mirror
/
rustc-dev-guide
mirror of https://github.com/rust-lang/rustc-dev-guide.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update coverage docs (#1122)

This commit is contained in:
Rich Kadel 2021-05-13 19:19:43 -07:00 committed by GitHub
parent 1e1ffdba44
commit 412d1a75ab
1 changed files with 72 additions and 70 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

140
src/llvm-coverage-instrumentation.md
View File

@ -28,11 +28,12 @@ them), and generate various reports for analysis, for example:
<br/> <br/>
Detailed instructions and examples are documented in the Detailed instructions and examples are documented in the
[Rust Unstable Book (under _source-based-code-coverage_)][unstable-book-sbcc]. [Rust Unstable Book (under
_compiler-flags/instrument-coverage_)][unstable-book-instrument-coverage].
[llvm-instrprof-increment]: https://llvm.org/docs/LangRef.html#llvm-instrprof-increment-intrinsic [llvm-instrprof-increment]: https://llvm.org/docs/LangRef.html#llvm-instrprof-increment-intrinsic
[Coverage Map]: https://llvm.org/docs/CoverageMappingFormat.html [coverage map]: https://llvm.org/docs/CoverageMappingFormat.html
[unstable-book-sbcc]: https://doc.rust-lang.org/nightly/unstable-book/compiler-flags/source-based-code-coverage.html [unstable-book-instrument-coverage]: https://doc.rust-lang.org/nightly/unstable-book/compiler-flags/instrument-coverage.html
## Rust symbol mangling ## Rust symbol mangling
@ -82,7 +83,7 @@ a span of code ([`CodeRegion`][code-region]). It counts the number of times a
branch is executed, and also specifies the exact location of that code span in branch is executed, and also specifies the exact location of that code span in
the Rust source code. the Rust source code.
Note that many of these `Coverage` statements will *not* be converted into Note that many of these `Coverage` statements will _not_ be converted into
physical counters (or any other executable instructions) in the final binary. physical counters (or any other executable instructions) in the final binary.
Some of them will be (see `CoverageKind::`[`Counter`][counter-coverage-kind]), Some of them will be (see `CoverageKind::`[`Counter`][counter-coverage-kind]),
but other counters can be computed on the fly, when generating a coverage but other counters can be computed on the fly, when generating a coverage
@ -111,7 +112,7 @@ fn some_func(flag: bool) {
In this example, four contiguous code regions are counted while only In this example, four contiguous code regions are counted while only
incrementing two counters. incrementing two counters.
CFG analysis is used to not only determine *where* the branches are, for CFG analysis is used to not only determine _where_ the branches are, for
conditional expressions like `if`, `else`, `match`, and `loop`, but also to conditional expressions like `if`, `else`, `match`, and `loop`, but also to
determine where expressions can be used in place of physical counters. determine where expressions can be used in place of physical counters.
@ -150,40 +151,41 @@ MIR `Statement` into some backend-specific action or instruction.
match statement.kind { match statement.kind {
... ...
mir::StatementKind::Coverage(box ref coverage) => { mir::StatementKind::Coverage(box ref coverage) => {
self.codegen_coverage(&mut bx, coverage.clone()); self.codegen_coverage(&mut bx, coverage.clone(), statement.source_info.scope);
bx bx
} }
``` ```
`codegen_coverage()` handles each `CoverageKind` as follows: `codegen_coverage()` handles each `CoverageKind` as follows:
* For all `CoverageKind`s, Coverage data (counter ID, expression equation - For all `CoverageKind`s, Coverage data (counter ID, expression equation
and ID, and code regions) are passed to the backend's `Builder`, to and ID, and code regions) are passed to the backend's `Builder`, to
populate data structures that will be used to generate the crate's populate data structures that will be used to generate the crate's
"Coverage Map". (See the [`FunctionCoverage`][function-coverage] `struct`.) "Coverage Map". (See the [`FunctionCoverage`][function-coverage] `struct`.)
* For `CoverageKind::Counter`s, an instruction is injected in the backend - For `CoverageKind::Counter`s, an instruction is injected in the backend
IR to increment the physical counter, by calling the `BuilderMethod` IR to increment the physical counter, by calling the `BuilderMethod`
[`instrprof_increment()`][instrprof-increment]. [`instrprof_increment()`][instrprof-increment].
