Skip to content

Commit 8075b28

Browse files
richkadelrichkadel
committed
Renamed compiler-flags file to name of compiler-flag: instrument-coverage
And redirect users from the old file name.
1 parent f58a362 commit 8075b28

File tree

2 files changed

+349
-344
lines changed

2 files changed

+349
-344
lines changed

‎src/doc/unstable-book/src/compiler-flags/instrument-coverage.md

+346
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,346 @@
1+
# `instrument-coverage`
2+
3+
The tracking issue for this feature is: [#79121].
4+
5+
[#79121]: https://github.com/rust-lang/rust/issues/79121
6+
7+
---
8+
9+
## Introduction
10+
11+
The Rust compiler includes two code coverage implementations:
12+
13+
- A GCC-compatible, gcov-based coverage implementation, enabled with `-Z profile`, which derives coverage data based on DebugInfo.
14+
- A source-based code coverage implementation, enabled with `-Z instrument-coverage`, which uses LLVM's native, efficient coverage instrumentation to generate very precise coverage data.
15+
16+
This document describes how to enable and use the LLVM instrumentation-based coverage, via the `-Z instrument-coverage` compiler flag.
17+
18+
## How it works
19+
20+
When `-Z instrument-coverage` is enabled, the Rust compiler enhances rust-based libraries and binaries by:
21+
22+
- Automatically injecting calls to an LLVM intrinsic ([`llvm.instrprof.increment`]), at functions and branches in compiled code, to increment counters when conditional sections of code are executed.
23+
- Embedding additional information in the data section of each library and binary (using the [LLVM Code Coverage Mapping Format] _Version 4_, supported _only_ in LLVM 11 and up), to define the code regions (start and end positions in the source code) being counted.
24+
25+
When running a coverage-instrumented program, the counter values are written to a `profraw` file at program termination. LLVM bundles tools that read the counter results, combine those results with the coverage map (embedded in the program binary), and generate coverage reports in multiple formats.
26+
27+
[`llvm.instrprof.increment`]: https://llvm.org/docs/LangRef.html#llvm-instrprof-increment-intrinsic
28+
[llvm code coverage mapping format]: https://llvm.org/docs/CoverageMappingFormat.html
29+
30+
> **Note**: `-Z instrument-coverage` also automatically enables `-Z symbol-mangling-version=v0` (tracking issue [#60705]). The `v0` symbol mangler is strongly recommended, but be aware that this demangler is also experimental. The `v0` demangler can be overridden by explicitly adding `-Z symbol-mangling-version=legacy`.
31+
32+
[#60705]: https://github.com/rust-lang/rust/issues/60705
33+
34+
## Enable coverage profiling in the Rust compiler
35+
36+
Rust's source-based code coverage requires the Rust "profiler runtime". Without it, compiling with `-Z instrument-coverage` generates an error that the profiler runtime is missing.
37+
38+
The Rust `nightly` distribution channel includes the profiler runtime, by default.
39+
40+
> **Important**: If you are building the Rust compiler from the source distribution, the profiler runtime is _not_ enabled in the default `config.toml.example`. Edit your `config.toml` file and ensure the `profiler` feature is set it to `true` (either under the `[build]` section, or under the settings for an individual `[target.<triple>]`):
41+
>
42+
> ```toml
43+
> # Build the profiler runtime (required when compiling with options that depend
44+
> # on this runtime, such as `-C profile-generate` or `-Z instrument-coverage`).
45+
> profiler = true
46+
> ```
47+
48+
### Building the demangler
49+
50+
LLVM coverage reporting tools generate results that can include function names and other symbol references, and the raw coverage results report symbols using the compiler's "mangled" version of the symbol names, which can be difficult to interpret. To work around this issue, LLVM coverage tools also support a user-specified symbol name demangler.
51+
52+
One option for a Rust demangler is [`rustfilt`], which can be installed with:
53+
