json/test/thirdparty/fastcov/fastcov.py

372 lines
16 KiB
Python
Raw Normal View History

2019-03-30 08:12:32 +00:00
#!/usr/bin/env python3
"""
Author: Bryan Gillespie
A massively parallel gcov wrapper for generating intermediate coverage formats fast
The goal of fastcov is to generate code coverage intermediate formats as fast as possible
(ideally < 1 second), even for large projects with hundreds of gcda objects. The intermediate
formats may then be consumed by a report generator such as lcov's genhtml, or a dedicated front
end such as coveralls.
Sample Usage:
$ cd build_dir
$ ./fastcov.py --zerocounters
$ <run unit tests>
2019-04-04 07:50:27 +00:00
$ ./fastcov.py --exclude /usr/include test/ --lcov -o report.info
2019-03-30 08:12:32 +00:00
$ genhtml -o code_coverage report.info
"""
import re
import os
import sys
import glob
import json
2019-04-05 22:24:19 +00:00
import time
2019-03-30 08:12:32 +00:00
import argparse
import threading
import subprocess
import multiprocessing
MINIMUM_GCOV = (9,0,0)
2019-04-05 22:24:19 +00:00
MINIMUM_CHUNK_SIZE = 5
2019-03-30 08:12:32 +00:00
# Interesting metrics
2019-04-05 22:24:19 +00:00
START_TIME = time.time()
2019-03-30 08:12:32 +00:00
GCOVS_TOTAL = []
GCOVS_SKIPPED = []
def chunks(l, n):
"""Yield successive n-sized chunks from l."""
for i in range(0, len(l), n):
yield l[i:i + n]
2019-04-05 22:24:19 +00:00
def stopwatch():
"""Return number of seconds since last time this was called"""
global START_TIME
end_time = time.time()
delta = end_time - START_TIME
START_TIME = end_time
return delta
def parseVersionFromLine(version_str):
"""Given a string containing a dotted integer version, parse out integers and return as tuple"""
version = re.search(r'(\d+\.\d+\.\d+)[^\.]', version_str)
if not version:
return (0,0,0)
return tuple(map(int, version.group(1).split(".")))
2019-03-30 08:12:32 +00:00
def getGcovVersion(gcov):
p = subprocess.Popen([gcov, "-v"], stdout=subprocess.PIPE)
output = p.communicate()[0].decode('UTF-8')
p.wait()
2019-04-05 22:24:19 +00:00
return parseVersionFromLine(output.split("\n")[0])
2019-03-30 08:12:32 +00:00
def removeFiles(files):
for file in files:
os.remove(file)
def getFilteredGcdaFiles(gcda_files, exclude):
def excludeGcda(gcda):
for ex in exclude:
if ex in gcda:
return False
return True
return list(filter(excludeGcda, gcda_files))
def getGcdaFiles(cwd, gcda_files):
if not gcda_files:
2019-04-05 05:54:46 +00:00
gcda_files = glob.glob(os.path.join(os.path.abspath(cwd), "**/*.gcda"), recursive=True)
2019-03-30 08:12:32 +00:00
return gcda_files
2019-04-04 07:50:27 +00:00
def gcovWorker(cwd, gcov, files, chunk, gcov_filter_options, branch_coverage):
gcov_args = "-it"
if branch_coverage:
gcov_args += "b"
p = subprocess.Popen([gcov, gcov_args] + chunk, cwd=cwd, stdout=subprocess.PIPE, stderr=subprocess.DEVNULL)
2019-03-30 08:12:32 +00:00
for line in iter(p.stdout.readline, b''):
intermediate_json = json.loads(line.decode(sys.stdout.encoding))
2019-04-05 22:24:19 +00:00
intermediate_json_files = processGcovs(cwd, intermediate_json["files"], gcov_filter_options)
2019-03-30 08:12:32 +00:00
for f in intermediate_json_files:
files.append(f) #thread safe, there might be a better way to do this though
GCOVS_TOTAL.append(len(intermediate_json["files"]))
GCOVS_SKIPPED.append(len(intermediate_json["files"])-len(intermediate_json_files))
p.wait()
2019-04-04 07:50:27 +00:00
def processGcdas(cwd, gcov, jobs, gcda_files, gcov_filter_options, branch_coverage):
2019-03-30 08:12:32 +00:00
chunk_size = max(MINIMUM_CHUNK_SIZE, int(len(gcda_files) / jobs) + 1)
threads = []
intermediate_json_files = []
for chunk in chunks(gcda_files, chunk_size):
2019-04-04 07:50:27 +00:00
t = threading.Thread(target=gcovWorker, args=(cwd, gcov, intermediate_json_files, chunk, gcov_filter_options, branch_coverage))
2019-03-30 08:12:32 +00:00
threads.append(t)
t.start()
2019-04-05 22:24:19 +00:00
log("Spawned %d gcov threads, each processing at most %d gcda files" % (len(threads), chunk_size))
2019-03-30 08:12:32 +00:00
for t in threads:
t.join()
return intermediate_json_files