```rust ```rust
pub fn codegen_coverage(&self, bx: &mut Bx, coverage: Coverage) { pub fn codegen_coverage(&self, bx: &mut Bx, coverage: Coverage, scope: SourceScope) {
...
let instance = ... // the scoped instance (current or inlined function)
let Coverage { kind, code_region } = coverage; let Coverage { kind, code_region } = coverage;
match kind { match kind {
CoverageKind::Counter { function_source_hash, id } => { CoverageKind::Counter { function_source_hash, id } => {
if let Some(code_region) = code_region { ...
bx.add_coverage_counter(self.instance, id, code_region); bx.add_coverage_counter(instance, id, code_region);
}
... ...
bx.instrprof_increment(fn_name, hash, num_counters, index); bx.instrprof_increment(fn_name, hash, num_counters, index);
} }
CoverageKind::Expression { id, lhs, op, rhs } => { CoverageKind::Expression { id, lhs, op, rhs } => {
bx.add_coverage_counter_expression(self.instance, id, lhs, op, rhs, code_region); bx.add_coverage_counter_expression(instance, id, lhs, op, rhs, code_region);
} }
CoverageKind::Unreachable => { CoverageKind::Unreachable => {
... bx.add_coverage_unreachable(
instance,
code_region.expect(...
``` ```
_code snippet trimmed for brevity_
> The function name `instrprof_increment()` is taken from the LLVM intrinsic > The function name `instrprof_increment()` is taken from the LLVM intrinsic
call of the same name ([`llvm.instrprof.increment`][llvm-instrprof-increment]), call of the same name ([`llvm.instrprof.increment`][llvm-instrprof-increment]),
@ -221,7 +223,7 @@ properly-configured variables in LLVM IR, according to very specific
details of the [_LLVM Coverage Mapping Format_][coverage-mapping-format] details of the [_LLVM Coverage Mapping Format_][coverage-mapping-format]
(Version 4).[^llvm-and-covmap-versions] (Version 4).[^llvm-and-covmap-versions]
[^llvm-and-covmap-versions]: The Rust compiler (as of <!-- date: 2021-01 --> [^llvm-and-covmap-versions]: The Rust compiler (as of
January 2021) supports _LLVM Coverage Mapping Format_ Version 4 (the most January 2021) supports _LLVM Coverage Mapping Format_ Version 4 (the most
up-to-date version of the format, at the time of this writing) for improved up-to-date version of the format, at the time of this writing) for improved
compatibility with other LLVM-based compilers (like _Clang_), and to take compatibility with other LLVM-based compilers (like _Clang_), and to take
@ -233,13 +235,16 @@ instrument-coverage` will generate an error message.
```rust ```rust
pub fn finalize<'ll, 'tcx>(cx: &CodegenCx<'ll, 'tcx>) { pub fn finalize<'ll, 'tcx>(cx: &CodegenCx<'ll, 'tcx>) {
...
if !tcx.sess.instrument_coverage_except_unused_functions() {
add_unused_functions(cx);
}
let mut function_coverage_map = match cx.coverage_context() { let mut function_coverage_map = match cx.coverage_context() {
Some(ctx) => ctx.take_function_coverage_map(), Some(ctx) => ctx.take_function_coverage_map(),
None => return, None => return,
}; };
... ...
add_unreachable_coverage(tcx, &mut function_coverage_map);
let mut mapgen = CoverageMapGenerator::new(); let mut mapgen = CoverageMapGenerator::new();
for (instance, function_coverage) in function_coverage_map { for (instance, function_coverage) in function_coverage_map {
@ -250,56 +255,51 @@ pub fn finalize<'ll, 'tcx>(cx: &CodegenCx<'ll, 'tcx>) {
``` ```
_code snippet trimmed for brevity_ _code snippet trimmed for brevity_
One notable step, performed by `mapgen::finalize()` before processing the One notable first step performed by `mapgen::finalize()` is the call to
`Instance`s and their `FunctionCoverage`s, is the call to [`add_unused_functions()`][add-unused-functions]:
[`add_unreachable_functions()`][add-unreachable-coverage].
When finalizing the coverage map, `FunctionCoverage` only has the `CodeRegion`s and counters for When finalizing the coverage map, `FunctionCoverage` only has the `CodeRegion`s
the functions that went through codegen; such as public functions and "used" functions and counters for the functions that went through codegen; such as public
(functions referenced by other "used" or public items). Any other functions (considered unused functions and "used" functions (functions referenced by other "used" or public
or "Unreachable") were still parsed and processed through the MIR stage. items). Any other functions (considered unused) were still parsed and processed
through the MIR stage.
