poky/scripts/contrib/patchreview.py
Richard Purdie c9303648f8 scripts/patchreview: Make json output human parsable
Sort dict keys in the json output and use tab spacing. This means
when commited into git, the diffs are human readable but it is more
compact filesize than space indentation.

(From OE-Core rev: bde2ecb203d8a1a29715c70ca3ded382982390cf)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
2022-05-23 14:00:38 +01:00

240 lines
8.5 KiB
Python
Executable File

#! /usr/bin/env python3
#
# SPDX-License-Identifier: GPL-2.0-only
#
# TODO
# - option to just list all broken files
# - test suite
# - validate signed-off-by
status_values = ("accepted", "pending", "inappropriate", "backport", "submitted", "denied", "inactive-upstream")
class Result:
# Whether the patch has an Upstream-Status or not
missing_upstream_status = False
# If the Upstream-Status tag is malformed in some way (string for bad bit)
malformed_upstream_status = None
# If the Upstream-Status value is unknown (boolean)
unknown_upstream_status = False
# The upstream status value (Pending, etc)
upstream_status = None
# Whether the patch has a Signed-off-by or not
missing_sob = False
# Whether the Signed-off-by tag is malformed in some way
malformed_sob = False
# The Signed-off-by tag value
sob = None
# Whether a patch looks like a CVE but doesn't have a CVE tag
missing_cve = False
def blame_patch(patch):
"""
From a patch filename, return a list of "commit summary (author name <author
email>)" strings representing the history.
"""
import subprocess
return subprocess.check_output(("git", "log",
"--follow", "--find-renames", "--diff-filter=A",
"--format=%s (%aN <%aE>)",
"--", patch)).decode("utf-8").splitlines()
def patchreview(path, patches):
import re, os.path
# General pattern: start of line, optional whitespace, tag with optional
# hyphen or spaces, maybe a colon, some whitespace, then the value, all case
# insensitive.
sob_re = re.compile(r"^[\t ]*(Signed[-_ ]off[-_ ]by:?)[\t ]*(.+)", re.IGNORECASE | re.MULTILINE)
status_re = re.compile(r"^[\t ]*(Upstream[-_ ]Status:?)[\t ]*([\w-]*)", re.IGNORECASE | re.MULTILINE)
cve_tag_re = re.compile(r"^[\t ]*(CVE:)[\t ]*(.*)", re.IGNORECASE | re.MULTILINE)
cve_re = re.compile(r"cve-[0-9]{4}-[0-9]{4,6}", re.IGNORECASE)
results = {}
for patch in patches:
fullpath = os.path.join(path, patch)
result = Result()
results[fullpath] = result
content = open(fullpath, encoding='ascii', errors='ignore').read()
# Find the Signed-off-by tag
match = sob_re.search(content)
if match:
value = match.group(1)
if value != "Signed-off-by:":
result.malformed_sob = value
result.sob = match.group(2)
else:
result.missing_sob = True
# Find the Upstream-Status tag
match = status_re.search(content)
if match:
value = match.group(1)
if value != "Upstream-Status:":
result.malformed_upstream_status = value
value = match.group(2).lower()
# TODO: check case
if value not in status_values:
result.unknown_upstream_status = True
result.upstream_status = value
else:
result.missing_upstream_status = True
# Check that patches which looks like CVEs have CVE tags
if cve_re.search(patch) or cve_re.search(content):
if not cve_tag_re.search(content):
result.missing_cve = True
# TODO: extract CVE list
return results
def analyse(results, want_blame=False, verbose=True):
"""
want_blame: display blame data for each malformed patch
verbose: display per-file results instead of just summary
"""
# want_blame requires verbose, so disable blame if we're not verbose
if want_blame and not verbose:
want_blame = False
total_patches = 0
missing_sob = 0
malformed_sob = 0
missing_status = 0
malformed_status = 0
missing_cve = 0
pending_patches = 0
for patch in sorted(results):
r = results[patch]
total_patches += 1
