rust/tests/run-make/coverage-reports/Makefile

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

179 lines
8.2 KiB
Makefile
Raw Normal View History

# needs-profiler-support
# ignore-windows-gnu
# FIXME(pietroalbini): this test currently does not work on cross-compiled
# targets because remote-test is not capable of sending back the *.profraw
# files generated by the LLVM instrumentation.
# ignore-cross-compile
# Rust coverage maps support LLVM Coverage Mapping Format versions 5 and 6,
# corresponding with LLVM versions 12 and 13, respectively.
# When upgrading LLVM versions, consider whether to enforce a minimum LLVM
# version during testing, with an additional directive at the top of this file
# that sets, for example: `min-llvm-version: 12.0`
# FIXME(mati865): MinGW GCC miscompiles compiler-rt profiling library but with Clang it works
# properly. Since we only have GCC on the CI ignore the test for now.
2022-07-24 19:00:00 -05:00
include ../coverage/coverage_tools.mk
BASEDIR=../coverage-reports
SOURCEDIR=../coverage
# The `llvm-cov show` flag `--debug`, used to generate the `counters` output files, is only
# enabled if LLVM assertions are enabled. This requires Rust config `llvm/optimize` and not
# `llvm/release_debuginfo`. Note that some CI builds disable debug assertions (by setting
# `NO_LLVM_ASSERTIONS=1`), so the tests must still pass even if the `--debug` flag is
# not supported. (Note that `counters` files are only produced in the `$(TMPDIR)`
# directory, for inspection and debugging support. They are *not* copied to `expected_*`
# files when `--bless`ed.)
LLVM_COV_DEBUG := $(shell \
"$(LLVM_BIN_DIR)"/llvm-cov show --debug 2>&1 | \
grep -q "Unknown command line argument '--debug'"; \
echo $$?)
ifeq ($(LLVM_COV_DEBUG), 1)
