diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 19:33:14 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 19:33:14 +0000 |
commit | 36d22d82aa202bb199967e9512281e9a53db42c9 (patch) | |
tree | 105e8c98ddea1c1e4784a60a5a6410fa416be2de /tools/lint/clang-format | |
parent | Initial commit. (diff) | |
download | firefox-esr-36d22d82aa202bb199967e9512281e9a53db42c9.tar.xz firefox-esr-36d22d82aa202bb199967e9512281e9a53db42c9.zip |
Adding upstream version 115.7.0esr.upstream/115.7.0esr
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'tools/lint/clang-format')
-rw-r--r-- | tools/lint/clang-format/__init__.py | 237 |
1 files changed, 237 insertions, 0 deletions
diff --git a/tools/lint/clang-format/__init__.py b/tools/lint/clang-format/__init__.py new file mode 100644 index 0000000000..d0a81f7b06 --- /dev/null +++ b/tools/lint/clang-format/__init__.py @@ -0,0 +1,237 @@ +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. + +import os +import re +import signal +import sys +import xml.etree.ElementTree as ET + +from mozboot.util import get_tools_dir +from mozlint import result +from mozlint.pathutils import expand_exclusions +from mozprocess import ProcessHandler + +CLANG_FORMAT_NOT_FOUND = """ +Could not find clang-format! It should've been installed automatically - \ +please report a bug here: +https://bugzilla.mozilla.org/enter_bug.cgi?product=Firefox%20Build%20System&component=Lint%20and%20Formatting +""".strip() + + +def setup(root, mach_command_context, **lintargs): + if get_clang_format_binary(): + return 0 + + from mozbuild.code_analysis.mach_commands import get_clang_tools + + rc, _ = get_clang_tools(mach_command_context) + if rc: + return 1 + + +class ClangFormatProcess(ProcessHandler): + def __init__(self, config, *args, **kwargs): + self.config = config + kwargs["stream"] = False + kwargs["universal_newlines"] = True + ProcessHandler.__init__(self, *args, **kwargs) + + def run(self, *args, **kwargs): + orig = signal.signal(signal.SIGINT, signal.SIG_IGN) + ProcessHandler.run(self, *args, **kwargs) + signal.signal(signal.SIGINT, orig) + + +def run_process(config, cmd): + proc = ClangFormatProcess(config, cmd) + proc.run() + try: + proc.wait() + except KeyboardInterrupt: + proc.kill() + + return proc.output + + +def get_clang_format_binary(): + """ + Returns the path of the first clang-format binary available + if not found returns None + """ + binary = os.environ.get("CLANG_FORMAT") + if binary: + return binary + + clang_tools_path = os.path.join(get_tools_dir(), "clang-tools") + bin_path = os.path.join(clang_tools_path, "clang-tidy", "bin") + binary = os.path.join(bin_path, "clang-format") + + if sys.platform.startswith("win"): + binary += ".exe" + + if not os.path.isfile(binary): + return None + + return binary + + +def is_ignored_path(ignored_dir_re, topsrcdir, f): + # Remove up to topsrcdir in pathname and match + if f.startswith(topsrcdir + "/"): + match_f = f[len(topsrcdir + "/") :] + else: + match_f = f + return re.match(ignored_dir_re, match_f) + + +def remove_ignored_path(paths, topsrcdir, log): + path_to_third_party = os.path.join(topsrcdir, ".clang-format-ignore") + + ignored_dir = [] + with open(path_to_third_party, "r") as fh: + for line in fh: + # In case it starts with a space + line = line.strip() + # Remove comments and empty lines + if line.startswith("#") or len(line) == 0: + continue + # The regexp is to make sure we are managing relative paths + ignored_dir.append(r"^[\./]*" + line.rstrip()) + + # Generates the list of regexp + ignored_dir_re = "(%s)" % "|".join(ignored_dir) + + path_list = [] + for f in paths: + if is_ignored_path(ignored_dir_re, topsrcdir, f): + # Early exit if we have provided an ignored directory + log.debug("Ignored third party code '{0}'".format(f)) + continue + path_list.append(f) + + return path_list + + +def lint(paths, config, fix=None, **lintargs): + log = lintargs["log"] + paths = list(expand_exclusions(paths, config, lintargs["root"])) + + # We ignored some specific files for a bunch of reasons. + # Not using excluding to avoid duplication + if lintargs.get("use_filters", True): + paths = remove_ignored_path(paths, lintargs["root"], log) + + # An empty path array can occur when the user passes in `-n`. If we don't + # return early in this case, rustfmt will attempt to read stdin and hang. + if not paths: + return [] + + binary = get_clang_format_binary() + + if not binary: + print(CLANG_FORMAT_NOT_FOUND) + if "MOZ_AUTOMATION" in os.environ: + return 1 + return [] + + cmd_args = [binary] + + base_command = cmd_args + ["--version"] + version = run_process(config, base_command) + log.debug("Version: {}".format(version)) + + cmd_args.append("--output-replacements-xml") + base_command = cmd_args + paths + log.debug("Command: {}".format(" ".join(cmd_args))) + output = run_process(config, base_command) + + def replacement(parser): + for end, e in parser.read_events(): + assert end == "end" + if e.tag == "replacement": + item = {k: int(v) for k, v in e.items()} + assert sorted(item.keys()) == ["length", "offset"] + item["with"] = (e.text or "").encode("utf-8") + yield item + + # When given multiple paths as input, --output-replacements-xml + # will output one xml per path, in the order they are given, but + # XML parsers don't know how to handle that, so do it manually. + parser = None + replacements = [] + for line in output: + if line.startswith("<?xml "): + if parser: + replacements.append(list(replacement(parser))) + parser = ET.XMLPullParser(["end"]) + parser.feed(line) + replacements.append(list(replacement(parser))) + + results = [] + fixed = 0 + for path, replacement in zip(paths, replacements): + if not replacement: + continue + with open(path, "rb") as fh: + data = fh.read() + + linenos = [] + patched_data = b"" + last_offset = 0 + lineno_before = 1 + lineno_after = 1 + + for item in replacement: + offset = item["offset"] + length = item["length"] + replace_with = item["with"] + since_last_offset = data[last_offset:offset] + replaced = data[offset : offset + length] + + lines_since_last_offset = since_last_offset.count(b"\n") + lineno_before += lines_since_last_offset + lineno_after += lines_since_last_offset + start_lineno = (lineno_before, lineno_after) + + lineno_before += replaced.count(b"\n") + lineno_after += replace_with.count(b"\n") + end_lineno = (lineno_before, lineno_after) + + if linenos and start_lineno[0] <= linenos[-1][1][0]: + linenos[-1] = (linenos[-1][0], end_lineno) + else: + linenos.append((start_lineno, end_lineno)) + + patched_data += since_last_offset + replace_with + last_offset = offset + len(replaced) + patched_data += data[last_offset:] + + lines_before = data.decode("utf-8", "replace").splitlines() + lines_after = patched_data.decode("utf-8", "replace").splitlines() + for (start_before, start_after), (end_before, end_after) in linenos: + diff = "".join( + "-" + l + "\n" for l in lines_before[start_before - 1 : end_before] + ) + diff += "".join( + "+" + l + "\n" for l in lines_after[start_after - 1 : end_after] + ) + + results.append( + result.from_config( + config, + path=path, + diff=diff, + level="warning", + lineno=start_before, + column=0, + ) + ) + + if fix: + with open(path, "wb") as fh: + fh.write(patched_data) + fixed += len(linenos) + + return {"results": results, "fixed": fixed} |