54+
```shell
55+
cargo install rustfilt
56+
```
57+
58+
Another option, if you are building from the Rust compiler source distribution, is to use the `rust-demangler` tool included in the Rust source distribution, which can be built with:
59+
60+
```shell
61+
$ ./x.py build rust-demangler
62+
```
63+
64+
[`rustfilt`]: https://crates.io/crates/rustfilt
65+
66+
## Compiling with coverage enabled
67+
68+
Set the `-Z instrument-coverage` compiler flag in order to enable LLVM source-based code coverage profiling.
69+
70+
The default option generates coverage for all functions, including unused (never called) functions and generics. The compiler flag supports an optional value to tailor this behavior. (See [`-Z instrument-coverage=<options>`](#-z-instrument-coverageoptions), below.)
71+
72+
With `cargo`, you can instrument your program binary _and_ dependencies at the same time.
73+
74+
For example (if your project's Cargo.toml builds a binary by default):
75+
76+
```shell
77+
$ cd your-project
78+
$ cargo clean
79+
$ RUSTFLAGS="-Z instrument-coverage" cargo build
80+
```
81+
82+
If `cargo` is not configured to use your `profiler`-enabled version of `rustc`, set the path explicitly via the `RUSTC` environment variable. Here is another example, using a `stage1` build of `rustc` to compile an `example` binary (from the [`json5format`] crate):
83+
84+
```shell
85+
$ RUSTC=$HOME/rust/build/x86_64-unknown-linux-gnu/stage1/bin/rustc \
86+
RUSTFLAGS="-Z instrument-coverage" \
87+
cargo build --example formatjson5
88+
```
89+
90+
> **Note**: that some compiler options, combined with `-Z instrument-coverage`, can produce LLVM IR and/or linked binaries that are incompatible with LLVM coverage maps. For example, coverage requires references to actual functions in LLVM IR. If any covered function is optimized out, the coverage tools may not be able to process the coverage results. If you need to pass additional options, with coverage enabled, test them early, to confirm you will get the coverage results you expect.
91+
92+
## Running the instrumented binary to generate raw coverage profiling data
93+
94+
In the previous example, `cargo` generated the coverage-instrumented binary `formatjson5`:
95+
96+
```shell
97+
$ echo "{some: 'thing'}" | target/debug/examples/formatjson5 -
98+
```
99+
100+
```json5
101+
{
102+
some: "thing",
103+
}
104+
```
105+
106+
After running this program, a new file, `default.profraw`, should be in the current working directory. It's often preferable to set a specific file name or path. You can change the output file using the environment variable `LLVM_PROFILE_FILE`:
107+
108+
```shell
109+
$ echo "{some: 'thing'}" \
110+
| LLVM_PROFILE_FILE="formatjson5.profraw" target/debug/examples/formatjson5 -
111+
...
112+
$ ls formatjson5.profraw
113+
formatjson5.profraw
114+
```
115+
116+
If `LLVM_PROFILE_FILE` contains a path to a non-existent directory, the missing directory structure will be created. Additionally, the following special pattern strings are rewritten:
117+
118+
- `%p` - The process ID.
119+
- `%h` - The hostname of the machine running the program.
120+
- `%t` - The value of the TMPDIR environment variable.
121+
- `%Nm` - the instrumented binary’s signature: The runtime creates a pool of N raw profiles, used for on-line profile merging. The runtime takes care of selecting a raw profile from the pool, locking it, and updating it before the program exits. `N` must be between `1` and `9`, and defaults to `1` if omitted (with simply `%m`).
122+
- `%c` - Does not add anything to the filename, but enables a mode (on some platforms, including Darwin) in which profile counter updates are continuously synced to a file. This means that if the instrumented program crashes, or is killed by a signal, perfect coverage information can still be recovered.
123+
124+
## Installing LLVM coverage tools
125+
126+
LLVM's supplies two tools—`llvm-profdata` and `llvm-cov`—that process coverage data and generate reports. There are several ways to find and/or install these tools, but note that the coverage mapping data generated by the Rust compiler requires LLVM version 11 or higher. (`llvm-cov --version` typically shows the tool's LLVM version number.):
