From 3d05d90a65ebbe6adef4e9edbda18db48de31df5 Mon Sep 17 00:00:00 2001
From: Michael Droettboom <mdboom@gmail.com>
Date: Wed, 13 Jul 2022 14:11:02 -0400
Subject: [PATCH] Refactor this as a Makefile target

---
 .github/workflows/c-coverage.yml              |  19 +---
 .gitignore                                    |   1 +
 Makefile.pre.in                               |  44 ++++++++
 ...2-07-13-13-24-01.gh-issue-94759.dTLMod.rst |   4 +
 configure                                     | 100 ++++++++++++++++++
 configure.ac                                  |   2 +
 6 files changed, 156 insertions(+), 14 deletions(-)
 create mode 100644 Misc/NEWS.d/next/Build/2022-07-13-13-24-01.gh-issue-94759.dTLMod.rst

diff --git a/.github/workflows/c-coverage.yml b/.github/workflows/c-coverage.yml
index 2f3718c515c5f1..814e95dfba8a96 100644
--- a/.github/workflows/c-coverage.yml
+++ b/.github/workflows/c-coverage.yml
@@ -40,26 +40,17 @@ jobs:
       uses: hendrikmuhs/ccache-action@v1
     - name: Configure clang (with coverage)
       run: |
-        echo "CC=/usr/lib/ccache/clang-12 -fprofile-instr-generate -fcoverage-mapping" >> $GITHUB_ENV
-        echo "CXX=/usr/lib/ccache/clang++-12 -fprofile-instr-generate -fcoverage-mapping" >> $GITHUB_ENV
+        echo "CC=clang" >> $GITHUB_ENV
+        echo "CXX=clang++" >> $GITHUB_ENV
     - name: Configure CPython
       run: ./configure --with-pydebug --with-openssl=$OPENSSL_DIR
-    - name: Build CPython
-      run: make -j4
-    - name: Collect coverage data
-      # Specify the LLVM_PROFILE_FILE using %m so multiple shared objects can write
-      # in parallel. Set the full path to the directory so results aren't written
-      # into temporary directories created by tests.
-      # Using "-j 1" is important, or the Github Action runs out of memory
-      run: LLVM_PROFILE_FILE=${PWD}/python%m.profraw xvfb-run ./python -m test -j 1
     - name: Generate coverage report
-      run: |
-        llvm-profdata-12 merge -sparse python*.profraw -o python.profdata
-        llvm-cov-12 show -format=html -output-dir=cpython-coverage -instr-profile=python.profdata -show-branches=count -show-regions python .
+      # Using "-j1" is important, or the Github Action runs out of memory
+      run: EXTRATESTOPTS=-j1 xvfb-run make coverage-report-llvm
     - name: Publish coverage-report
       uses: JamesIves/github-pages-deploy-action@v4
       with:
-        folder: cpython-coverage
+        folder: llvm-cov-report
         repository-name: '' # TODO Destination
         token: ${{ secrets.COVERAGE_DEPLOY_TOKEN }} # TODO: Use an organization-level token
         single-commit: true
diff --git a/.gitignore b/.gitignore
index 0ddfd717d737f8..592324ea731149 100644
--- a/.gitignore
+++ b/.gitignore
@@ -21,6 +21,7 @@
 *.gc??
 *.profclang?
 *.profraw
+*.profdata
 *.dyn
 .gdb_history
 .purify
diff --git a/Makefile.pre.in b/Makefile.pre.in
index 77696fca940787..a5e30deed72b3f 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -51,6 +51,8 @@ PGO_PROF_USE_FLAG=@PGO_PROF_USE_FLAG@
 LLVM_PROF_MERGER=@LLVM_PROF_MERGER@
 LLVM_PROF_FILE=@LLVM_PROF_FILE@
 LLVM_PROF_ERR=@LLVM_PROF_ERR@
+LLVM_PROFDATA=@LLVM_PROFDATA@
+LLVM_COV=@LLVM_COV@
 DTRACE=         @DTRACE@
 DFLAGS=         @DFLAGS@
 DTRACE_HEADERS= @DTRACE_HEADERS@
@@ -321,6 +323,10 @@ COVERAGE_REPORT=$(abs_builddir)/lcov-report
 COVERAGE_LCOV_OPTIONS=--rc lcov_branch_coverage=1
 COVERAGE_REPORT_OPTIONS=--rc lcov_branch_coverage=1 --branch-coverage --title "CPython $(VERSION) LCOV report [commit $(shell $(GITVERSION))]"
 