need_blame = False
# Build statistics
if r.missing_sob:
missing_sob += 1
if r.malformed_sob:
malformed_sob += 1
if r.missing_upstream_status:
missing_status += 1
if r.malformed_upstream_status or r.unknown_upstream_status:
malformed_status += 1
# Count patches with no status as pending
pending_patches +=1
if r.missing_cve:
missing_cve += 1
if r.upstream_status == "pending":
pending_patches += 1
# Output warnings
if r.missing_sob:
need_blame = True
if verbose:
print("Missing Signed-off-by tag (%s)" % patch)
if r.malformed_sob:
need_blame = True
if verbose:
print("Malformed Signed-off-by '%s' (%s)" % (r.malformed_sob, patch))
if r.missing_cve:
need_blame = True
if verbose:
print("Missing CVE tag (%s)" % patch)
if r.missing_upstream_status:
need_blame = True
if verbose:
print("Missing Upstream-Status tag (%s)" % patch)
if r.malformed_upstream_status:
need_blame = True
if verbose:
print("Malformed Upstream-Status '%s' (%s)" % (r.malformed_upstream_status, patch))
if r.unknown_upstream_status:
need_blame = True
if verbose:
print("Unknown Upstream-Status value '%s' (%s)" % (r.upstream_status, patch))
if want_blame and need_blame:
print("\n".join(blame_patch(patch)) + "\n")
def percent(num):
try:
return "%d (%d%%)" % (num, round(num * 100.0 / total_patches))
except ZeroDivisionError:
return "N/A"
if verbose:
print()
print("""Total patches found: %d
Patches missing Signed-off-by: %s
Patches with malformed Signed-off-by: %s
Patches missing CVE: %s
Patches missing Upstream-Status: %s
Patches with malformed Upstream-Status: %s
Patches in Pending state: %s""" % (total_patches,
percent(missing_sob),
percent(malformed_sob),
percent(missing_cve),
percent(missing_status),
percent(malformed_status),
percent(pending_patches)))
def histogram(results):
from toolz import recipes, dicttoolz
import math
counts = recipes.countby(lambda r: r.upstream_status, results.values())
bars = dicttoolz.valmap(lambda v: "#" * int(math.ceil(float(v) / len(results) * 100)), counts)
for k in bars:
print("%-20s %s (%d)" % (k.capitalize() if k else "No status", bars[k], counts[k]))
if __name__ == "__main__":
import argparse, subprocess, os
args = argparse.ArgumentParser(description="Patch Review Tool")
args.add_argument("-b", "--blame", action="store_true", help="show blame for malformed patches")
args.add_argument("-v", "--verbose", action="store_true", help="show per-patch results")
args.add_argument("-g", "--histogram", action="store_true", help="show patch histogram")
args.add_argument("-j", "--json", help="update JSON")
args.add_argument("directory", help="directory to scan")
args = args.parse_args()
patches = subprocess.check_output(("git", "-C", args.directory, "ls-files", "recipes-*/**/*.patch", "recipes-*/**/*.diff")).decode("utf-8").split()
results = patchreview(args.directory, patches)
analyse(results, want_blame=args.blame, verbose=args.verbose)
if args.json:
import json, os.path, collections
if os.path.isfile(args.json):
data = json.load(open(args.json))
else:
data = []
row = collections.Counter()
row["total"] = len(results)
row["date"] = subprocess.check_output(["git", "-C", args.directory, "show", "-s", "--pretty=format:%cd", "--date=format:%s"]).decode("utf-8").strip()
row["commit"] = subprocess.check_output(["git", "-C", args.directory, "show", "-s", "--pretty=format:%H"]).decode("utf-8").strip()
for r in results.values():
if r.upstream_status in status_values:
row[r.upstream_status] += 1
if r.malformed_upstream_status or r.missing_upstream_status:
row['malformed-upstream-status'] += 1
if r.malformed_sob or r.missing_sob:
row['malformed-sob'] += 1
data.append(row)
json.dump(data, open(args.json, "w"), sort_keys=True, indent="\t")
if args.histogram:
print()
histogram(results)