miri: make vtable addresses not globally unique
Miri currently gives vtables a unique global address. That's not actually matching reality though. So this PR enables Miri to generate different addresses for the same type-trait pair.
To avoid generating an unbounded number of `AllocId` (and consuming unbounded amounts of memory), we use the "salt" technique that we also already use for giving constants non-unique addresses: the cache is keyed on a "salt" value n top of the actually relevant key, and Miri picks a random salt (currently in the range `0..16`) each time it needs to choose an `AllocId` for one of these globals -- that means we'll get up to 16 different addresses for each vtable. The salt scheme is integrated into the global allocation deduplication logic in `tcx`, and also used for functions and string literals. (So this also fixes the problem that casting the same function to a fn ptr over and over will consume unbounded memory.)
r? `@saethlin`
Fixes https://github.com/rust-lang/miri/issues/3737
Line numbers of multiply-inlined functions were fixed in #114643 by using a
single DISubprogram. That, however, triggered assertions because parameters
weren't deduplicated. The "solution" to that in #115417 was to insert a
DILexicalScope below the DISubprogram and parent all of the parameters to that
scope. That fixed the assertion, but debuggers (including gdb and lldb) don't
recognize variables that are not parented to the subprogram itself as parameters,
even if they are emitted with DW_TAG_formal_parameter.
Consider the program:
use std::env;
fn square(n: i32) -> i32 {
n * n
}
fn square_no_inline(n: i32) -> i32 {
n * n
}
fn main() {
let x = square(env::vars().count() as i32);
let y = square_no_inline(env::vars().count() as i32);
println!("{x} == {y}");
}
When making a release build with debug=2 and rustc 1.82.0-nightly (8b3870784 2024-08-07)
(gdb) r
Starting program: /ephemeral/tmp/target/release/tmp
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Breakpoint 1, tmp::square () at src/main.rs:5
5 n * n
(gdb) info args
No arguments.
(gdb) info locals
n = 31
(gdb) c
Continuing.
Breakpoint 2, tmp::square_no_inline (n=31) at src/main.rs:10
10 n * n
(gdb) info args
n = 31
(gdb) info locals
No locals.
This issue is particularly annoying because it removes arguments from stack traces.
The DWARF for the inlined function looks like this:
< 2><0x00002132 GOFF=0x00002132> DW_TAG_subprogram
DW_AT_linkage_name _ZN3tmp6square17hc507052ff3d2a488E
DW_AT_name square
DW_AT_decl_file 0x0000000f /ephemeral/tmp/src/main.rs
DW_AT_decl_line 0x00000004
DW_AT_type 0x00001a56<.debug_info+0x00001a56>
DW_AT_inline DW_INL_inlined
< 3><0x00002142 GOFF=0x00002142> DW_TAG_lexical_block
< 4><0x00002143 GOFF=0x00002143> DW_TAG_formal_parameter
DW_AT_name n
DW_AT_decl_file 0x0000000f /ephemeral/tmp/src/main.rs
DW_AT_decl_line 0x00000004
DW_AT_type 0x00001a56<.debug_info+0x00001a56>
< 4><0x0000214e GOFF=0x0000214e> DW_TAG_null
< 3><0x0000214f GOFF=0x0000214f> DW_TAG_null
That DW_TAG_lexical_block inhibits every debugger I've tested from recognizing
'n' as a parameter.
This patch removes the additional lexical scope. Parameters can be easily
deduplicated by a tuple of their scope and the argument index, at the trivial
cost of taking a Hash + Eq bound on DIScope.
bootstrap: clear miri ui-test deps when miri sysroot gets rebuilt
Second attempt after https://github.com/rust-lang/rust/pull/128683: seems like it's not the compiler changing that we care about, but the sysroot changing.
I did some local testing with sysroot rebuilds and it works fine for at least those cases I checked.
r? ``@onur-ozkan``
Store `do_not_recommend`-ness in impl header
Alternative to #128674
It's less flexible, but also less invasive. Hopefully it's also performant. I'd recommend we think separately about the design for how to gate arbitrary diagnostic attributes moving forward.