2019-04-05 22:24:19 +00:00
def processGcov(cwd, gcov, files, gcov_filter_options):
# Add absolute path
gcov["file_abs"] = os.path.abspath(os.path.join(cwd, gcov["file"]))
2019-04-04 07:50:27 +00:00
# If explicit sources were passed, check for match
if gcov_filter_options["sources"]:
2019-04-05 22:24:19 +00:00
if gcov["file_abs"] in gcov_filter_options["sources"]:
2019-04-04 07:50:27 +00:00
files.append(gcov)
return
# Check include filter
if gcov_filter_options["include"]:
for ex in gcov_filter_options["include"]:
if ex in gcov["file"]:
files.append(gcov)
break
return
# Check exclude filter
for ex in gcov_filter_options["exclude"]:
2019-03-30 08:12:32 +00:00
if ex in gcov["file"]:
return
2019-04-04 07:50:27 +00:00
2019-03-30 08:12:32 +00:00
files.append(gcov)
2019-04-05 22:24:19 +00:00
def processGcovs(cwd, gcov_files, gcov_filter_options):
2019-03-30 08:12:32 +00:00
files = []
for gcov in gcov_files:
2019-04-05 22:24:19 +00:00
processGcov(cwd, gcov, files, gcov_filter_options)
2019-03-30 08:12:32 +00:00
return files
2019-04-05 22:24:19 +00:00
def dumpBranchCoverageToLcovInfo(f, branches):
2019-04-04 07:50:27 +00:00
branch_miss = 0
2019-04-05 22:24:19 +00:00
for line_num, branch_counts in branches.items():
for i, count in enumerate(branch_counts):
2019-04-04 07:50:27 +00:00
#Branch (<line number>, <block number>, <branch number>, <taken>)
2019-04-05 22:24:19 +00:00
f.write("BRDA:%s,%d,%d,%d\n" % (line_num, int(i/2), i, count))
branch_miss += int(count == 0)
f.write("BRF:%d\n" % len(branches)) #Branches Found
f.write("BRH:%d\n" % (len(branches) - branch_miss)) #Branches Hit
2019-04-04 07:50:27 +00:00
2019-04-05 22:24:19 +00:00
def dumpToLcovInfo(fastcov_json, output):
2019-03-30 08:12:32 +00:00
with open(output, "w") as f:
2019-04-05 22:24:19 +00:00
for sf, data in fastcov_json["sources"].items():
2019-04-04 07:50:27 +00:00
f.write("SF:%s\n" % sf) #Source File
2019-03-30 08:12:32 +00:00
fn_miss = 0
2019-04-05 22:24:19 +00:00
for function, fdata in data["functions"].items():
f.write("FN:%d,%s\n" % (fdata["start_line"], function)) #Function Start Line
f.write("FNDA:%d,%s\n" % (fdata["execution_count"], function)) #Function Hits
fn_miss += int(fdata["execution_count"] == 0)
f.write("FNF:%d\n" % len(data["functions"])) #Functions Found
f.write("FNH:%d\n" % (len(data["functions"]) - fn_miss)) #Functions Hit
2019-04-04 07:50:27 +00:00
2019-04-05 22:24:19 +00:00
if data["branches"]:
dumpBranchCoverageToLcovInfo(f, data["branches"])
2019-04-04 07:50:27 +00:00
2019-03-30 08:12:32 +00:00
line_miss = 0
2019-04-05 22:24:19 +00:00
for line_num, count in data["lines"].items():
f.write("DA:%s,%d\n" % (line_num, count)) #Line
line_miss += int(count == 0)
f.write("LF:%d\n" % len(data["lines"])) #Lines Found
f.write("LH:%d\n" % (len(data["lines"]) - line_miss)) #Lines Hit
2019-03-30 08:12:32 +00:00
f.write("end_of_record\n")
2019-04-05 22:24:19 +00:00
def exclMarkerWorker(fastcov_sources, chunk):
for source in chunk:
# If there are no covered lines, skip
if not fastcov_sources[source]["lines"]:
continue
start_line = 0
end_line = 0
with open(source) as f:
for i, line in enumerate(f, 1): #Start enumeration at line 1
if not "LCOV_EXCL" in line:
continue
if "LCOV_EXCL_LINE" in line:
if str(i) in fastcov_sources[source]["lines"]:
del fastcov_sources[source]["lines"][str(i)]
if str(i) in fastcov_sources[source]["branches"]:
del fastcov_sources[source]["branches"][str(i)]
elif "LCOV_EXCL_START" in line:
start_line = i
elif "LCOV_EXCL_STOP" in line:
end_line = i
if not start_line:
end_line = 0
continue
for key in ["lines", "branches"]:
for line_num in list(fastcov_sources[source][key].keys()):
if int(line_num) <= end_line and int(line_num) >= start_line:
del fastcov_sources[source][key][line_num]
start_line = end_line = 0
def scanExclusionMarkers(fastcov_json, jobs):
chunk_size = max(MINIMUM_CHUNK_SIZE, int(len(fastcov_json["sources"]) / jobs) + 1)
threads = []
for chunk in chunks(list(fastcov_json["sources"].keys()), chunk_size):
t = threading.Thread(target=exclMarkerWorker, args=(fastcov_json["sources"], chunk))
threads.append(t)
t.start()
