2 files changed, 210 insertions, 0 deletions
diff --git a/src/rocksdb/coverage/coverage_test.sh b/src/rocksdb/coverage/coverage_test.sh
new file mode 100755
index 000000000..aa5f68c77
--- /dev/null
+++ b/src/rocksdb/coverage/coverage_test.sh
@@ -0,0 +1,82 @@
+#!/usr/bin/env bash
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+
+# Exit on error.
+set -e
+
+if [ -n "$USE_CLANG" ]; then
+  echo "Error: Coverage test is supported only for gcc."
+  exit 1
+fi
+
+ROOT=".."
+# Fetch right version of gcov
+if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
+  source $ROOT/build_tools/fbcode_config_platform009.sh
+  GCOV=$GCC_BASE/bin/gcov
+else
+  GCOV=$(which gcov)
+fi
+echo -e "Using $GCOV"
+
+COVERAGE_DIR="$PWD/COVERAGE_REPORT"
+mkdir -p $COVERAGE_DIR
+
+# Find all gcno files to generate the coverage report
+
+PYTHON=${1:-`which python3`}
+echo -e "Using $PYTHON"
+GCNO_FILES=`find $ROOT -name "*.gcno"`
+$GCOV --preserve-paths --relative-only --no-output $GCNO_FILES 2>/dev/null |
+  # Parse the raw gcov report to more human readable form.
+  $PYTHON $ROOT/coverage/parse_gcov_output.py |
+  # Write the output to both stdout and report file.
+  tee $COVERAGE_DIR/coverage_report_all.txt &&
+echo -e "Generated coverage report for all files: $COVERAGE_DIR/coverage_report_all.txt\n"
+
+# TODO: we also need to get the files of the latest commits.
+# Get the most recently committed files.
+LATEST_FILES=`
+  git show --pretty="format:" --name-only HEAD |
+  grep -v "^$" |
+  paste -s -d,`
+RECENT_REPORT=$COVERAGE_DIR/coverage_report_recent.txt
+
+echo -e "Recently updated files: $LATEST_FILES\n" > $RECENT_REPORT
+$GCOV --preserve-paths --relative-only --no-output $GCNO_FILES 2>/dev/null |
+  $PYTHON $ROOT/coverage/parse_gcov_output.py -interested-files $LATEST_FILES |
+  tee -a $RECENT_REPORT &&
+echo -e "Generated coverage report for recently updated files: $RECENT_REPORT\n"
+
+# Unless otherwise specified, we'll not generate html report by default
+if [ -z "$HTML" ]; then
+  exit 0
+fi
+
+# Generate the html report. If we cannot find lcov in this machine, we'll simply
+# skip this step.
+echo "Generating the html coverage report..."
+
+LCOV=$(which lcov || true 2>/dev/null)
+if [ -z $LCOV ]
+then
+  echo "Skip: Cannot find lcov to generate the html report."
+  exit 0
+fi
+
+LCOV_VERSION=$(lcov -v | grep 1.1 || true)
+if [ $LCOV_VERSION ]
+then
+  echo "Not supported lcov version. Expect lcov 1.1."
+  exit 0
+fi
+
+(cd $ROOT; lcov --no-external \
+     --capture  \
+     --directory $PWD \
+     --gcov-tool $GCOV \
+     --output-file $COVERAGE_DIR/coverage.info)
+
+genhtml $COVERAGE_DIR/coverage.info -o $COVERAGE_DIR
+
+echo "HTML Coverage report is generated in $COVERAGE_DIR"
diff --git a/src/rocksdb/coverage/parse_gcov_output.py b/src/rocksdb/coverage/parse_gcov_output.py
new file mode 100644
index 000000000..b9788ec81
--- /dev/null
+++ b/src/rocksdb/coverage/parse_gcov_output.py
@@ -0,0 +1,128 @@
+#!/usr/bin/env python
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+
+from __future__ import print_function
+
+import optparse
+import re
+import sys
+
+# the gcov report follows certain pattern. Each file will have two lines
+# of report, from which we can extract the file name, total lines and coverage
+# percentage.