Normalize struct tail properly for `dyn` ptr-to-ptr casting in new solver
Realized that the new solver didn't handle ptr-to-ptr casting correctly.
r? lcnr
Built on #128694
Rollup of 10 pull requests
Successful merges:
- #128149 (nontemporal_store: make sure that the intrinsic is truly just a hint)
- #128394 (Unify run button display with "copy code" button and with mdbook buttons)
- #128537 (const vector passed through to codegen)
- #128632 (std: do not overwrite style in `get_backtrace_style`)
- #128878 (Slightly refactor `Flags` in bootstrap)
- #128886 (Get rid of some `#[allow(rustc::untranslatable_diagnostic)]`)
- #128929 (Fix codegen-units tests that were disabled 8 years ago)
- #128937 (Fix warnings in rmake tests on `x86_64-unknown-linux-gnu`)
- #128978 (Use `assert_matches` around the compiler more)
- #128994 (Fix bug in `Parser::look_ahead`.)
r? `@ghost`
`@rustbot` modify labels: rollup
Fix bug in `Parser::look_ahead`.
The special case was failing to handle invisible delimiters on one path.
Fixes (but doesn't close until beta backported) #128895.
r? `@davidtwco`
Fix warnings in rmake tests on `x86_64-unknown-linux-gnu`
r? `@jieyouxu`
This PR fixes some warnings I saw in rmake tests. I didn't deny more warnings in this PR until `@jieyouxu` gives their opinion, but maybe we should actually deny all warnings in `rmake.rs` files?
I've also only looked at non-ignored tests on `x86_64-unknown-linux-gnu`, and denying warnings would require a try build for all targets 😓.
Fix codegen-units tests that were disabled 8 years ago
I don't know if any of these tests still have value. They were disabled by https://github.com/rust-lang/rust/pull/33890, and we've survived without them for a while. But considering how small this test suite is, maybe it's worth having them.
I also had to add some normalization to the codegen-units tests output. I think the fact that I had to add some underscores how poor our test coverage is.
Slightly refactor `Flags` in bootstrap
The next step for https://github.com/rust-lang/rust/issues/126819 is to track commands executed inside `Config::parse`. This is quite challenging, because (tracked) command execution needs to access some state that is stored inside `Config`, which creates a sort of a chicken-and-egg problem.
I would like to first untangle `Config::parse` a little bit, which is what this PR starts with.
Tracking issue: https://github.com/rust-lang/rust/issues/126819
r? `@onur-ozkan`
std: do not overwrite style in `get_backtrace_style`
If another thread calls `set_backtrace_style` while a `get_backtrace_style` is reading the environment variables, `get_backtrace_style` will overwrite the value. Use an atomic CAS to avoid this.
const vector passed through to codegen
This allows constant vectors using a repr(simd) type to be propagated
through to the backend by reusing the functionality used to do a similar
thing for the simd_shuffle intrinsic
#118209
r? RalfJung
nontemporal_store: make sure that the intrinsic is truly just a hint
The `!nontemporal` flag for stores in LLVM *sounds* like it is just a hint, but actually, it is not -- at least on x86, non-temporal stores need very special treatment by the programmer or else the Rust memory model breaks down. LLVM still treats these stores as-if they were normal stores for optimizations, which is [highly dubious](https://github.com/llvm/llvm-project/issues/64521). Let's avoid all that dubiousness by making our own non-temporal stores be truly just a hint, which is possible on some targets (e.g. ARM). On all other targets, non-temporal stores become regular stores.
~~Blocked on https://github.com/rust-lang/stdarch/pull/1541 propagating to the rustc repo, to make sure the `_mm_stream` intrinsics are unaffected by this change.~~
Fixes https://github.com/rust-lang/rust/issues/114582
Cc `@Amanieu` `@workingjubilee`
fix: Missing non-exhaustive let diagnostics inside async or unsafe block
The reason that this test doesn't have a pointer deref case is because the following code;
```rust
fn test(ptr: *const Result<i32, !>) {
unsafe {
let Ok(_x) = *ptr;
}
}
```
is getting a block with no stmts but tail one in here(thus, no diagnostic error),
0daeb5c0b0/crates/hir-ty/src/diagnostics/expr.rs (L256-L257)
while the following is getting a block with a single stmt without tail 🤔
```rust
fn test(x: Result<i32, &'static !>) {
let Ok(_y) = x;
}
```
I'll make a more deep inspection and file this as a new issue
_Originally posted by `@ShoyuVanilla` in https://github.com/rust-lang/rust-analyzer/pull/17853#discussion_r1712993585_
