdown file is stale, so you remove that as well
- CI (but not tidy) tells you not to remove an error description and copy what E0001 does
Let's be nice to people and directly tell them what to do rather than making them follow misleading breadcrumbs.
r? ``@GuillaumeGomez``
Load include_bytes! directly into an Lrc
This PR deletes an innocent-looking `.into()` that was converting from a `Vec<u8>` to `Lrc<[u8]>`. This has significant runtime and memory overhead when using `include_bytes!` to pull in a large binary file.
Add a specialization for encoding byte arrays in rmeta
This specialization already exists for FileEncoder, but since EncodeContext is implemented by forwarding down to FileEncoder, using EncodeContext used to bypass the specialization.
