121 lines · python
1#!/usr/bin/env python2# ===----------------------------------------------------------------------===##3#4# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.5# See https://llvm.org/LICENSE.txt for license information.6# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception7#8# ===----------------------------------------------------------------------===##9 10from typing import List, Tuple, Optional11import argparse12import io13import itertools14import os15import pathlib16import re17import sys18 19libcxx_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))20sys.path.append(os.path.join(libcxx_root, "utils"))21from libcxx.header_information import Header22 23def parse_line(line: str) -> Tuple[int, str]:24 """25 Parse a single line of --trace-includes output.26 27 Returns the inclusion level and the raw file name being included.28 """29 match = re.match(r"(\.+) (.+)", line)30 if not match:31 raise ArgumentError(f"Line {line} contains invalid data.")32 33 # The number of periods in front of the header name is the nesting level of34 # that header.35 return (len(match.group(1)), match.group(2))36 37def make_cxx_v1_relative(header: str) -> Optional[str]:38 """39 Returns the path of the header as relative to <whatever>/c++/v1, or None if the path40 doesn't contain c++/v1.41 42 We use that heuristic to figure out which headers are libc++ headers.43 """44 # On Windows, the path separators can either be forward slash or backslash.45 # If it is a backslash, Clang prints it escaped as two consecutive46 # backslashes, and they need to be escaped in the RE. (Use a raw string for47 # the pattern to avoid needing another level of escaping on the Python string48 # literal level.)49 pathsep = r"(?:/|\\\\)"50 CXX_V1_REGEX = r"^.*c\+\+" + pathsep + r"v[0-9]+" + pathsep + r"(.+)$"51 match = re.match(CXX_V1_REGEX, header)52 if not match:53 return None54 else:55 return match.group(1)56 57def parse_file(file: io.TextIOBase) -> List[Tuple[Header, Header]]:58 """59 Parse a file containing --trace-includes output to generate a list of the60 transitive includes contained in it.61 """62 result = []63 includer = None64 for line in file.readlines():65 (level, header) = parse_line(line)66 relative = make_cxx_v1_relative(header)67 68 # Not a libc++ header69 if relative is None:70 continue71 72 # If we're at the first level, remember this header as being the one who includes other headers.73 # There's usually exactly one, except if the compiler is passed a file with `-include`.74 if level == 1:75 includer = Header(relative)76 continue77 78 # Otherwise, take note that this header is being included by the top-level includer.79 else:80 assert includer is not None81 result.append((includer, Header(relative)))82 return result83 84def print_csv(includes: List[Tuple[Header, Header]]) -> None:85 """86 Print the transitive includes as space-delimited CSV.87 88 This function only prints public libc++ headers that are not C compatibility headers.89 """90 # Sort and group by includer91 by_includer = lambda t: t[0]92 includes = itertools.groupby(sorted(includes, key=by_includer), key=by_includer)93 94 for (includer, includees) in includes:95 includees = map(lambda t: t[1], includees)96 for h in sorted(set(includees)):97 if h.is_public() and not h.is_C_compatibility():98 print(f"{includer} {h}")99 100def main(argv):101 parser = argparse.ArgumentParser(102 description="""103 Given a list of headers produced by --trace-includes, produce a list of libc++ headers in that output.104 105 Note that -fshow-skipped-includes must also be passed to the compiler in order to get sufficient106 information for this script to run.107 108 The output of this script is provided in space-delimited CSV format where each line contains:109 110 <header performing inclusion> <header being included>111 """)112 parser.add_argument("inputs", type=argparse.FileType("r"), nargs='+', default=None,113 help="One or more files containing the result of --trace-includes")114 args = parser.parse_args(argv)115 116 includes = [line for file in args.inputs for line in parse_file(file)]117 print_csv(includes)118 119if __name__ == "__main__":120 main(sys.argv[1:])121