fix least significant digits of f128 associated constants
While the numbers are parsed to the correct value, the decimal numbers in the source were rounded to zero instead of to the nearest, making the literals different from the values shown in the documentation.
Rollup of 6 pull requests
Successful merges:
- #124921 (offset_from: always allow pointers to point to the same address)
- #127407 (Make parse error suggestions verbose and fix spans)
- #127684 (consolidate miri-unleashed tests for mutable refs into one file)
- #127729 (Stop using the `gen` identifier in the compiler)
- #127736 (Add myself to the review rotation)
- #127758 (coverage: Restrict `ExpressionUsed` simplification to `Code` mappings)
r? `@ghost`
`@rustbot` modify labels: rollup
coverage: Restrict `ExpressionUsed` simplification to `Code` mappings
In the future, branch and MC/DC mappings might have expressions that don't correspond to any single point in the control-flow graph. That makes it trickier to keep track of which expressions should expect an `ExpressionUsed` node.
We therefore sidestep that complexity by only performing `ExpressionUsed` simplification for expressions associated directly with ordinary `Code` mappings.
(This simplification step is inherited from the original coverage implementation, which only supported `Code` mappings anyway, so there's no particular reason to extend it to other kinds of mappings unless we specifically choose to.)
Relevant to:
- #124154
- #126677
- #124278
```@rustbot``` label +A-code-coverage
Stop using the `gen` identifier in the compiler
In preparation for edition 2024, this PR previews the fallout of removing usages of `gen` since it's being reserved as a keyword.
There are two notable changes here:
1. Had to rename `fn gen(..)` in gen/kill analysis to `gen_`. Not certain there's a better name than that.
2. There are (false?[^1]) positives in `rustc_macros` when using synstructure, which uses `gen impl` to mark an implementation. We could suppress this in a one-off way, or perhaps just ignore `gen` in macros altogether, since if an identifier ends up in expanded code then it'll get properly denied anyways.
Not relevant to the compiler, but it's gonna be really annoying to change `rand`'s `gen` fn in the library and miri...
[^1]: I haven't looked at the synstructure proc macro code itself so I'm not certain if it'll start to fail when converted to ed2024 (or, e.g., when syn starts parsing `gen` as a kw).
Make parse error suggestions verbose and fix spans
Go over all structured parser suggestions and make them verbose style.
When suggesting to add or remove delimiters, turn them into multiple suggestion parts.
offset_from: always allow pointers to point to the same address
This PR implements the last remaining part of the t-opsem consensus in https://github.com/rust-lang/unsafe-code-guidelines/issues/472: always permits offset_from when both pointers have the same address, no matter how they are computed. This is required to achieve *provenance monotonicity*.
Tracking issue: https://github.com/rust-lang/rust/issues/117945
### What is provenance monotonicity and why does it matter?
Provenance monotonicity is the property that adding arbitrary provenance to any no-provenance pointer must never make the program UB. More specifically, in the program state, data in memory is stored as a sequence of [abstract bytes](https://rust-lang.github.io/unsafe-code-guidelines/glossary.html#abstract-byte), where each byte can optionally carry provenance. When a pointer is stored in memory, all of the bytes it is stored in carry that provenance. Provenance monotonicity means: if we take some byte that does not have provenance, and give it some arbitrary provenance, then that cannot change program behavior or introduce UB into a UB-free program.
We care about provenance monotonicity because we want to allow the optimizer to remove provenance-stripping operations. Removing a provenance-stripping operation effectively means the program after the optimization has provenance where the program before the optimization did not -- since the provenance removal does not happen in the optimized program. IOW, the compiler transformation added provenance to previously provenance-free bytes. This is exactly what provenance monotonicity lets us do.
We care about removing provenance-stripping operations because `*ptr = *ptr` is, in general, (likely) a provenance-stripping operation. Specifically, consider `ptr: *mut usize` (or any integer type), and imagine the data at `*ptr` is actually a pointer (i.e., we are type-punning between pointers and integers). Then `*ptr` on the right-hand side evaluates to the data in memory *without* any provenance (because [integers do not have provenance](https://rust-lang.github.io/rfcs/3559-rust-has-provenance.html#integers-do-not-have-provenance)). Storing that back to `*ptr` means that the abstract bytes `ptr` points to are the same as before, except their provenance is now gone. This makes `*ptr = *ptr` a provenance-stripping operation (Here we assume `*ptr` is fully initialized. If it is not initialized, evaluating `*ptr` to a value is UB, so removing `*ptr = *ptr` is trivially correct.)
### What does `offset_from` have to do with provenance monotonicity?
With `ptr = without_provenance(N)`, `ptr.offset_from(ptr)` is always well-defined and returns 0. By provenance monotonicity, I can now add provenance to the two arguments of `offset_from` and it must still be well-defined. Crucially, I can add *different* provenance to the two arguments, and it must still be well-defined. In other words, this must always be allowed: `ptr1.with_addr(N).offset_from(ptr2.with_addr(N))` (and it returns 0). But the current spec for `offset_from` says that the two pointers must either both be derived from an integer or both be derived from the same allocation, which is not in general true for arbitrary `ptr1`, `ptr2`.