The set of unreachable functions is computed via the set difference of all MIR The set of unused functions is computed via the set difference of all MIR
`DefId`s (`tcx` query `mir_keys`) minus the codegenned `DefId`s `DefId`s (`tcx` query `mir_keys`) minus the codegenned `DefId`s (`tcx` query
(`tcx` query `collect_and_partition_mono_items`). `add_unreachable_functions()` `codegened_and_inlined_items`). `add_unused_functions()` computes the set of
computes the set of unreachable functions, queries the `tcx` for the unused functions, queries the `tcx` for the previously-computed `CodeRegions`,
previously-computed `CodeRegions`, for each unreachable MIR, and adds those code for each unused MIR, synthesizes an LLVM function (with no internal statements,
regions to one of the non-generic codegenned functions (non-generic avoids since it will not be called), and adds a new `FunctionCoverage`, with
potentially injecting the unreachable coverage multiple times for multiple `Unreachable` code regions.
instantiations).
[compile-codegen-unit]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/base/fn.compile_codegen_unit.html [compile-codegen-unit]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/base/fn.compile_codegen_unit.html
[coverageinfo-finalize]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/context/struct.CodegenCx.html#method.coverageinfo_finalize [coverageinfo-finalize]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/context/struct.CodegenCx.html#method.coverageinfo_finalize
[mapgen-finalize]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/coverageinfo/mapgen/fn.finalize.html [mapgen-finalize]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/coverageinfo/mapgen/fn.finalize.html
[coverage-mapping-format]: https://llvm.org/docs/CoverageMappingFormat.html [coverage-mapping-format]: https://llvm.org/docs/CoverageMappingFormat.html
[add-unreachable-coverage]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/coverageinfo/mapgen/fn.add_unreachable_coverage.html [add-unused-functions]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_llvm/coverageinfo/mapgen/fn.add_unused_functions.html
## Testing LLVM Coverage ## Testing LLVM Coverage
Coverage instrumentation in the MIR is validated by a `mir-opt` test: Coverage instrumentation in the MIR is validated by a `mir-opt` test:
[`instrument-coverage`][mir-opt-test]. [`instrument-coverage`][mir-opt-test].
More complete testing of end-to-end coverage instrumentation and reports are done More complete testing of end-to-end coverage instrumentation and reports are
in the `run-make-fulldeps` tests, with sample Rust programs (to be instrumented) done in the `run-make-fulldeps` tests, with sample Rust programs (to be
in the [`coverage`][coverage-test-samples] directory, and the actual tests and expected instrumented) in the [`coverage`][coverage-test-samples] directory, and the
results in [`coverage-reports`]. actual tests and expected results in [`coverage-reports`].
In addition to testing the final result, two intermediate results are also validated Finally, the [`coverage-llvmir`] test compares compiles a simple Rust program
to catch potential regression errors early: Minimum `CoverageSpan`s computed during with `-Z instrument-coverage` and compares the compiled program's LLVM IR to
the `InstrumentCoverage` MIR pass are saved in `mir_dump` [Spanview][spanview-debugging] expected LLVM IR instructions and structured data for a coverage-enabled
files and compared to expected results in [`coverage-spanview`]. program, including various checks for Coverage Map-related metadata and the LLVM
intrinsic calls to increment the runtime counters.
Finally, the [`coverage-llvmir`] test compares compiles a simple Rust program with
`-Z instrument-coverage` and compares the compiled program's LLVM IR to expected
LLVM IR instructions and structured data for a coverage-enabled program, including
various checks for Coverage Map-related metadata and the LLVM intrinsic calls to
increment the runtime counters.
Expected results for both the `mir-opt` tests and the `coverage*` tests under Expected results for both the `mir-opt` tests and the `coverage*` tests under
`run-make-fulldeps` can be refreshed by running: `run-make-fulldeps` can be refreshed by running:
```shell ```shell
$ ./x.py test src/test/<test-type> --blessed $ ./x.py test mir-opt --blessed
$ ./x.py test src/test/run-make-fulldeps/coverage --blessed
``` ```
[mir-opt-test]: https://github.com/rust-lang/rust/blob/master/src/test/mir-opt/instrument_coverage.rs [mir-opt-test]: https://github.com/rust-lang/rust/blob/master/src/test/mir-opt/instrument_coverage.rs
@ -396,16 +396,18 @@ contrast with the [`SimplifyCfg`][simplify-cfg] MIR pass, this step does
not alter the MIR itself, because the `CoverageGraph` aggressively simplifies not alter the MIR itself, because the `CoverageGraph` aggressively simplifies
the CFG, and ignores nodes that are not relevant to coverage. For example: the CFG, and ignores nodes that are not relevant to coverage. For example:
* The BCB CFG ignores (excludes) branches considered not relevant - The BCB CFG ignores (excludes) branches considered not relevant
to the current coverage solution. It excludes unwind-related code[^78544] to the current coverage solution. It excludes unwind-related code[^78544]
that is injected by the Rust compiler but has no physical source that is injected by the Rust compiler but has no physical source
code to count, which allows a `Call`-terminated BasicBlock code to count, which allows a `Call`-terminated BasicBlock
to be merged with its successor, within a single BCB. to be merged with its successor, within a single BCB.
