Move MIR towards a single kind of local
This PR modifies MIR to handle function arguments (`Arg`), user-defined variable bindings (`Var`), compiler-generated temporaries (`Tmp`), as well as the return value pointer equally. All of them are replaced with a single `Local` type, a few functions for iterating over different kinds of locals, and a way to get the kind of local we're dealing with (mainly used in the constant qualification/propagation passes).
~~I haven't managed to fix one remaining issue: A `StorageDead` not getting emitted for a variable (see the `TODO` in the test). If that's fixed, this is basically good to go.~~ Found the issue (an off-by-one error), fix incoming.
r? @eddyb for changes to constant qualification and propagation I'm not quite sure about
Allow more non-inline modules in blocks
Currently, non-inline modules without a `#[path]` attribute are not allowed in blocks.
This PR allows non-inline modules that have an ancestor module with a `#[path]` attribute, provided there is not a nearer ancestor block.
For example,
```rust
fn main() {
#[path = "..."] mod foo {
mod bar; //< allowed by this PR
fn f() {
mod bar; //< still an error
}
}
}
```
Fixes#36772.
r? @nikomatsakis
Allow supplying an error destination via the compiler driver
Allows replacing stderr with a buffer from the client.
Also, some refactoring around run_compiler.
Remove CString drop test.
The test relies on the undefined behavior, and so may fail in some
circumstances. This can be worked around by stubbing a memory allocator
in the test, but it is a bit of work, and LLVM could still theoretically
eliminate the write of the zero byte in release mode (which is
intended).
So let's just remove the test and mark the function as inline. It
shouldn't be optimized away when inlined into the debug build of user's
code.
Supersedes #36607
r? @alexcrichton
remove ExactSizeIterator from RangeInclusive<{u,i}{32,size}>
Fixes#36386.
This is a [breaking-change] for nightly users of `#![feature(inclusive_range_syntax)]` and/or `#![feature(inclusive_range)]`.
rustc: Tweak expansion order of custom derive
This commit alters the expansion order of custom macros-1.1 style `#[derive]`
modes. Instead of left-to-right the expansion now happens in three categories,
each of which is internally left-to-right:
* Old-style custom derive (`#[derive_Foo]`) is expanded
* New-style custom derive (macros 1.1) is expanded
* Built in derive modes are expanded
This gives built in derive modes maximal knowledge about the struct that's being
expanded and also avoids pesky issues like exposing `#[structural_match]` or
`#[rustc_copy_clone_marker]`.
cc #35900
rustc: Use a special filename for macros 1.1
This "special filename" is surrounded by `<>` to ensure that
`FileMap::is_real_file` returns `false`. This way the "files" parsed here aren't
emitted as dep info `.d` files and don't confuse Cargo about non-existent files.
Closes#36625
This commit alters the expansion order of custom macros-1.1 style `#[derive]`
modes. Instead of left-to-right the expansion now happens in three categories,
each of which is internally left-to-right:
* Old-style custom derive (`#[derive_Foo]`) is expanded
* New-style custom derive (macros 1.1) is expanded
* Built in derive modes are expanded
This gives built in derive modes maximal knowledge about the struct that's being
expanded and also avoids pesky issues like exposing `#[structural_match]` or
`#[rustc_copy_clone_marker]`.
cc #35900
This "special filename" is surrounded by `<>` to ensure that
`FileMap::is_real_file` returns `false`. This way the "files" parsed here aren't
emitted as dep info `.d` files and don't confuse Cargo about non-existent files.
Closes#36625
Haiku: Initial work at OS support
These changes should be non-invasive to non-Haiku platforms. These patches were hand reworked from Neil's original Rust 1.9.0 patches. I've done some style cleanup and design updates along the way.
There are a few small additional patches to libc, rust-installer and compiler-rt that will be submitted once this one is accepted.
Haiku can be compiled on Linux, and a full gcc cross-compiler with a Haiku target is available, which means bootstrapping should be fairly easy. The patches here have already successfully bootstrapped under our haiku x86_gcc2 architecture. http://rust-on-haiku.com/wiki/PortingRust
I'll be focusing on our more modern gcc5 x86 and x86 architectures for now.
As for support, we're not seeking official support for now. We understand Haiku isn't a top-tier OS choice, however having these patches upstream greatly reduces the amount of patchwork we have to do. Mesa has Haiku code upstream, and we submit patches to keep it going. Mesa doesn't test on Haiku and we're ok with that :-)
rustc: implement -C link-arg
this flag lets you pass a _single_ argument to the linker but can be
used _repeatedly_. For example, instead of using:
```
rustc -C link-args='-l bar' (..)
```
you could write
```
rustc -C link-arg='-l' -C link-arg='bar' (..)
```
This new flag can be used with RUSTFLAGS where `-C link-args` has
problems with "nested" spaces:
```
RUSTFLAGS='-C link-args="-Tlayout.ld -nostartfiles"'
```
This passes three arguments to rustc: `-C` `link-args="-Tlayout.ld` and
`-nostartfiles"` to `rustc`. That's not what we meant. But this does
what we want:
```
RUSTFLAGS='-C link-arg=-Tlayout.ld -C link-arg=-nostartfiles`
```
cc rust-lang/rfcs#1509
r? @alexcrichton
cc @Zoxc
This needs a test. Any suggestion?
Change error message for intrinsic signature.
Makes it so the signature of the intrinsic in the user's code is
"found," while the signature that rustc knows about is "expected."
Before this patch, the code
```
extern "platform-intrinsic" {
fn x86_mm_movemask_ps() -> i32;
}
```
would give the error
```
error[E0444]: platform-specific intrinsic has invalid number of arguments: found 1, expected 0
--> test.rs:4:5
|
4 | fn x86_mm_movemask_ps() -> i32;
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
error: aborting due to previous error
```
After this patch, it says "found 0, expected 1".
The test relies on the undefined behavior, and so may fail in some
circumstances. This can be worked around by stubbing a memory allocator
in the test, but it is a bit of work, and LLVM could still theoretically
eliminate the write of the zero byte in release mode (which is
intended).
So let's just remove the test and mark the function as inline. It
shouldn't be optimized away when inlined into the debug build of user's
code.
