[session] use a dedicated class to track cnxset
We introduce a new CnxSetTracker to track `cnxset` used by Transaction and
allows to wait for them. This new class does not use Thread ID not thread
joining to work. This allows to use multiple transaction per thread and a
transaction in multiple thread.
The class itself is totally threadsafe by the Transaction is still not
thread safe.
The old _threads_in_transaction attribute is dropped in favor of a new logic
based on this object. The registration of cnxset used is not done by the
Transaction itself. tx.cnset is a property handling the Consistency of its value
with the CnxSetTracker instance.
Note: The CnxSetTracker instance only track transaction id, not transaction
itself, So not reference cycle are created.
# md5crypt.py
#
# 0423.2000 by michal wallace http://www.sabren.com/
# based on perl's Crypt::PasswdMD5 by Luis Munoz (lem@cantv.net)
# based on /usr/src/libcrypt/crypt.c from FreeBSD 2.2.5-RELEASE
#
# MANY THANKS TO
#
# Carey Evans - http://home.clear.net.nz/pages/c.evans/
# Dennis Marti - http://users.starpower.net/marti1/
#
# For the patches that got this thing working!
#
# modification by logilab:
# * remove usage of the string module
# * don't include the magic string in the output string
# for true crypt.crypt compatibility
# * use hashlib module instead of md5
#########################################################
"""md5crypt.py - Provides interoperable MD5-based crypt() function
SYNOPSIS
import md5crypt.py
cryptedpassword = md5crypt.md5crypt(password, salt);
DESCRIPTION
unix_md5_crypt() provides a crypt()-compatible interface to the
rather new MD5-based crypt() function found in modern operating systems.
It's based on the implementation found on FreeBSD 2.2.[56]-RELEASE and
contains the following license in it:
"THE BEER-WARE LICENSE" (Revision 42):
<phk@login.dknet.dk> wrote this file. As long as you retain this notice you
can do whatever you want with this stuff. If we meet some day, and you think
this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
"""
MAGIC = '$1$' # Magic string
ITOA64 = "./0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
from hashlib import md5 # pylint: disable=E0611
def to64 (v, n):
ret = ''
while (n - 1 >= 0):
n = n - 1
ret = ret + ITOA64[v & 0x3f]
v = v >> 6
return ret
def crypt(pw, salt):
if isinstance(pw, unicode):
pw = pw.encode('utf-8')
# Take care of the magic string if present
if salt.startswith(MAGIC):
salt = salt[len(MAGIC):]
# salt can have up to 8 characters:
salt = salt.split('$', 1)[0]
salt = salt[:8]
ctx = pw + MAGIC + salt
final = md5(pw + salt + pw).digest()
for pl in xrange(len(pw), 0, -16):
if pl > 16:
ctx = ctx + final[:16]
else:
ctx = ctx + final[:pl]
# Now the 'weird' xform (??)
i = len(pw)
while i:
if i & 1:
ctx = ctx + chr(0) #if ($i & 1) { $ctx->add(pack("C", 0)); }
else:
ctx = ctx + pw[0]
i = i >> 1
final = md5(ctx).digest()
# The following is supposed to make
# things run slower.
# my question: WTF???
for i in xrange(1000):
ctx1 = ''
if i & 1:
ctx1 = ctx1 + pw
else:
ctx1 = ctx1 + final[:16]
if i % 3:
ctx1 = ctx1 + salt
if i % 7:
ctx1 = ctx1 + pw
if i & 1:
ctx1 = ctx1 + final[:16]
else:
ctx1 = ctx1 + pw
final = md5(ctx1).digest()
# Final xform
passwd = ''
passwd = passwd + to64((int(ord(final[0])) << 16)
|(int(ord(final[6])) << 8)
|(int(ord(final[12]))),4)
passwd = passwd + to64((int(ord(final[1])) << 16)
|(int(ord(final[7])) << 8)
|(int(ord(final[13]))), 4)
passwd = passwd + to64((int(ord(final[2])) << 16)
|(int(ord(final[8])) << 8)
|(int(ord(final[14]))), 4)
passwd = passwd + to64((int(ord(final[3])) << 16)
|(int(ord(final[9])) << 8)
|(int(ord(final[15]))), 4)
passwd = passwd + to64((int(ord(final[4])) << 16)
|(int(ord(final[10])) << 8)
|(int(ord(final[5]))), 4)
passwd = passwd + to64((int(ord(final[11]))), 2)
return passwd