log("Spawned %d threads each scanning at most %d source files" % (len(threads), chunk_size))
for t in threads:
t.join()
def distillFunction(function_raw, functions):
function_name = function_raw["name"]
if function_name not in functions:
functions[function_name] = {
"start_line": function_raw["start_line"],
"execution_count": function_raw["execution_count"]
}
else:
functions[function_name]["execution_count"] += function_raw["execution_count"]
def distillLine(line_raw, lines, branches):
line_number = str(line_raw["line_number"])
if line_number not in lines:
lines[line_number] = line_raw["count"]
else:
lines[line_number] += line_raw["count"]
for i, branch in enumerate(line_raw["branches"]):
if line_number not in branches:
branches[line_number] = []
blen = len(branches[line_number])
glen = len(line_raw["branches"])
if blen < glen:
branches[line_number] += [0] * (glen - blen)
branches[line_number][i] += branch["count"]
def distillSource(source_raw, sources):
source_name = source_raw["file_abs"]
if source_name not in sources:
sources[source_name] = {
"functions": {},
"branches": {},
"lines": {},
}
for function in source_raw["functions"]:
distillFunction(function, sources[source_name]["functions"])
for line in source_raw["lines"]:
distillLine(line, sources[source_name]["lines"], sources[source_name]["branches"])
def distillReport(report_raw):
report_json = {
"sources": {}
}
for source in report_raw:
distillSource(source, report_json["sources"])
return report_json
def dumpToJson(intermediate, output):
2019-03-30 08:12:32 +00:00
with open(output, "w") as f:
json.dump(intermediate, f)
def log(line):
if not args.quiet:
2019-04-05 22:24:19 +00:00
print("[{:.3f}s] {}".format(stopwatch(), line))
2019-03-30 08:12:32 +00:00
2019-04-04 07:50:27 +00:00
def getGcovFilterOptions(args):
return {
2019-04-05 22:24:19 +00:00
"sources": set([os.path.abspath(s) for s in args.sources]), #Make paths absolute, use set for fast lookups
2019-04-04 07:50:27 +00:00
"include": args.includepost,
"exclude": args.excludepost,
}
2019-03-30 08:12:32 +00:00
def main(args):
# Need at least gcov 9.0.0 because that's when gcov JSON and stdout streaming was introduced
current_gcov_version = getGcovVersion(args.gcov)
if current_gcov_version < MINIMUM_GCOV:
sys.stderr.write("Minimum gcov version {} required, found {}\n".format(".".join(map(str, MINIMUM_GCOV)), ".".join(map(str, current_gcov_version))))
exit(1)
2019-04-05 22:24:19 +00:00
# Get list of gcda files to process
2019-03-30 08:12:32 +00:00
gcda_files = getGcdaFiles(args.directory, args.gcda_files)
2019-04-05 22:24:19 +00:00
log("Found {} .gcda files ".format(len(gcda_files)))
2019-03-30 08:12:32 +00:00
2019-04-05 22:24:19 +00:00
# If gcda filtering is enabled, filter them out now
2019-03-30 08:12:32 +00:00
if args.excludepre:
gcda_files = getFilteredGcdaFiles(gcda_files, args.excludepre)
2019-04-05 22:24:19 +00:00
log("{} .gcda files after filtering".format(len(gcda_files)))
2019-03-30 08:12:32 +00:00
# We "zero" the "counters" by simply deleting all gcda files
if args.zerocounters:
removeFiles(gcda_files)
2019-04-05 22:24:19 +00:00
log("{} .gcda files removed".format(len(gcda_files)))
2019-03-30 08:12:32 +00:00
return
2019-04-05 22:24:19 +00:00
# Fire up one gcov per cpu and start processing gcdas
2019-04-04 07:50:27 +00:00
gcov_filter_options = getGcovFilterOptions(args)
intermediate_json_files = processGcdas(args.cdirectory, args.gcov, args.jobs, gcda_files, gcov_filter_options, args.branchcoverage)
2019-03-30 08:12:32 +00:00
2019-04-05 22:24:19 +00:00
# Summarize processing results
2019-03-30 08:12:32 +00:00
gcov_total = sum(GCOVS_TOTAL)
gcov_skipped = sum(GCOVS_SKIPPED)
2019-04-05 22:24:19 +00:00
log("Processed {} .gcov files ({} total, {} skipped)".format(gcov_total - gcov_skipped, gcov_total, gcov_skipped))
# Distill all the extraneous info gcov gives us down to the core report
fastcov_json = distillReport(intermediate_json_files)
log("Aggregated raw gcov JSON into fastcov JSON report")
2019-03-30 08:12:32 +00:00
2019-04-05 22:24:19 +00:00
# Dump to desired file format
2019-03-30 08:12:32 +00:00
