Attached patches use the new cloexec parameter added by the PEP 433 (see issue #17036). cloexec_fs_walk.patch: [security] don't leak a file descriptors of directories to a child processes cloexec_listening_socket.patch: [security] don't leak a listening socket to child processes, see also #12107 cloexec_log_file.patch: [security] don't leak the file descriptor of a log file to child processes cloexec_subprocess.patch: [security/bugs] don't leak file descriptors to child processes cloexec_misc.patch: misc mmodules "security" is a strong word: if subprocess is called with close_fds=True, there is no such problem at all. It's more a theorical problem if a process is created in another thread without using the subprocess module (but directly low level functions).