127+
128+
- The LLVM tools may be installed (or installable) directly to your OS (such as via `apt-get`, for Linux).
129+
- If you are building the Rust compiler from source, you can optionally use the bundled LLVM tools, built from source. Those tool binaries can typically be found in your build platform directory at something like: `rust/build/x86_64-unknown-linux-gnu/llvm/bin/llvm-*`.
130+
- You can install compatible versions of these tools via `rustup`.
131+
132+
The `rustup` option is guaranteed to install a compatible version of the LLVM tools, but they can be hard to find. We recommend [`cargo-binutils`], which installs Rust-specific wrappers around these and other LLVM tools, so you can invoke them via `cargo` commands!
133+
134+
```shell
135+
$ rustup component add llvm-tools-preview
136+
$ cargo install cargo-binutils
137+
$ cargo profdata -- --help # note the additional "--" preceding the tool-specific arguments
138+
```
139+
140+
[`cargo-binutils`]: https://crates.io/crates/cargo-binutils
141+
142+
## Creating coverage reports
143+
144+
Raw profiles have to be indexed before they can be used to generate coverage reports. This is done using [`llvm-profdata merge`] (or `cargo profdata -- merge`), which can combine multiple raw profiles and index them at the same time:
145+
146+
```shell
147+
$ llvm-profdata merge -sparse formatjson5.profraw -o formatjson5.profdata
148+
```
149+
150+
Finally, the `.profdata` file is used, in combination with the coverage map (from the program binary) to generate coverage reports using [`llvm-cov report`] (or `cargo cov -- report`), for a coverage summaries; and [`llvm-cov show`] (or `cargo cov -- show`), to see detailed coverage of lines and regions (character ranges) overlaid on the original source code.
151+
152+
These commands have several display and filtering options. For example:
153+
154+
```shell
155+
$ llvm-cov show -Xdemangler=rustfilt target/debug/examples/formatjson5 \
156+
-instr-profile=formatjson5.profdata \
157+
-show-line-counts-or-regions \
158+
-show-instantiations \
159+
-name=add_quoted_string
160+
```
161+
162+
<img alt="Screenshot of sample `llvm-cov show` result, for function add_quoted_string" src="img/llvm-cov-show-01.png" class="center"/>
163+
<br/>
164+
<br/>
165+
166+
Some of the more notable options in this example include:
167+
168+
- `--Xdemangler=rustfilt` - the command name or path used to demangle Rust symbols (`rustfilt` in the example, but this could also be a path to the `rust-demangler` tool)
169+
- `target/debug/examples/formatjson5` - the instrumented binary (from which to extract the coverage map)
170+
- `--instr-profile=<path-to-file>.profdata` - the location of the `.profdata` file created by `llvm-profdata merge` (from the `.profraw` file generated by the instrumented binary)
171+
- `--name=<exact-function-name>` - to show coverage for a specific function (or, consider using another filter option, such as `--name-regex=<pattern>`)
172+
173+
[`llvm-profdata merge`]: https://llvm.org/docs/CommandGuide/llvm-profdata.html#profdata-merge
174+
[`llvm-cov report`]: https://llvm.org/docs/CommandGuide/llvm-cov.html#llvm-cov-report
175+
[`llvm-cov show`]: https://llvm.org/docs/CommandGuide/llvm-cov.html#llvm-cov-show
176+
177+
> **Note**: Coverage can also be disabled on an individual function by annotating the function with the [`no_coverage` attribute] (which requires the feature flag `#![feature(no_coverage)]`).
178+
179+
[`no_coverage` attribute]: ../language-features/no-coverage.md
180+
181+
## Interpreting reports
182+
183+
There are four statistics tracked in a coverage summary:
184+
185+
- Function coverage is the percentage of functions that have been executed at least once. A function is considered to be executed if any of its instantiations are executed.
186+
- Instantiation coverage is the percentage of function instantiations that have been executed at least once. Generic functions and functions generated from macros are two kinds of functions that may have multiple instantiations.