DEBUG_FLAG=--debug
endif
# FIXME(richkadel): I'm adding `--ignore-filename-regex=` line(s) for specific test(s) that produce
# `llvm-cov` results for multiple files (for example `uses_crate.rs` and `used_crate/mod.rs`) as a
# workaround for two problems causing tests to fail on Windows:
#
# 1. When multiple files appear in the `llvm-cov show` results, each file's coverage results can
# appear in different a different order. Whether this is random or, somehow, platform-specific,
# the Windows output flips the order of the files, compared to Linux. In the `uses_crate.rs`
# test, the only test-unique (interesting) results we care about are the results for only one
# of the two files, `mod/uses_crate.rs`, so the workaround is to ignore all but this one file.
# In the future, we may want a more sophisticated solution that splits apart `llvm-cov show`
# results into separate results files for each result (taking care not to create new file
# paths that might be too long for Windows MAX_PATH limits when creating these new sub-results,
# as well).
# 2. When multiple files appear in the `llvm-cov show` results, the results for each file are
# prefixed with their filename, including platform-specific path separators (`\` for Windows,
# and `/` everywhere else). This could be filtered or normalized of course, but by ignoring
# coverage results for all but one of the file, the filenames are no longer included anyway.
# If this changes (if/when we decide to support `llvm-cov show` results for multiple files),
# the file path separator differences may need to be addressed.
#
# Since this is only a workaround, I decided to implement the override by adding an option for
# each file to be ignored, using a `--ignore-filename-regex=` entry for each one, rather than
# implement some more sophisticated solution with a new custom test directive in the test file
# itself (similar to `expect-exit-status`) because that would add a lot of complexity and still
# be a workaround, with the same result, with no benefit.
#
# Yes these `--ignore-filename-regex=` options are included in all invocations of `llvm-cov show`
# for now, but it is effectively ignored for all tests that don't include this file anyway.
#
# (Note that it's also possible the `_counters.<test>.txt` and `<test>.json` files (if generated)
# may order results from multiple files inconsistently, which might also have to be accommodated
# if and when we allow `llvm-cov` to produce results for multiple files. Note, the path separators
# appear to be normalized to `/` in those files, thankfully.)
LLVM_COV_IGNORE_FILES=\
--ignore-filename-regex='(uses_crate.rs|uses_inline_crate.rs|unused_mod.rs)'
all: $(patsubst $(SOURCEDIR)/lib/%.rs,%,$(wildcard $(SOURCEDIR)/lib/*.rs)) $(patsubst $(SOURCEDIR)/%.rs,%,$(wildcard $(SOURCEDIR)/*.rs))
# Ensure there are no `expected` results for tests that may have been removed or renamed
.PHONY: clear_expected_if_blessed
clear_expected_if_blessed:
ifdef RUSTC_BLESS_TEST
rm -f expected_*
endif
-include clear_expected_if_blessed
%: $(SOURCEDIR)/lib/%.rs
# Compile the test library with coverage instrumentation
$(RUSTC) $(SOURCEDIR)/lib/$@.rs \
coverage bug fixes and optimization support Adjusted LLVM codegen for code compiled with `-Zinstrument-coverage` to address multiple, somewhat related issues. Fixed a significant flaw in prior coverage solution: Every counter generated a new counter variable, but there should have only been one counter variable per function. This appears to have bloated .profraw files significantly. (For a small program, it increased the size by about 40%. I have not tested large programs, but there is anecdotal evidence that profraw files were way too large. This is a good fix, regardless, but hopefully it also addresses related issues. Fixes: #82144 Invalid LLVM coverage data produced when compiled with -C opt-level=1 Existing tests now work up to at least `opt-level=3`. This required a detailed analysis of the LLVM IR, comparisons with Clang C++ LLVM IR when compiled with coverage, and a lot of trial and error with codegen adjustments. The biggest hurdle was figuring out how to continue to support coverage results for unused functions and generics. Rust's coverage results have three advantages over Clang's coverage results: 1. Rust's coverage map does not include any overlapping code regions, making coverage counting unambiguous. 2. Rust generates coverage results (showing zero counts) for all unused functions, including generics. (Clang does not generate coverage for uninstantiated template functions.) 3. Rust's unused functions produce minimal stubbed functions in LLVM IR, sufficient for including in the coverage results; while Clang must generate the complete LLVM IR for each unused function, even though it will never be called. This PR removes the previous hack of attempting to inject coverage into some other existing function instance, and generates dedicated instances for each unused function. This change, and a few other adjustments (similar to what is required for `-C link-dead-code`, but with lower impact), makes it possible to support LLVM optimizations. Fixes: #79651 Coverage report: "Unexecuted instantiation:..." for a generic function from multiple crates Fixed by removing the aforementioned hack. Some "Unexecuted instantiation" notices are unavoidable, as explained in the `used_crate.rs` test, but `-Zinstrument-coverage` has new options to back off support for either unused generics, or all unused functions, which avoids the notice, at the cost of less coverage of unused functions. Fixes: #82875 Invalid LLVM coverage data produced with crate brotli_decompressor Fixed by disabling the LLVM function attribute that forces inlining, if `-Z instrument-coverage` is enabled. This attribute is applied to Rust functions with `#[inline(always)], and in some cases, the forced inlining breaks coverage instrumentation and reports.
2021-03-15 18:32:45 -05:00
$$( sed -n 's/^\/\/ compile-flags: \([^#]*\).*/\1/p' $(SOURCEDIR)/lib/$@.rs ) \
--crate-type rlib -Cinstrument-coverage --target $(TARGET)
%: $(SOURCEDIR)/%.rs
# Compile the test program with coverage instrumentation
$(RUSTC) $(SOURCEDIR)/$@.rs \
coverage bug fixes and optimization support Adjusted LLVM codegen for code compiled with `-Zinstrument-coverage` to address multiple, somewhat related issues. Fixed a significant flaw in prior coverage solution: Every counter generated a new counter variable, but there should have only been one counter variable per function. This appears to have bloated .profraw files significantly. (For a small program, it increased the size by about 40%. I have not tested large programs, but there is anecdotal evidence that profraw files were way too large. This is a good fix, regardless, but hopefully it also addresses related issues. Fixes: #82144 Invalid LLVM coverage data produced when compiled with -C opt-level=1 Existing tests now work up to at least `opt-level=3`. This required a detailed analysis of the LLVM IR, comparisons with Clang C++ LLVM IR when compiled with coverage, and a lot of trial and error with codegen adjustments. The biggest hurdle was figuring out how to continue to support coverage results for unused functions and generics. Rust's coverage results have three advantages over Clang's coverage results: 1. Rust's coverage map does not include any overlapping code regions, making coverage counting unambiguous. 2. Rust generates coverage results (showing zero counts) for all unused functions, including generics. (Clang does not generate coverage for uninstantiated template functions.) 3. Rust's unused functions produce minimal stubbed functions in LLVM IR, sufficient for including in the coverage results; while Clang must generate the complete LLVM IR for each unused function, even though it will never be called. This PR removes the previous hack of attempting to inject coverage into some other existing function instance, and generates dedicated instances for each unused function. This change, and a few other adjustments (similar to what is required for `-C link-dead-code`, but with lower impact), makes it possible to support LLVM optimizations. Fixes: #79651 Coverage report: "Unexecuted instantiation:..." for a generic function from multiple crates Fixed by removing the aforementioned hack. Some "Unexecuted instantiation" notices are unavoidable, as explained in the `used_crate.rs` test, but `-Zinstrument-coverage` has new options to back off support for either unused generics, or all unused functions, which avoids the notice, at the cost of less coverage of unused functions. Fixes: #82875 Invalid LLVM coverage data produced with crate brotli_decompressor Fixed by disabling the LLVM function attribute that forces inlining, if `-Z instrument-coverage` is enabled. This attribute is applied to Rust functions with `#[inline(always)], and in some cases, the forced inlining breaks coverage instrumentation and reports.