if args.lcov:
2019-04-05 22:24:19 +00:00
scanExclusionMarkers(fastcov_json, args.jobs)
log("Scanned {} source files for exclusion markers".format(len(fastcov_json["sources"])))
dumpToLcovInfo(fastcov_json, args.output)
log("Created lcov info file '{}'".format(args.output))
elif args.gcov_raw:
dumpToJson(intermediate_json_files, args.output)
log("Created gcov raw json file '{}'".format(args.output))
2019-03-30 08:12:32 +00:00
else:
2019-04-05 22:24:19 +00:00
dumpToJson(fastcov_json, args.output)
log("Created fastcov json file '{}'".format(args.output))
2019-03-30 08:12:32 +00:00
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='A parallel gcov wrapper for fast coverage report generation')
parser.add_argument('-z', '--zerocounters', dest='zerocounters', action="store_true", help='Recursively delete all gcda files')
2019-04-04 07:50:27 +00:00
# Enable Branch Coverage
parser.add_argument('-b', '--branch-coverage', dest='branchcoverage', action="store_true", help='Include branch counts in the coverage report')
# Filtering Options
2019-04-05 22:24:19 +00:00
parser.add_argument('-s', '--source-files', dest='sources', nargs="+", metavar='', default=[], help='Filter: Specify exactly which source files should be included in the final report. Paths must be either absolute or relative to current directory.')
parser.add_argument('-e', '--exclude', dest='excludepost', nargs="+", metavar='', default=[], help='Filter: Exclude source files from final report if they contain one of the provided substrings (i.e. /usr/include test/, etc.)')
parser.add_argument('-i', '--include', dest='includepost', nargs="+", metavar='', default=[], help='Filter: Only include source files in final report that contain one of the provided substrings (i.e. src/ etc.)')
parser.add_argument('-f', '--gcda-files', dest='gcda_files', nargs="+", metavar='', default=[], help='Filter: Specify exactly which gcda files should be processed instead of recursively searching the search directory.')
parser.add_argument('-E', '--exclude-gcda', dest='excludepre', nargs="+", metavar='', default=[], help='Filter: Exclude gcda files from being processed via simple find matching (not regex)')
2019-03-30 08:12:32 +00:00
2019-04-04 07:50:27 +00:00
parser.add_argument('-g', '--gcov', dest='gcov', default='gcov', help='Which gcov binary to use')
2019-03-30 08:12:32 +00:00
parser.add_argument('-d', '--search-directory', dest='directory', default=".", help='Base directory to recursively search for gcda files (default: .)')
2019-04-05 22:24:19 +00:00
parser.add_argument('-c', '--compiler-directory', dest='cdirectory', default=".", help='Base directory compiler was invoked from (default: .) \
This needs to be set if invoking fastcov from somewhere other than the base compiler directory.')
2019-03-30 08:12:32 +00:00
parser.add_argument('-j', '--jobs', dest='jobs', type=int, default=multiprocessing.cpu_count(), help='Number of parallel gcov to spawn (default: %d).' % multiprocessing.cpu_count())
2019-04-05 22:24:19 +00:00
parser.add_argument('-m', '--minimum-chunk-size', dest='minimum_chunk', type=int, default=5, help='Minimum number of files a thread should process (default: 5). \
If you have only 4 gcda files but they are monstrously huge, you could change this value to a 1 so that each thread will only process 1 gcda. Otherise fastcov will spawn only 1 thread to process all of them.')
2019-03-30 08:12:32 +00:00
2019-04-05 22:24:19 +00:00
parser.add_argument('-l', '--lcov', dest='lcov', action="store_true", help='Output in lcov info format instead of fastcov json')
parser.add_argument('-r', '--gcov-raw', dest='gcov_raw', action="store_true", help='Output in gcov raw json instead of fastcov json')
parser.add_argument('-o', '--output', dest='output', default="coverage.json", help='Name of output file (default: coverage.json)')
2019-03-30 08:12:32 +00:00
parser.add_argument('-q', '--quiet', dest='quiet', action="store_true", help='Suppress output to stdout')
2019-04-05 22:24:19 +00:00
2019-03-30 08:12:32 +00:00
args = parser.parse_args()
main(args)