187+
- Line coverage is the percentage of code lines that have been executed at least once. Only executable lines within function bodies are considered to be code lines.
188+
- Region coverage is the percentage of code regions that have been executed at least once. A code region may span multiple lines: for example, in a large function body with no control flow. In other cases, a single line can contain multiple code regions: `return x || (y && z)` has countable code regions for `x` (which may resolve the expression, if `x` is `true`), `|| (y && z)` (executed only if `x` was `false`), and `return` (executed in either situation).
189+
190+
Of these four statistics, function coverage is usually the least granular while region coverage is the most granular. The project-wide totals for each statistic are listed in the summary.
191+
192+
## Test coverage
193+
194+
A typical use case for coverage analysis is test coverage. Rust's source-based coverage tools can both measure your tests' code coverage as percentage, and pinpoint functions and branches not tested.
195+
196+
The following example (using the [`json5format`] crate, for demonstration purposes) show how to generate and analyze coverage results for all tests in a crate.
197+
198+
Since `cargo test` both builds and runs the tests, we set both the additional `RUSTFLAGS`, to add the `-Z instrument-coverage` flag, and `LLVM_PROFILE_FILE`, to set a custom filename for the raw profiling data generated during the test runs. Since there may be more than one test binary, apply `%m` in the filename pattern. This generates unique names for each test binary. (Otherwise, each executed test binary would overwrite the coverage results from the previous binary.)
199+
200+
```shell
201+
$ RUSTFLAGS="-Z instrument-coverage" \
202+
LLVM_PROFILE_FILE="json5format-%m.profraw" \
203+
cargo test --tests
204+
```
205+
206+
Make note of the test binary file paths, displayed after the word "`Running`" in the test output:
207+
208+
```text
209+
...
210+
Compiling json5format v0.1.3 ($HOME/json5format)
211+
Finished test [unoptimized + debuginfo] target(s) in 14.60s
212+
213+
Running target/debug/deps/json5format-fececd4653271682
214+
running 25 tests
215+
...
216+
test result: ok. 25 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
217+
218+
Running target/debug/deps/lib-30768f9c53506dc5
219+
running 31 tests
220+
...
221+
test result: ok. 31 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
222+
```
223+
224+
You should have one or more `.profraw` files now, one for each test binary. Run the `profdata` tool to merge them:
225+
226+
```shell
227+
$ cargo profdata -- merge \
228+
-sparse json5format-*.profraw -o json5format.profdata
229+
```
230+
231+
Then run the `cov` tool, with the `profdata` file and all test binaries:
232+
233+
```shell
234+
$ cargo cov -- report \
235+
--use-color --ignore-filename-regex='/.cargo/registry' \
236+
--instr-profile=json5format.profdata \
237+
--object target/debug/deps/lib-30768f9c53506dc5 \
238+
--object target/debug/deps/json5format-fececd4653271682
239+
$ cargo cov -- show \
240+
--use-color --ignore-filename-regex='/.cargo/registry' \
241+
--instr-profile=json5format.profdata \
242+
--object target/debug/deps/lib-30768f9c53506dc5 \
243+
--object target/debug/deps/json5format-fececd4653271682 \
244+
--show-instantiations --show-line-counts-or-regions \
245+
--Xdemangler=rustfilt | less -R
246+
```
247+
248+
> **Note**: The command line option `--ignore-filename-regex=/.cargo/registry`, which excludes the sources for dependencies from the coverage results.\_
249+
250+
### Tips for listing the binaries automatically
251+
252+
For `bash` users, one suggested way to automatically complete the `cov` command with the list of binaries is with a command like:
253+
254+
```bash
255+
$ cargo cov -- report \
256+
$( \
257+
for file in \
258+
$( \
259+
RUSTFLAGS="-Z instrument-coverage" \
260+
cargo test --tests --no-run --message-format=json \