2021-03-15 18:32:45 -05:00
$$( sed -n 's/^\/\/ compile-flags: \([^#]*\).*/\1/p' $(SOURCEDIR)/$@.rs ) \
-L "$(TMPDIR)" -Cinstrument-coverage --target $(TARGET)
# Run it in order to generate some profiling data,
# with `LLVM_PROFILE_FILE=<profdata_file>` environment variable set to
# output the coverage stats for this run.
LLVM_PROFILE_FILE="$(TMPDIR)"/$@.profraw \
Updates to experimental coverage counter injection This is a combination of 18 commits. Commit #2: Additional examples and some small improvements. Commit #3: fixed mir-opt non-mir extensions and spanview title elements Corrected a fairly recent assumption in runtest.rs that all MIR dump files end in .mir. (It was appending .mir to the graphviz .dot and spanview .html file names when generating blessed output files. That also left outdated files in the baseline alongside the files with the incorrect names, which I've now removed.) Updated spanview HTML title elements to match their content, replacing a hardcoded and incorrect name that was left in accidentally when originally submitted. Commit #4: added more test examples also improved Makefiles with support for non-zero exit status and to force validation of tests unless a specific test overrides it with a specific comment. Commit #5: Fixed rare issues after testing on real-world crate Commit #6: Addressed PR feedback, and removed temporary -Zexperimental-coverage -Zinstrument-coverage once again supports the latest capabilities of LLVM instrprof coverage instrumentation. Also fixed a bug in spanview. Commit #7: Fix closure handling, add tests for closures and inner items And cleaned up other tests for consistency, and to make it more clear where spans start/end by breaking up lines. Commit #8: renamed "typical" test results "expected" Now that the `llvm-cov show` tests are improved to normally expect matching actuals, and to allow individual tests to override that expectation. Commit #9: test coverage of inline generic struct function Commit #10: Addressed review feedback * Removed unnecessary Unreachable filter. * Replaced a match wildcard with remining variants. * Added more comments to help clarify the role of successors() in the CFG traversal Commit #11: refactoring based on feedback * refactored `fn coverage_spans()`. * changed the way I expand an empty coverage span to improve performance * fixed a typo that I had accidently left in, in visit.rs Commit #12: Optimized use of SourceMap and SourceFile Commit #13: Fixed a regression, and synched with upstream Some generated test file names changed due to some new change upstream. Commit #14: Stripping out crate disambiguators from demangled names These can vary depending on the test platform. Commit #15: Ignore llvm-cov show diff on test with generics, expand IO error message Tests with generics produce llvm-cov show results with demangled names that can include an unstable "crate disambiguator" (hex value). The value changes when run in the Rust CI Windows environment. I added a sed filter to strip them out (in a prior commit), but sed also appears to fail in the same environment. Until I can figure out a workaround, I'm just going to ignore this specific test result. I added a FIXME to follow up later, but it's not that critical. I also saw an error with Windows GNU, but the IO error did not specify a path for the directory or file that triggered the error. I updated the error messages to provide more info for next, time but also noticed some other tests with similar steps did not fail. Looks spurious. Commit #16: Modify rust-demangler to strip disambiguators by default Commit #17: Remove std::process::exit from coverage tests Due to Issue #77553, programs that call std::process::exit() do not generate coverage results on Windows MSVC. Commit #18: fix: test file paths exceeding Windows max path len
2020-09-01 18:15:17 -05:00
$(call RUN,$@) || \
( \
status=$$?; \
grep -q "^\/\/ expect-exit-status-$$status" $(SOURCEDIR)/$@.rs || \
( >&2 echo "program exited with an unexpected exit status: $$status"; \
false \
) \
)