+def parse_gcov_report(gcov_input):
+    per_file_coverage = {}
+    total_coverage = None
+
+    for line in sys.stdin:
+        line = line.strip()
+
+        # --First line of the coverage report (with file name in it)?
+        match_obj = re.match("^File '(.*)'$", line)
+        if match_obj:
+            # fetch the file name from the first line of the report.
+            current_file = match_obj.group(1)
+            continue
+
+        # -- Second line of the file report (with coverage percentage)
+        match_obj = re.match("^Lines executed:(.*)% of (.*)", line)
+
+        if match_obj:
+            coverage = float(match_obj.group(1))
+            lines = int(match_obj.group(2))
+
+            if current_file is not None:
+                per_file_coverage[current_file] = (coverage, lines)
+                current_file = None
+            else:
+                # If current_file is not set, we reach the last line of report,
+                # which contains the summarized coverage percentage.
+                total_coverage = (coverage, lines)
+            continue
+
+        # If the line's pattern doesn't fall into the above categories. We
+        # can simply ignore them since they're either empty line or doesn't
+        # find executable lines of the given file.
+        current_file = None
+
+    return per_file_coverage, total_coverage
+
+
+def get_option_parser():
+    usage = (
+        "Parse the gcov output and generate more human-readable code "
+        + "coverage report."
+    )
+    parser = optparse.OptionParser(usage)
+
+    parser.add_option(
+        "--interested-files",
+        "-i",
+        dest="filenames",
+        help="Comma separated files names. if specified, we will display "
+        + "the coverage report only for interested source files. "
+        + "Otherwise we will display the coverage report for all "
+        + "source files.",
+    )
+    return parser
+
+
+def display_file_coverage(per_file_coverage, total_coverage):
+    # To print out auto-adjustable column, we need to know the longest
+    # length of file names.
+    max_file_name_length = max(len(fname) for fname in per_file_coverage.keys())
+
+    # -- Print header
+    # size of separator is determined by 3 column sizes:
+    # file name, coverage percentage and lines.
+    header_template = "%" + str(max_file_name_length) + "s\t%s\t%s"
+    separator = "-" * (max_file_name_length + 10 + 20)
+    print(
+        header_template % ("Filename", "Coverage", "Lines")
+    )  # noqa: E999 T25377293 Grandfathered in
+    print(separator)
+
+    # -- Print body
+    # template for printing coverage report for each file.
+    record_template = "%" + str(max_file_name_length) + "s\t%5.2f%%\t%10d"
+
+    for fname, coverage_info in per_file_coverage.items():
+        coverage, lines = coverage_info
+        print(record_template % (fname, coverage, lines))
+
+    # -- Print footer
+    if total_coverage:
+        print(separator)
+        print(record_template % ("Total", total_coverage[0], total_coverage[1]))
+
+
+def report_coverage():
+    parser = get_option_parser()
+    (options, args) = parser.parse_args()
+
+    interested_files = set()
+    if options.filenames is not None:
+        interested_files = {f.strip() for f in options.filenames.split(",")}
+
+    # To make things simple, right now we only read gcov report from the input
+    per_file_coverage, total_coverage = parse_gcov_report(sys.stdin)
+
+    # Check if we need to display coverage info for interested files.
+    if len(interested_files):
+        per_file_coverage = dict(
+            (fname, per_file_coverage[fname])
+            for fname in interested_files
+            if fname in per_file_coverage
+        )
+        # If we only interested in several files, it makes no sense to report
+        # the total_coverage
+        total_coverage = None
+
+    if not len(per_file_coverage):
+        print("Cannot find coverage info for the given files.", file=sys.stderr)
+        return
+    display_file_coverage(per_file_coverage, total_coverage)
+
+
+if __name__ == "__main__":
+    report_coverage()