On 15.07.2017 00:38, Mark Wooding wrote: > Package: python2.7 > Version: 2.7.13-2 > Severity: normal > > [stratocaster ~]nm --dynamic /usr/bin/python2.7 | grep md5 > 00091732 T init_md5 > 001d6890 T md5_append > 001d67d0 T md5_finish > 000919f3 T md5_init > [stratocaster ~] > > This causes trouble when trying to use bindings to a crypto library > which also wants to define `md5_init', because they disagree about how > the hashing state ought to be arranged: the outcome is a wrong answer on > amd64, and a segfault on i386. While it can certainly be argued that > neither Python nor the other library has especially good taste in > naming, `md5_init' is at least a proper, documented part of the > library's public interface, directly relevant to its purpose (and has > been since 1999), though the same can't be argued for Python: > > [stratocaster ~]grep -ri md5 /usr/include/python2.7/ > [stratocaster ~] > > Indeed, it seems that the Python package has leaked its MD5 symbols > before: see bug #440272. > > I'm deploying an awful hack involving `RTLD_DEEPBIND' in the meantime, > but it would be nice to get this fixed properly.
python uses <module_name>_init for extension modules / builtins to initilize these. So it's part of the ABI as well. Not sure how to change that.