# Run it through rustdoc as well to cover doctests.
# `%p` is the pid, and `%m` the binary signature. We suspect that the pid alone
# might result in overwritten files and failed tests, as rustdoc spawns each
# doctest as its own process, so make sure the filename is as unique as possible.
LLVM_PROFILE_FILE="$(TMPDIR)"/$@-%p-%m.profraw \
$(RUSTDOC) --crate-name workaround_for_79771 --test $(SOURCEDIR)/$@.rs \
coverage bug fixes and optimization support Adjusted LLVM codegen for code compiled with `-Zinstrument-coverage` to address multiple, somewhat related issues. Fixed a significant flaw in prior coverage solution: Every counter generated a new counter variable, but there should have only been one counter variable per function. This appears to have bloated .profraw files significantly. (For a small program, it increased the size by about 40%. I have not tested large programs, but there is anecdotal evidence that profraw files were way too large. This is a good fix, regardless, but hopefully it also addresses related issues. Fixes: #82144 Invalid LLVM coverage data produced when compiled with -C opt-level=1 Existing tests now work up to at least `opt-level=3`. This required a detailed analysis of the LLVM IR, comparisons with Clang C++ LLVM IR when compiled with coverage, and a lot of trial and error with codegen adjustments. The biggest hurdle was figuring out how to continue to support coverage results for unused functions and generics. Rust's coverage results have three advantages over Clang's coverage results: 1. Rust's coverage map does not include any overlapping code regions, making coverage counting unambiguous. 2. Rust generates coverage results (showing zero counts) for all unused functions, including generics. (Clang does not generate coverage for uninstantiated template functions.) 3. Rust's unused functions produce minimal stubbed functions in LLVM IR, sufficient for including in the coverage results; while Clang must generate the complete LLVM IR for each unused function, even though it will never be called. This PR removes the previous hack of attempting to inject coverage into some other existing function instance, and generates dedicated instances for each unused function. This change, and a few other adjustments (similar to what is required for `-C link-dead-code`, but with lower impact), makes it possible to support LLVM optimizations. Fixes: #79651 Coverage report: "Unexecuted instantiation:..." for a generic function from multiple crates Fixed by removing the aforementioned hack. Some "Unexecuted instantiation" notices are unavoidable, as explained in the `used_crate.rs` test, but `-Zinstrument-coverage` has new options to back off support for either unused generics, or all unused functions, which avoids the notice, at the cost of less coverage of unused functions. Fixes: #82875 Invalid LLVM coverage data produced with crate brotli_decompressor Fixed by disabling the LLVM function attribute that forces inlining, if `-Z instrument-coverage` is enabled. This attribute is applied to Rust functions with `#[inline(always)], and in some cases, the forced inlining breaks coverage instrumentation and reports.
2021-03-15 18:32:45 -05:00
$$( sed -n 's/^\/\/ compile-flags: \([^#]*\).*/\1/p' $(SOURCEDIR)/$@.rs ) \
-L "$(TMPDIR)" -Cinstrument-coverage \
-Z unstable-options --persist-doctests=$(TMPDIR)/rustdoc-$@
# Postprocess the profiling data so it can be used by the llvm-cov tool
"$(LLVM_BIN_DIR)"/llvm-profdata merge --sparse \
"$(TMPDIR)"/$@*.profraw \
-o "$(TMPDIR)"/$@.profdata
# Generate a coverage report using `llvm-cov show`.
"$(LLVM_BIN_DIR)"/llvm-cov show \
$(DEBUG_FLAG) \
$(LLVM_COV_IGNORE_FILES) \
--compilation-dir=. \
--Xdemangler="$(RUST_DEMANGLER)" \
--show-line-counts-or-regions \
--instr-profile="$(TMPDIR)"/$@.profdata \
$(call BIN,"$(TMPDIR)"/$@) \
$$( \
2022-11-26 20:47:52 -06:00
for file in $(TMPDIR)/rustdoc-$@/*/rust_out*; do \
[ -x "$$file" ] && printf "%s %s " -object $$file; \
done \
) \
2> "$(TMPDIR)"/show_coverage_stderr.$@.txt \
| "$(PYTHON)" $(BASEDIR)/normalize_paths.py \
| "$(PYTHON)" $(BASEDIR)/sort_subviews.py \
> "$(TMPDIR)"/actual_show_coverage.$@.txt || \
( status=$$? ; \
>&2 cat "$(TMPDIR)"/show_coverage_stderr.$@.txt ; \
exit $$status \
)
