from __future__ import absolute_import import os import re import warnings from pip._internal.utils.typing import MYPY_CHECK_RUNNING if MYPY_CHECK_RUNNING: from typing import Optional, Tuple def glibc_version_string(): # type: () -> Optional[str] "Returns glibc version string, or None if not using glibc." return glibc_version_string_confstr() or glibc_version_string_ctypes() def glibc_version_string_confstr(): # type: () -> Optional[str] "Primary implementation of glibc_version_string using os.confstr." # os.confstr is quite a bit faster than ctypes.DLL. It's also less likely # to be broken or missing. This strategy is used in the standard library # platform module: # https://github.com/python/cpython/blob/fcf1d003bf4f0100c9d0921ff3d70e1127ca1b71/Lib/platform.py#L175-L183 try: # os.confstr("CS_GNU_LIBC_VERSION") returns a string like "glibc 2.17": _, version = os.confstr("CS_GNU_LIBC_VERSION").split() except (AttributeError, OSError, ValueError): # os.confstr() or CS_GNU_LIBC_VERSION not available (or a bad value)... return None return version def glibc_version_string_ctypes(): # type: () -> Optional[str] "Fallback implementation of glibc_version_string using ctypes." try: import ctypes except ImportError: return None # ctypes.CDLL(None) internally calls dlopen(NULL), and as the dlopen # manpage says, "If filename is NULL, then the returned handle is for the # main program". This way we can let the linker do the work to figure out # which libc our process is actually using. process_namespace = ctypes.CDLL(None) try: gnu_get_libc_version = process_namespace.gnu_get_libc_version except AttributeError: # Symbol doesn't exist -> therefore, we are not linked to # glibc. return None # Call gnu_get_libc_version, which returns a string like "2.5" gnu_get_libc_version.restype = ctypes.c_char_p version_str = gnu_get_libc_version() # py2 / py3 compatibility: if not isinstance(version_str, str): version_str = version_str.decode("ascii") return version_str # Separated out from have_compatible_glibc for easier unit testing def check_glibc_version(version_str, required_major, minimum_minor): # type: (str, int, int) -> bool # Parse string and check against requested version. # # We use a regexp instead of str.split because we want to discard any # random junk that might come after the minor version -- this might happen # in patched/forked versions of glibc (e.g. Linaro's version of glibc # uses version strings like "2.20-2014.11"). See gh-3588. m = re.match(r"(?P[0-9]+)\.(?P[0-9]+)", version_str) if not m: warnings.warn("Expected glibc version with 2 components major.minor," " got: %s" % version_str, RuntimeWarning) return False return (int(m.group("major")) == required_major and int(m.group("minor")) >= minimum_minor) def have_compatible_glibc(required_major, minimum_minor): # type: (int, int) -> bool version_str = glibc_version_string() if version_str is None: return False return check_glibc_version(version_str, required_major, minimum_minor) # platform.libc_ver regularly returns completely nonsensical glibc # versions. E.g. on my computer, platform says: # # ~$ python2.7 -c 'import platform; print(platform.libc_ver())' # ('glibc', '2.7') # ~$ python3.5 -c 'import platform; print(platform.libc_ver())' # ('glibc', '2.9') # # But the truth is: # # ~$ ldd --version # ldd (Debian GLIBC 2.22-11) 2.22 # # This is unfortunate, because it means that the linehaul data on libc # versions that was generated by pip 8.1.2 and earlier is useless and # misleading. Solution: instead of using platform, use our code that actually # works. def libc_ver(): # type: () -> Tuple[str, str] """Try to determine the glibc version Returns a tuple of strings (lib, version) which default to empty strings in case the lookup fails. """ glibc_version = glibc_version_string() if glibc_version is None: return ("", "") else: return ("glibc", glibc_version)