rustdoc: change doctests locating rustc binary
We previously used the "naive" approach of replacing the `current_exe()`'s file name with rustc, but now load from the sysroot by default (`$sysroot/bin/rustc`). The functionality of locating the sysroot overlaps/is the same as the functionality used by codegen backend loading; this ensures that any failure cases we've introduced are not exceeding those, and that improvements to finding the sysroot for loading codegen backends likewise enhance rustdoc.
The second commit adds an unstable `--test-builder` flag to rustdoc, and is largely separate (I can split into separate PR, but it's a simple and related change). This is largely intended for "advanced" users at this point (I'm not sure if we'll ever stabilize it); it permits use of a different rustc binary for rustdoc compilation of doctests than the rustdoc binary used when loading. Note, that this may not be what you want as the parsers and such differ (and rustdoc uses its own libsyntax, etc.). However, I've been told that running doctests in miri may be assisted by this change, so I've implemented it; I'll file a tracking issue for it if there's interest in it (and we land this PR).
Avoid more `Symbol`-to-string operations
These commits avoid various `Symbol`-to-string conversions, by doing more operations directly on `Symbol`s. This requires adding a few more static `Symbol`s to the binary.
r? @petrochenkov
Add i686-unknown-uefi target
This adds a new rustc target-configuration called 'i686-unknown_uefi'.
This is similar to existing x86_64-unknown_uefi target.
The i686-unknown-uefi target can be used to build Intel Architecture
32bit UEFI application. The ABI defined in UEFI environment (aka IA32)
is similar to cdecl.
We choose i686-unknown-uefi-gnu instead of i686-unknown-uefi to avoid
the intrinsics generated by LLVM. The detail of root-cause and solution
analysis is added as comment in the code.
For x86_64-unknown-uefi, we cannot use -gnu, because the ABI between
MSVC and GNU is totally different, and UEFI chooses ABI similar to MSVC.
For i686-unknown-uefi, the UEFI chooses cdecl ABI, which is same as
MSVC and GNU. According to LLVM code, the only differences between MSVC
and GNU are fmodf(f32), longjmp() and TLS, which have no impact to UEFI.
As such, using i686-unknown-uefi-gnu is the simplest way to pass the build.
Adding the undefined symbols, such as _aulldiv() to rust compiler-builtins
is out of scope. But it may be considered later.
The scope of this patch is limited to support target-configuration.
No standard library support is added in this patch. Such work can be
done in future enhancement.
Cc: Josh Triplett <josh.triplett@intel.com>
Reviewed-by: Josh Triplett <josh.triplett@intel.com>
It's a large type -- 176 bytes on 64-bit. And it's passed around and
returned from a lot of functions, including within PResult.
This commit boxes it, which reduces memory traffic. In particular,
`PResult` shrinks to 16 bytes in the best case; this reduces instruction
counts by up to 2% on various workloads.
check_match: refactor + improve non-exhaustive diagnostics for default binding modes
Refactor `check_match` a bit with more code-reuse and improve the diagnostics for a non-exhaustive pattern match by peeling off any references from the scrutinee type so that the "defined here" label is added in more cases. For example:
```rust
error[E0004]: non-exhaustive patterns: `&mut &B` not covered
--> foo.rs:4:11
|
1 | enum E { A, B }
| ---------------
| | |
| | not covered
| `E` defined here
...
4 | match x {
| ^ pattern `&mut &B` not covered
|
= help: ensure that all possible cases are being handled, possibly by adding wildcards or more match arms
```
Moreover, wrt. "defined here", we give irrefutable pattern matching (i.e. in `let`, `for`, and `fn` parameters) a more consistent treatment in line with `match`.
r? @estebank
std: Add a `backtrace` module
This commit adds a `backtrace` module to the standard library, as
designed in [RFC 2504]. The `Backtrace` type is intentionally very
conservative, effectively only allowing capturing it and printing it.
Additionally this commit also adds a `backtrace` method to the `Error`
trait which defaults to returning `None`, as specified in [RFC 2504].
More information about the design here can be found in [RFC 2504] and in
the [tracking issue].
Implementation-wise this is all based on the `backtrace` crate and very
closely mirrors the `backtrace::Backtrace` type on crates.io. Otherwise
it's pretty standard in how it handles everything internally.