* A `Goto`-terminated `BasicBlock` can be merged with its successor - A `Goto`-terminated `BasicBlock` can be merged with its successor
***as long as*** it has the only incoming edge to the successor `BasicBlock`. **_as long as_** it has the only incoming edge to the successor
* Some BasicBlock terminators support Rust-specific concerns--like borrow-checking--that are `BasicBlock`.
not relevant to coverage analysis. `FalseUnwind`, for example, can be treated the same as - Some BasicBlock terminators support Rust-specific concerns--like
a `Goto` (potentially merged with its successor into the same BCB). borrow-checking--that are not relevant to coverage analysis. `FalseUnwind`,
for example, can be treated the same as a `Goto` (potentially merged with
its successor into the same BCB).
[^78544]: (Note, however, that Issue [#78544][rust-lang/rust#78544] considers [^78544]: (Note, however, that Issue [#78544][rust-lang/rust#78544] considers
providing future support for coverage of programs that intentionally providing future support for coverage of programs that intentionally
@ -448,24 +450,24 @@ directional edges (the arrows) leading from each node to its `successors()`.
The nodes contain information in sections: The nodes contain information in sections:
1. The gray header has a label showing the BCB ID (or _index_ for looking up 1. The gray header has a label showing the BCB ID (or _index_ for looking up
its `BasicCoverageBlockData`). its `BasicCoverageBlockData`).
2. The first content section shows the assigned `Counter` or `Expression` for 2. The first content section shows the assigned `Counter` or `Expression` for
each contiguous section of code. (There may be more than one `Expression` each contiguous section of code. (There may be more than one `Expression`
incremented by the same `Counter` for discontiguous sections of code representing incremented by the same `Counter` for discontiguous sections of code
the same sequential actions.) Note the code is represented by the line and representing the same sequential actions.) Note the code is represented by
column ranges (for example: `52:28-52:33`, representing the original source the line and column ranges (for example: `52:28-52:33`, representing the
line 52, for columns 28-33). These are followed by the MIR `Statement` or original source line 52, for columns 28-33). These are followed by the MIR
`Terminator` represented by that source range. (How these coverage regions `Statement` or `Terminator` represented by that source range. (How these
are determined is discussed in the following section.) coverage regions are determined is discussed in the following section.)
3. The final section(s) show the MIR `BasicBlock`s (by ID/index and its 3. The final section(s) show the MIR `BasicBlock`s (by ID/index and its
`TerminatorKind`) contained in this BCB. The last BCB is separated out because `TerminatorKind`) contained in this BCB. The last BCB is separated out
its `successors()` determine the edges leading out of the BCB, and into because its `successors()` determine the edges leading out of the BCB, and
the `leading_bb()` (first `BasicBlock`) of each successor BCB. into the `leading_bb()` (first `BasicBlock`) of each successor BCB.
Note, to find the `BasicCoverageBlock` from a final BCB `Terminator`'s Note, to find the `BasicCoverageBlock` from a final BCB `Terminator`'s
successor `BasicBlock`, there is an index and helper successor `BasicBlock`, there is an index and helper
function--[`bcb_from_bb()`][bcb-from-bb]--to look up a `BasicCoverageBlock` from _any_ function--[`bcb_from_bb()`][bcb-from-bb]--to look up a `BasicCoverageBlock` from
contained `BasicBlock`. *any* contained `BasicBlock`.
[directed-graph]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_data_structures/graph/trait.DirectedGraph.html [directed-graph]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_data_structures/graph/trait.DirectedGraph.html
[graph-traits]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_data_structures/graph/index.html#traits [graph-traits]: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_data_structures/graph/index.html#traits