261+
| jq -r "select(.profile.test == true) | .filenames[]" \
262+
| grep -v dSYM - \
263+
); \
264+
do \
265+
printf "%s %s " -object $file; \
266+
done \
267+
) \
268+
--instr-profile=json5format.profdata --summary-only # and/or other options
269+
```
270+
271+
Adding `--no-run --message-format=json` to the _same_ `cargo test` command used to run
272+
the tests (including the same environment variables and flags) generates output in a JSON
273+
format that `jq` can easily query.
274+
275+
The `printf` command takes this list and generates the `--object <binary>` arguments
276+
for each listed test binary.
277+
278+
### Including doc tests
279+
280+
The previous examples run `cargo test` with `--tests`, which excludes doc tests.[^79417]
281+
282+
To include doc tests in the coverage results, drop the `--tests` flag, and apply the
283+
`-Z instrument-coverage` flag, and some doc-test-specific options in the
284+
`RUSTDOCFLAGS` environment variable. (The `cargo profdata` command does not change.)
285+
286+
```bash
287+
$ RUSTFLAGS="-Z instrument-coverage" \
288+
RUSTDOCFLAGS="-Z instrument-coverage -Z unstable-options --persist-doctests target/debug/doctestbins" \
289+
LLVM_PROFILE_FILE="json5format-%m.profraw" \
290+
cargo test
291+
$ cargo profdata -- merge \
292+
-sparse json5format-*.profraw -o json5format.profdata
293+
```
294+
295+
The `-Z unstable-options --persist-doctests` flag is required, to save the test binaries
296+
(with their coverage maps) for `llvm-cov`.
297+
298+
```bash
299+
$ cargo cov -- report \
300+
$( \
301+
for file in \
302+
$( \
303+
RUSTFLAGS="-Z instrument-coverage" \
304+
RUSTDOCFLAGS="-Z instrument-coverage -Z unstable-options --persist-doctests target/debug/doctestbins" \
305+
cargo test --no-run --message-format=json \
306+
| jq -r "select(.profile.test == true) | .filenames[]" \
307+
| grep -v dSYM - \
308+
) \
309+
target/debug/doctestbins/*/rust_out; \
310+
do \
311+
[[ -x $file ]] && printf "%s %s " -object $file; \
312+
done \
313+
) \
314+
--instr-profile=json5format.profdata --summary-only # and/or other options
315+
```
316+
317+
> **Note**: The differences in this `cargo cov` command, compared with the version without
318+
> doc tests, include:
319+
320+
- The `cargo test ... --no-run` command is updated with the same environment variables
321+
and flags used to _build_ the tests, _including_ the doc tests. (`LLVM_PROFILE_FILE`
322+
is only used when _running_ the tests.)
323+
- The file glob pattern `target/debug/doctestbins/*/rust_out` adds the `rust_out`
324+
binaries generated for doc tests (note, however, that some `rust_out` files may not
325+
be executable binaries).
326+
- `[[ -x $file ]] &&` filters the files passed on to the `printf`, to include only
327+
executable binaries.
328+
329+
[^79417]:
330+
There is ongoing work to resolve a known issue
331+
[(#79417)](https://github.com/rust-lang/rust/issues/79417) that doc test coverage
332+
generates incorrect source line numbers in `llvm-cov show` results.
333+
334+
## `-Z instrument-coverage=<options>`
335+
336+
- `-Z instrument-coverage=all`: Instrument all functions, including unused functions and unused generics. (This is the same as `-Z instrument-coverage`, with no value.)
337+
- `-Z instrument-coverage=except-unused-generics`: Instrument all functions except unused generics.
338+
- `-Z instrument-coverage=except-unused-functions`: Instrument only used (called) functions and instantiated generic functions.
339+
- `-Z instrument-coverage=off`: Do not instrument any functions. (This is the same as simply not including the `-Z instrument-coverage` option.)
340+
341+
## Other references
342+
343+
Rust's implementation and workflow for source-based code coverage is based on the same library and tools used to implement [source-based code coverage in Clang]. (This document is partially based on the Clang guide.)
344+
345+
[source-based code coverage in clang]: https://clang.llvm.org/docs/SourceBasedCodeCoverage.html
346+
[`json5format`]: https://crates.io/crates/json5format

0 commit comments

Comments
 (0)