+# report files for llvm-cov coverage report
+COVERAGE_INFO_LLVM=	$(abs_builddir)/coverage.profdata
+COVERAGE_REPORT_LLVM=$(abs_builddir)/llvm-cov-report
+COVERAGE_REPORT_OPTIONS_LLVM=-show-branches=count -show-regions
 
 # === Definitions added by makesetup ===
 
@@ -693,6 +699,44 @@ coverage-report: regen-token regen-frozen
 	@ # build lcov report
 	$(MAKE) coverage-lcov
 
+# Compile and calculate coverage with llvm-cov
+.PHONY=check-clang coverage-llvm coverage-profdata coverage-report-llvm
+
+# Check whether the compiler is clang, and if not, error out.
+check-clang:
+	($(CC) --version | grep clang) || \
+	(echo "LLVM coverage only works with clang. Set CC=clang and CXX=clang++ and re-run ./configure"; exit 1)
+
+coverage-llvm: check-clang
+	@echo "Building with support for coverage checking:"
+	$(MAKE) clean
+	@ # Override CC rather than CFLAGS since these flags must come first
+	$(MAKE) @DEF_MAKE_RULE@ CC="$(CC) -fprofile-instr-generate -fcoverage-mapping"
+
+coverage-profdata:
+	@echo "Creating Coverage HTML report with llvm-profdata/llvm-cov:"
+	@rm -f $(COVERAGE_INFO_LLVM)
+	@rm -rf $(COVERAGE_REPORT_LLVM)
+	@ # Merge coverage results
+	$(LLVM_PROFDATA) merge -sparse python*.profraw -o $(COVERAGE_INFO_LLVM)
+	@ # Generate HTML
+	$(LLVM_COV) show -format=html -output-dir=$(COVERAGE_REPORT_LLVM) -instr-profile=$(COVERAGE_INFO_LLVM) $(COVERAGE_REPORT_OPTIONS_LLVM) python .
+	@echo
+	@echo "llvm-cov report at $(COVERAGE_REPORT_LLVM)/index.html"
+	@echo
+
+# Force regeneration of parser and importlib
+# Specify the LLVM_PROFILE_FILE using %m so multiple shared objects can write
+# in parallel. Set the full path to the directory so results aren't written
+# into temporary directories created by tests.
+coverage-report-llvm: regen-token regen-importlib
+	@ # build with coverage info
+	$(MAKE) coverage-llvm
+	@ # run tests, ignore failures
+	LLVM_PROFILE_FILE=${PWD}/python%m.profraw $(TESTRUNNER) $(TESTOPTS) || true
+	@ # build llvm-cov report
+	$(MAKE) coverage-profdata
+
 # Run "Argument Clinic" over all source files
 .PHONY=clinic
 clinic: check-clean-src $(srcdir)/Modules/_blake2/blake2s_impl.c
diff --git a/Misc/NEWS.d/next/Build/2022-07-13-13-24-01.gh-issue-94759.dTLMod.rst b/Misc/NEWS.d/next/Build/2022-07-13-13-24-01.gh-issue-94759.dTLMod.rst
new file mode 100644
index 00000000000000..74a17daec2dbf9
--- /dev/null
+++ b/Misc/NEWS.d/next/Build/2022-07-13-13-24-01.gh-issue-94759.dTLMod.rst
@@ -0,0 +1,4 @@
+A new Makefile target ``coverage-report-llvm`` will use ``clang`` and
+``llvm-cov`` to generate a coverage report. This provides more details about
+branch coverage and subexpressions than the existing ``gcc`` and ``lcov``
+based ``coverage-report``.
diff --git a/configure b/configure
index bffc30a590fbe4..f2c787b1e63e58 100755
--- a/configure
+++ b/configure
@@ -882,6 +882,7 @@ CFLAGS_NODIST
 BASECFLAGS
 CFLAGS_ALIASING
 OPT
+LLVM_COV
 LLVM_PROF_FOUND
 LLVM_PROFDATA
 LLVM_PROF_ERR
@@ -7924,6 +7925,105 @@ case $CC in
     ;;
 esac
 
