mirror of
https://github.com/rust-lang/rust.git
synced 2025-10-02 10:18:25 +00:00
bc4a6431eb
Convert moves of references to copies in ReferencePropagation
This is a fix for https://github.com/rust-lang/rust/issues/141101.
The root cause of this miscompile is that the SsaLocals analysis that MIR transforms use is supposed to detect locals that are only written to once, in their single assignment. But that analysis is subtly wrong; it does not consider `Operand::Move` to be a write even though the meaning ascribed to `Operand::Move` (at least as a function parameter) by Miri is that the callee may have done arbitrary writes to the caller's Local that the Operand wraps (because `Move` is pass-by-pointer). So Miri conwiders `Operand::Move` to be a write but both the MIR visitor system considers it a read, and so does SsaLocals.
I have tried fixing this by changing the `PlaceContext` that is ascribed to an `Operand::Move` to a `MutatingUseContext` but that seems to have borrow checker implications, and changing SsaLocals seems to have wide-ranging regressions in MIR optimizations.
So instead of doing those, this PR adds a new kludge to ReferencePropagation, which follows the same line of thinking as the kludge in CopyProp that solves this same problem inside that pass: a5584a8fe1/compiler/rustc_mir_transform/src/copy_prop.rs (L65-L98)
This folder contains tests for MIR optimizations.
The mir-opt test format emits MIR to extra files that you can automatically update by specifying
--bless on the command line (just like ui tests updating .stderr files).
--blessable test format
By default 32 bit and 64 bit targets use the same dump files, which can be problematic in the presence of pointers in constants or other bit width dependent things. In that case you can add
// EMIT_MIR_FOR_EACH_BIT_WIDTH
to your test, causing separate files to be generated for 32bit and 64bit systems.
Testing a particular MIR pass
If you are only testing the behavior of a particular mir-opt pass on some specific input (as is usually the case), you should add
//@ test-mir-pass: PassName
to the top of the file. This makes sure that other passes don't run which means you'll get the input you expected and your test won't break when other code changes. This also lets you test passes that are disabled by default.
Emit a diff of the mir for a specific optimization
This is what you want most often when you want to see how an optimization changes the MIR.
// EMIT_MIR $file_name_of_some_mir_dump.diff
Emit mir after a specific optimization
Use this if you are just interested in the final state after an optimization.
// EMIT_MIR $file_name_of_some_mir_dump.after.mir
Emit mir before a specific optimization
This exists mainly for completeness and is rarely useful.
// EMIT_MIR $file_name_of_some_mir_dump.before.mir
FileCheck directives
The LLVM FileCheck tool is used to verify the contents of output MIR against CHECK directives
present in the test file. This works on the runtime MIR, generated by --emit=mir, and not
on the output of a individual passes.
Use // skip-filecheck to prevent FileCheck from running.
To check MIR for function foo, start with a // CHECK-LABEL fn foo( directive.
{{regex}} syntax allows to match regex.
[[name:regex]] syntax allows to bind name to a string matching regex, and refer to it
as [[name]] in later directives, regex should be written not to match a leading space.
Use [[my_local:_.*]] to name a local, and [[my_bb:bb.*]] to name a block.
Documentation for FileCheck is available here: https://www.llvm.org/docs/CommandGuide/FileCheck.html