[RFC 2504]: https://github.com/rust-lang/rfcs/blob/master/text/2504-fix-error.md
[tracking issue]: https://github.com/rust-lang/rust/issues/53487
cc #53487
Rollup of 6 pull requests
Successful merges:
- #64060 (Improve hygiene of `alloc::format!`)
- #64072 (Replace file_stem by file_name in rustdoc markdown)
- #64129 (vxWorks: set DEFAULT_MIN_STACK_SIZE to 256K and use min_stack to pass initial stack size to rtpSpawn)
- #64188 (rustc: Allow the cdylib crate type with wasm32-wasi)
- #64326 (Fixed documentation within c_str::from_ptr)
- #64349 (documentation for AtomicPtr CAS operations)
Failed merges:
r? @ghost
documentation for AtomicPtr CAS operations
The examples in the documentation for AtomicPtr CAS operations only show code that does *not* perform the CAS operation. I suggest to change them so that they actually do exchange the AtomicPtr's value.
rustc: Allow the cdylib crate type with wasm32-wasi
The wasm32-wasi target respects configuration around `crt-static` in
general, but is defaulted to being static. This interacted badly with
code which validated the `cdylib` crate type for `wasm32-wasi`,
erroneously saying that the `cdylib` crate type wasn't supported on
`wasm32-wasi` by default. This commit sets the appropriate flag in
`wasm32_wasi`'s target specification to indicate that the `cdylib` crate
type is supported regardless of `crt-static`
Closes#64187
vxWorks: set DEFAULT_MIN_STACK_SIZE to 256K and use min_stack to pass initial stack size to rtpSpawn
vxWorks: set DEFAULT_MIN_STACK_SIZE to 256K and use min_stack to pass initial stack size to rtpSpawn
r? @alexcrichton
cc @n-salim
Replace file_stem by file_name in rustdoc markdown
Before this PR, a file name like `some.file.md` will be output to a file named `some.html` with is not correct because the expected output file must be `some.file.html`
This ensures that the failure cases for finding the codegen backend and
for finding the rustc binary are essentially the same, and since we
almost always will load the codegen backend, this is essentially meaning
that the rustc change is not a regression.
rustc_mir: buffer -Zdump-mir output instead of pestering the kernel constantly.
This brings `mir-opt` tests from `197s` (over 3 minutes!) to `2.85s`, on my build server.
That's a nice speedup of about `69x` and so it definitely fixes#58485, for me.
It's such a beginner mistake I feel like maybe `clippy` should lint against it?
(cc @Manishearth @oli-obk)
lldb: avoid mixing "Hit breakpoint" message with other output.
This is to get `src/test/debuginfo/lexical-scopes-in-block-expression.rs` working.
It used to work like a week ago, and the main change that happened was I switched from Python 2 to Python 3 (I don't remember why, but I did get rid of the build dir entirely, and it fixed something else).
The error was:
```
error: line not found in debugger output: [...]$27 = 10
```
Relevant part of the output:
```
print val
(long) $26 = 15
print ten
(long) $27 = 10 Hit breakpoint 15.1: where = a`lexical_scopes_in_block_expression::main::hcdd5c3caa9166e73 + 1223 at lexical-scopes-in-block-expression.rs:504:4, address = 0x00005555555556e7, resolved, hit count = 1
Hit breakpoint 16.1: where = a`lexical_scopes_in_block_expression::main::hcdd5c3caa9166e73 + 631 at lexical-scopes-in-block-expression.rs:510:8, address = 0x0000555555555497, resolved, hit count = 1
```
There are most `print` commands and their outputs before, and more `Hit breakpoint` messages afterwards, so I assume what happens is the `Hit breakpoint` messages should be interleaved but somehow they ended up being buffered after all of the other output.
As a stopgap measure I'm adding a newline before each `Hit breakpoint` so they don't end up on the same line as the last `print` output (which breaks our pattern-matching).
r? @michaelwoerister
test/c-variadic: Fix patterns on powerpc64
On architectures such as powerpc64 that use extend_integer_width_to in
their C ABI processing, integer parameters shorter than the native
register width will be annotated with the ArgAttribute::SExt or
ArgAttribute::ZExt attribute, and that attribute will be included in the
generated LLVM IR.
In this test, all relevant parameters are `i32`, which will get the
`signext` annotation on the relevant 64-bit architectures. Match both
the annotated and non-annotated case, but enforce that the annotation is
applied consistently.
Make `abs`, `wrapping_abs`, `overflowing_abs` const functions
This makes `abs`, `wrapping_abs` and `overflowing_abs` const functions like #58044 makes `wrapping_neg` and `overflowing_neg` const functions.
`abs` is made const by returning `(self ^ -1) - -1` = `!self + 1` = `-self` for negative numbers and `(self ^ 0) - 0` = `self` for non-negative numbers. The subexpression `self >> ($BITS - 1)` evaluates to `-1` for negative numbers and `0` otherwise. The subtraction overflows when `self` is `min_value()`, as we would be subtracting `max_value() - -1`; this is when `abs` should overflow.
`wrapping_abs` and `overflowing_abs` make use of `wrapping_sub` and `overflowing_sub` instead of the subtraction operator.
