The iglob function, which we use to find C source files in the kernel tree, always follows symbolic links. This can cause unintentional recursions whenever a symbolic link points to a parent directory. A common scenario is building the kernel with the output set to a directory inside the kernel tree, which will contain such a symlink. Instead of using the iglob function, use os.walk to traverse the directory tree, which by default doesn't follow symbolic links. fnmatch is then used to match the glob on the filename, as well as ignore hidden files (which were ignored by default with iglob). This approach runs just as fast as using iglob.
Unified · Split
| @@ -1,8 +1,8 @@ | |||
| 1 | #!/usr/bin/env python3 | 1 | #!/usr/bin/env python3 |
| 2 | # SPDX-License-Identifier: GPL-2.0-only | 2 | # SPDX-License-Identifier: GPL-2.0-only |
| 3 | 3 | ||
| 4 | import fnmatch | ||
| 4 | import os | 5 | import os |
| 5 | import glob | ||
| 6 | import re | 6 | import re |
| 7 | import argparse | 7 | import argparse |
| 8 | 8 | ||
| @@ -81,10 +81,20 @@ def print_compat(filename, compatibles): · 72 unmodified lines | |||
| 81 | else: | 81 | else: |
| 82 | print(*compatibles, sep='\n') | 82 | print(*compatibles, sep='\n') |
| 83 | 83 | ||
| 84 | def glob_without_symlinks(root, glob): | ||
| 85 | for path, dirs, files in os.walk(root): | ||
| 86 | # Ignore hidden directories | ||
| 87 | for d in dirs: | ||
| 88 | if fnmatch.fnmatch(d, ".*"): | ||
| 89 | dirs.remove(d) | ||
| 90 | for f in files: | ||
| 91 | if fnmatch.fnmatch(f, glob): | ||
| 92 | yield os.path.join(path, f) | ||
| 93 | |||
| 84 | def files_to_parse(path_args): | 94 | def files_to_parse(path_args): |
| 85 | for f in path_args: | 95 | for f in path_args: |
| 86 | if os.path.isdir(f): | 96 | if os.path.isdir(f): |
| 87 | for filename in glob.iglob(f + "/**/*.c", recursive=True): | 97 | for filename in glob_without_symlinks(f, "*.c"): |
| 88 | yield filename | 98 | yield filename |
| 89 | else: | 99 | else: |
| 90 | yield f | 100 | yield f |