ifdef DEBUG_FLAG
# The first line (beginning with "Args:" contains hard-coded, build-specific
# file paths. Strip that line and keep the remaining lines with counter debug
# data.
tail -n +2 "$(TMPDIR)"/show_coverage_stderr.$@.txt \
> "$(TMPDIR)"/actual_show_coverage_counters.$@.txt
endif
ifdef RUSTC_BLESS_TEST
cp "$(TMPDIR)"/actual_show_coverage.$@.txt \
expected_show_coverage.$@.txt
else
# Compare the show coverage output (`--bless` refreshes `typical` files).
#
# `llvm-cov show` normally prints instantiation groups in an unpredictable
# order, but we have used `sort_subviews.py` to sort them, so we can still
# check the output directly with `diff`.
#
# Some of the test cases are currently not working (since #110393) and have
# been marked with `// ignore-llvm-cov-show-diffs` so that they don't fail
# the build.
$(DIFF) \
expected_show_coverage.$@.txt "$(TMPDIR)"/actual_show_coverage.$@.txt || \
Updates to experimental coverage counter injection This is a combination of 18 commits. Commit #2: Additional examples and some small improvements. Commit #3: fixed mir-opt non-mir extensions and spanview title elements Corrected a fairly recent assumption in runtest.rs that all MIR dump files end in .mir. (It was appending .mir to the graphviz .dot and spanview .html file names when generating blessed output files. That also left outdated files in the baseline alongside the files with the incorrect names, which I've now removed.) Updated spanview HTML title elements to match their content, replacing a hardcoded and incorrect name that was left in accidentally when originally submitted. Commit #4: added more test examples also improved Makefiles with support for non-zero exit status and to force validation of tests unless a specific test overrides it with a specific comment. Commit #5: Fixed rare issues after testing on real-world crate Commit #6: Addressed PR feedback, and removed temporary -Zexperimental-coverage -Zinstrument-coverage once again supports the latest capabilities of LLVM instrprof coverage instrumentation. Also fixed a bug in spanview. Commit #7: Fix closure handling, add tests for closures and inner items And cleaned up other tests for consistency, and to make it more clear where spans start/end by breaking up lines. Commit #8: renamed "typical" test results "expected" Now that the `llvm-cov show` tests are improved to normally expect matching actuals, and to allow individual tests to override that expectation. Commit #9: test coverage of inline generic struct function Commit #10: Addressed review feedback * Removed unnecessary Unreachable filter. * Replaced a match wildcard with remining variants. * Added more comments to help clarify the role of successors() in the CFG traversal Commit #11: refactoring based on feedback * refactored `fn coverage_spans()`. * changed the way I expand an empty coverage span to improve performance * fixed a typo that I had accidently left in, in visit.rs Commit #12: Optimized use of SourceMap and SourceFile Commit #13: Fixed a regression, and synched with upstream Some generated test file names changed due to some new change upstream. Commit #14: Stripping out crate disambiguators from demangled names These can vary depending on the test platform. Commit #15: Ignore llvm-cov show diff on test with generics, expand IO error message Tests with generics produce llvm-cov show results with demangled names that can include an unstable "crate disambiguator" (hex value). The value changes when run in the Rust CI Windows environment. I added a sed filter to strip them out (in a prior commit), but sed also appears to fail in the same environment. Until I can figure out a workaround, I'm just going to ignore this specific test result. I added a FIXME to follow up later, but it's not that critical. I also saw an error with Windows GNU, but the IO error did not specify a path for the directory or file that triggered the error. I updated the error messages to provide more info for next, time but also noticed some other tests with similar steps did not fail. Looks spurious. Commit #16: Modify rust-demangler to strip disambiguators by default Commit #17: Remove std::process::exit from coverage tests Due to Issue #77553, programs that call std::process::exit() do not generate coverage results on Windows MSVC. Commit #18: fix: test file paths exceeding Windows max path len
2020-09-01 18:15:17 -05:00
( grep -q '^\/\/ ignore-llvm-cov-show-diffs' $(SOURCEDIR)/$@.rs && \
>&2 echo 'diff failed, but suppressed with `// ignore-llvm-cov-show-diffs` in $(SOURCEDIR)/$@.rs' \
) || \
( >&2 echo 'diff failed, and not suppressed without `// ignore-llvm-cov-show-diffs` in $(SOURCEDIR)/$@.rs'; \
false \
)
endif