+if test -n "$ac_tool_prefix"; then
+  # Extract the first word of "${ac_tool_prefix}llvm-cov", so it can be a program name with args.
+set dummy ${ac_tool_prefix}llvm-cov; ac_word=$2
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $ac_word" >&5
+$as_echo_n "checking for $ac_word... " >&6; }
+if ${ac_cv_path_LLVM_COV+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  case $LLVM_COV in
+  [\\/]* | ?:[\\/]*)
+  ac_cv_path_LLVM_COV="$LLVM_COV" # Let the user override the test with a path.
+  ;;
+  *)
+  as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in ${llvm_path}
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+    for ac_exec_ext in '' $ac_executable_extensions; do
+  if as_fn_executable_p "$as_dir/$ac_word$ac_exec_ext"; then
+    ac_cv_path_LLVM_COV="$as_dir/$ac_word$ac_exec_ext"
+    $as_echo "$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+  done
+IFS=$as_save_IFS
+
+  ;;
+esac
+fi
+LLVM_COV=$ac_cv_path_LLVM_COV
+if test -n "$LLVM_COV"; then
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: $LLVM_COV" >&5
+$as_echo "$LLVM_COV" >&6; }
+else
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5
+$as_echo "no" >&6; }
+fi
+
+
+fi
+if test -z "$ac_cv_path_LLVM_COV"; then
+  ac_pt_LLVM_COV=$LLVM_COV
+  # Extract the first word of "llvm-cov", so it can be a program name with args.
+set dummy llvm-cov; ac_word=$2
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $ac_word" >&5
+$as_echo_n "checking for $ac_word... " >&6; }
+if ${ac_cv_path_ac_pt_LLVM_COV+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  case $ac_pt_LLVM_COV in
+  [\\/]* | ?:[\\/]*)
+  ac_cv_path_ac_pt_LLVM_COV="$ac_pt_LLVM_COV" # Let the user override the test with a path.
+  ;;
+  *)
+  as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in ${llvm_path}
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+    for ac_exec_ext in '' $ac_executable_extensions; do
+  if as_fn_executable_p "$as_dir/$ac_word$ac_exec_ext"; then
+    ac_cv_path_ac_pt_LLVM_COV="$as_dir/$ac_word$ac_exec_ext"
+    $as_echo "$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+  done
+IFS=$as_save_IFS
+
+  ;;
+esac
+fi
+ac_pt_LLVM_COV=$ac_cv_path_ac_pt_LLVM_COV
+if test -n "$ac_pt_LLVM_COV"; then
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_pt_LLVM_COV" >&5
+$as_echo "$ac_pt_LLVM_COV" >&6; }
+else
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5
+$as_echo "no" >&6; }
+fi
+
+  if test "x$ac_pt_LLVM_COV" = x; then
+    LLVM_COV="''"
+  else
+    case $cross_compiling:$ac_tool_warned in
+yes:)
+{ $as_echo "$as_me:${as_lineno-$LINENO}: WARNING: using cross tools not prefixed with host triplet" >&5
+$as_echo "$as_me: WARNING: using cross tools not prefixed with host triplet" >&2;}
+ac_tool_warned=yes ;;
+esac
+    LLVM_COV=$ac_pt_LLVM_COV
+  fi
+else
+  LLVM_COV="$ac_cv_path_LLVM_COV"
+fi
+
+
 # XXX Shouldn't the code above that fiddles with BASECFLAGS and OPT be
 # merged with this chunk of code?
 
diff --git a/configure.ac b/configure.ac
index f940c9b2ddf001..b1ec4be8723a81 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1957,6 +1957,8 @@ case $CC in
     LLVM_PROF_FILE=""
     ;;
 esac
+AC_SUBST(LLVM_COV)
+AC_PATH_TOOL(LLVM_COV, llvm-cov, '', ${llvm_path})
 
 # XXX Shouldn't the code above that fiddles with BASECFLAGS and OPT be
 # merged with this chunk of code?