To obtain provenance monotonicity, this PR hence changes the spec for offset_from to say that if both pointers have the same address, the function is always well-defined.
### What further consequences does this have?
It means the compiler can no longer transform `end2 = begin.offset(end.offset_from(begin))` into `end2 = end`. However, it can still be transformed into `end2 = begin.with_addr(end.addr())`, which later parts of the backend (when provenance has been erased) can trivially turn into `end2 = end`.
The only alternative I am aware of is a fundamentally different handling of zero-sized accesses, where a "no provenance" pointer is not allowed to do zero-sized accesses and instead we have a special provenance that indicates "may be used for zero-sized accesses (and nothing else)". `offset` and `offset_from` would then always be UB on a "no provenance" pointer, and permit zero-sized offsets on a "zero-sized provenance" pointer. This achieves provenance monotonicity. That is, however, a breaking change as it contradicts what we landed in https://github.com/rust-lang/rust/pull/117329. It's also a whole bunch of extra UB, which doesn't seem worth it just to achieve that transformation.
### What about the backend?
LLVM currently doesn't have an intrinsic for pointer difference, so we anyway cast to integer and subtract there. That's never UB so it is compatible with any relaxation we may want to apply.
If LLVM gets a `ptrsub` in the future, then plausibly it will be consistent with `ptradd` and [consider two equal pointers to be inbounds](https://github.com/rust-lang/rust/pull/124921#issuecomment-2205795829).
In the future, branch and MC/DC mappings might have expressions that don't
correspond to any single point in the control-flow graph. That makes it
trickier to keep track of which expressions should expect an `ExpressionUsed`
node.
We therefore sidestep that complexity by only performing `ExpressionUsed`
simplification for expressions associated directly with ordinary `Code`
mappings.
Make os/windows and pal/windows default to `#![deny(unsafe_op_in_unsafe_fn)]`
This is to prevent regressions in modules that currently pass. I did also fix up a few trivial places where the module contained only one or two simple wrappers. In more complex cases we should try to ensure the `unsafe` blocks are appropriately scoped and have any appropriate safety comments.
This does not fix the windows bits of #127747 but it should help prevent regressions until that is done and also make it more obvious specifically which modules need attention.
std: `#![deny(unsafe_op_in_unsafe_fn)]` in platform-independent code
This applies the `unsafe_op_in_unsafe_fn` lint in all places in std that _do not have platform-specific cfg in their code_. For all such places, the lint remains allowed, because they need further work to address the relevant concerns. This list includes:
- `std::backtrace_rs` (internal-only)
- `std::sys` (internal-only)
- `std::os`
Notably this eliminates all "unwrapped" unsafe operations in `std::io` and `std::sync`, which will make them much more auditable in the future. Such has *also* been left for future work. While I made a few safety comments along the way on interfaces I have grown sufficiently familiar with, in most cases I had no context, nor particular confidence the unsafety was correct.
In the cases where I was able to determine the unsafety was correct without having prior context, it was obviously redundant. For example, an unsafe function calling another unsafe function that has the exact same contract, forwarding its caller's requirements just as it forwards its actual call.
Windows: Remove some unnecessary type aliases
Back in the olden days, C did not have fixed-width types so these type aliases were at least potentially useful. Nowadays, and especially in Rust, we don't need the aliases and they don't help with anything. Notably the windows bindings we use also don't bother with the aliases. And even when we have used aliases they're often only used once then forgotten about.
The only one that gives me pause is `DWORD` because it's used a fair bit. But it's still used inconsistently and we implicitly assume it's a `u32` anyway (e.g. `as` casting from an `i32`).
Fill out target-spec metadata for all targets
**What does this PR try to resolve?**
This PR completes the target-spec metadata fields for all targets. This is required for a corresponding Cargo PR which adds a check for whether a target supports building the standard library when the `-Zbuild-std=std` flag is passed ([see this issue](https://github.com/rust-lang/wg-cargo-std-aware/issues/87). This functionality in Cargo is reliant on the output of `--print=target-spec-json`.
**How should we test and review this PR?**
Check that a given target-spec metadata has been updated with:
```
$ ./x.py build library/std
$ build/host/stage1/bin/rustc --print=target-spec-json --target <target_name> -Zunstable-options
```
**Additional Information**
A few things to note:
* Where a targets 'std' or 'host tools' support is listed as '?' in the rust docs, these are left as 'None' with this PR. The corresponding changes in cargo will only reject an attempt to build std if the 'std' field is 'Some(false)'. In the case it is 'None', cargo will continue trying to build
* There's no rush for this to be merged. I understand that the format for this is not finalised yet.
* Related: #120745
These constifications were blocked on classification functions being
added. Now that those methods are available, constify them.
This brings things more in line with `f32` and `f64`.
std: removes logarithms family function edge cases handling for solaris.
Issue had been fixed over time with solaris, 11.x behaves correctly
(and we support it as minimum), illumos works correctly too.
