Christian Tismer | ea62ce7 | 2018-06-09 20:32:25 +0200 | [diff] [blame] | 1 | """ |
| 2 | pep384_macrocheck.py |
| 3 | |
| 4 | This programm tries to locate errors in the relevant Python header |
| 5 | files where macros access type fields when they are reachable from |
| 6 | the limided API. |
| 7 | |
| 8 | The idea is to search macros with the string "->tp_" in it. |
| 9 | When the macro name does not begin with an underscore, |
| 10 | then we have found a dormant error. |
| 11 | |
| 12 | Christian Tismer |
| 13 | 2018-06-02 |
| 14 | """ |
| 15 | |
| 16 | import sys |
| 17 | import os |
| 18 | import re |
| 19 | |
| 20 | |
| 21 | DEBUG = False |
| 22 | |
| 23 | def dprint(*args, **kw): |
| 24 | if DEBUG: |
| 25 | print(*args, **kw) |
| 26 | |
| 27 | def parse_headerfiles(startpath): |
| 28 | """ |
| 29 | Scan all header files which are reachable fronm Python.h |
| 30 | """ |
| 31 | search = "Python.h" |
| 32 | name = os.path.join(startpath, search) |
| 33 | if not os.path.exists(name): |
| 34 | raise ValueError("file {} was not found in {}\n" |
| 35 | "Please give the path to Python's include directory." |
| 36 | .format(search, startpath)) |
| 37 | errors = 0 |
| 38 | with open(name) as python_h: |
| 39 | while True: |
| 40 | line = python_h.readline() |
| 41 | if not line: |
| 42 | break |
| 43 | found = re.match(r'^\s*#\s*include\s*"(\w+\.h)"', line) |
| 44 | if not found: |
| 45 | continue |
| 46 | include = found.group(1) |
| 47 | dprint("Scanning", include) |
| 48 | name = os.path.join(startpath, include) |
| 49 | if not os.path.exists(name): |
| 50 | name = os.path.join(startpath, "../PC", include) |
| 51 | errors += parse_file(name) |
| 52 | return errors |
| 53 | |
| 54 | def ifdef_level_gen(): |
| 55 | """ |
| 56 | Scan lines for #ifdef and track the level. |
| 57 | """ |
| 58 | level = 0 |
| 59 | ifdef_pattern = r"^\s*#\s*if" # covers ifdef and ifndef as well |
| 60 | endif_pattern = r"^\s*#\s*endif" |
| 61 | while True: |
| 62 | line = yield level |
| 63 | if re.match(ifdef_pattern, line): |
| 64 | level += 1 |
| 65 | elif re.match(endif_pattern, line): |
| 66 | level -= 1 |
| 67 | |
| 68 | def limited_gen(): |
| 69 | """ |
| 70 | Scan lines for Py_LIMITED_API yes(1) no(-1) or nothing (0) |
| 71 | """ |
| 72 | limited = [0] # nothing |
| 73 | unlimited_pattern = r"^\s*#\s*ifndef\s+Py_LIMITED_API" |
| 74 | limited_pattern = "|".join([ |
| 75 | r"^\s*#\s*ifdef\s+Py_LIMITED_API", |
| 76 | r"^\s*#\s*(el)?if\s+!\s*defined\s*\(\s*Py_LIMITED_API\s*\)\s*\|\|", |
| 77 | r"^\s*#\s*(el)?if\s+defined\s*\(\s*Py_LIMITED_API" |
| 78 | ]) |
| 79 | else_pattern = r"^\s*#\s*else" |
| 80 | ifdef_level = ifdef_level_gen() |
| 81 | status = next(ifdef_level) |
| 82 | wait_for = -1 |
| 83 | while True: |
| 84 | line = yield limited[-1] |
| 85 | new_status = ifdef_level.send(line) |
| 86 | dir = new_status - status |
| 87 | status = new_status |
| 88 | if dir == 1: |
| 89 | if re.match(unlimited_pattern, line): |
| 90 | limited.append(-1) |
| 91 | wait_for = status - 1 |
| 92 | elif re.match(limited_pattern, line): |
| 93 | limited.append(1) |
| 94 | wait_for = status - 1 |
| 95 | elif dir == -1: |
| 96 | # this must have been an endif |
| 97 | if status == wait_for: |
| 98 | limited.pop() |
| 99 | wait_for = -1 |
| 100 | else: |
| 101 | # it could be that we have an elif |
| 102 | if re.match(limited_pattern, line): |
| 103 | limited.append(1) |
| 104 | wait_for = status - 1 |
| 105 | elif re.match(else_pattern, line): |
| 106 | limited.append(-limited.pop()) # negate top |
| 107 | |
| 108 | def parse_file(fname): |
| 109 | errors = 0 |
| 110 | with open(fname) as f: |
| 111 | lines = f.readlines() |
| 112 | type_pattern = r"^.*?->\s*tp_" |
| 113 | define_pattern = r"^\s*#\s*define\s+(\w+)" |
| 114 | limited = limited_gen() |
| 115 | status = next(limited) |
| 116 | for nr, line in enumerate(lines): |
| 117 | status = limited.send(line) |
| 118 | line = line.rstrip() |
| 119 | dprint(fname, nr, status, line) |
| 120 | if status != -1: |
| 121 | if re.match(define_pattern, line): |
| 122 | name = re.match(define_pattern, line).group(1) |
| 123 | if not name.startswith("_"): |
| 124 | # found a candidate, check it! |
| 125 | macro = line + "\n" |
| 126 | idx = nr |
| 127 | while line.endswith("\\"): |
| 128 | idx += 1 |
| 129 | line = lines[idx].rstrip() |
| 130 | macro += line + "\n" |
| 131 | if re.match(type_pattern, macro, re.DOTALL): |
| 132 | # this type field can reach the limited API |
| 133 | report(fname, nr + 1, macro) |
| 134 | errors += 1 |
| 135 | return errors |
| 136 | |
| 137 | def report(fname, nr, macro): |
| 138 | f = sys.stderr |
| 139 | print(fname + ":" + str(nr), file=f) |
| 140 | print(macro, file=f) |
| 141 | |
| 142 | if __name__ == "__main__": |
| 143 | p = sys.argv[1] if sys.argv[1:] else "../../Include" |
| 144 | errors = parse_headerfiles(p) |
| 145 | if errors: |
| 146 | # somehow it makes sense to raise a TypeError :-) |
| 147 | raise TypeError("These {} locations contradict the limited API." |
| 148 | .format(errors)) |