-
Notifications
You must be signed in to change notification settings - Fork 12.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Rollup of 3 pull requests #124194
Rollup of 3 pull requests #124194
Conversation
Currently, the matcher `//[rev-foo,rev-bar]~` does not get selected by the regex. Change the matcher to also match strings that contain a `-`.h
Implement Modified Condition/Decision Coverage This is an implementation based on llvm backend support (>= 18) by `@evodius96` and branch coverage support by `@Zalathar.` ### Major changes: * Add -Zcoverage-options=mcdc as switch. Now coverage options accept either `no-branch`, `branch`, or `mcdc`. `mcdc` also enables `branch` because it is essential to work. * Add coverage mapping for MCDCBranch and MCDCDecision. Note that MCDCParameter evolves from llvm 18 to llvm 19. The mapping in rust side mainly references to 19 and is casted to 18 types in llvm wrapper. * Add wrapper for mcdc instrinc functions from llvm. And inject associated statements to mir. * Add BcbMappingKind::Decision, I'm not sure is it proper but can't find a better way temporarily. * Let coverage-dump support parsing MCDCBranch and MCDCDecision from llvm ir. * Add simple tests to check whether mcdc works. * Same as clang, currently rustc does not generate instrument for decision with more than 6 condtions or only 1 condition due to considerations of resource. ### Implementation Details 1. To get information about conditions and decisions, `MCDCState` in `BranchInfoBuilder` is used during hir lowering to mir. For expressions with logical op we call `Builder::visit_coverage_branch_operation` to record its sub conditions, generate condition ids for them and save their spans (to construct the span of whole decision). This process mainly references to the implementation in clang and is described in comments over `MCDCState::record_conditions`. Also true marks and false marks introduced by branch coverage are used to detect where the decision evaluation ends: the next id of the condition == 0. 2. Once the `MCDCState::decision_stack` popped all recorded conditions, we can ensure that the decision is checked over and push it into `decision_spans`. We do not manually insert decision span to avoid complexity from then_else_break in nested if scopes. 3. When constructing CoverageSpans, add condition info to BcbMappingKind::Branch and decision info to BcbMappingKind::Decision. If the branch mapping has non-zero condition id it will be transformed to MCDCBranch mapping and insert `CondBitmapUpdate` statements to its evaluated blocks. While decision bcb mapping will insert `TestVectorBitmapUpdate` in all its end blocks. ### Usage ```bash echo "[build]\nprofiler=true" >> config.toml ./x build --stage 1 ./x test tests/coverage/mcdc_if.rs ``` to build the compiler and run tests. ```shell export PATH=path/to/llvm-build:$PATH rustup toolchain link mcdc build/host/stage1 cargo +mcdc rustc --bin foo -- -Cinstrument-coverage -Zcoverage-options=mcdc cd target/debug LLVM_PROFILE_FILE="foo.profraw" ./foo llvm-profdata merge -sparse foo.profraw -o foo.profdata llvm-cov show ./foo -instr-profile=foo.profdata --show-mcdc ``` to check "foo" code. ### Problems to solve For now decision mapping will insert statements to its all end blocks, which may be optimized by inserting a final block of the decision. To do this we must also trace the evaluated value at each end of the decision and join them separately. This implementation is not heavily tested so there should be some unrevealed issues. We are going to check our rust products in the next. Please let me know if you had any suggestions or comments.
…, r=oli-obk Fix capturing duplicated lifetimes via parent in `precise_captures` (`impl use<'...>`) For technical reasons related to the way that `Self` and `T::Assoc` are lowered from HIR -> `rustc_middle::ty`, an opaque may mention in its bounds both the original early-bound lifetime from the parent `impl`/`fn`, *and* the *duplicated* early-bound lifetime on the opaque. This is fine -- and has been fine since `@cjgillot` rewrote the way we handled opaque lifetime captures, and we went further to allow this behavior explicitly in rust-lang#115659. It's worthwhile to read this PR's technical section to recall how this duplication works and when it acts surprisingly. The problem here is that the check that make sure that `impl use<'a, 'b>` lists all of the opaque's captured lifetimes wasn't smart enough to consider both these captured lifetimes and the original lifetimes they're duplicated from to be equal. This PR fixes that. r? oli-obk
…, r=jieyouxu Match hyphen in multi-revision comment matchers Currently, the matcher `//[rev-foo,rev-bar]~` does not get selected by the regex. Change the matcher to include `-`.
@bors r+ rollup=never p=3 |
☀️ Test successful - checks-actions |
📌 Perf builds for each rolled up PR:
previous master: a61b14d15e In the case of a perf regression, run the following command for each PR you suspect might be the cause: |
Finished benchmarking commit (584f183): comparison URL. Overall result: ❌✅ regressions and improvements - no action needed@rustbot label: -perf-regression Instruction countThis is a highly reliable metric that was used to determine the overall result at the top of this comment.
Max RSS (memory usage)This benchmark run did not return any relevant results for this metric. CyclesThis benchmark run did not return any relevant results for this metric. Binary sizeThis benchmark run did not return any relevant results for this metric. Bootstrap: 670.71s -> 672.18s (0.22%) |
Successful merges:
precise_captures
(impl use<'...>
) #124104 (Fix capturing duplicated lifetimes via parent inprecise_captures
(impl use<'...>
))r? @ghost
@rustbot modify labels: rollup
Create a similar rollup