/usr/include/poker-eval/enumerate.h incorrectly assumes sizeof(int)==sizeof(int*)

[Scott Wolchok]

The offending line is line 624:

  _curElem = (int **) malloc(num_sets * sizeof(int)); \

It should read:

  _curElem = (int **) malloc(num_sets * sizeof(int*)); \

because the array being allocated is an array of int*, as shown by the cast, not an array of int. On an x86 system, this is probably fine, but on amd64, sizeof(int) != sizeof(int*).

This causes a crash in a program I wrote using python-pypoker-eval to evaluate pre-flop odds in Texas Hold'em, but only when more than 2 hands are passed to it.

The program (preflop_matchup.py):

#!/usr/bin/env python

import itertools
import sys

import pokereval

def matchup(hands):
  pok = pokereval.PokerEval()
  answer = pok.poker_eval(game='holdem',
                          pockets=hands,
                          board=['__', '__', '__', '__', '__'])
  samples = float(answer['info'][0])
  for hand, results in zip(hands, answer['eval']):
    win_pct = results['scoop'] / samples * 100
    tie_pct = results['tiehi'] / samples * 100
    print '%s: %f%% to win, %f%% to tie' % (hand, win_pct, tie_pct)

def grouper(n, iterable, fillvalue=None):
    "grouper(3, 'ABCDEFG', 'x') --> ABC DEF Gxx"
    args = [iter(iterable)] * n
    return itertools.izip_longest(fillvalue=fillvalue, *args)

def main(argv):
  hands = []
  for item in argv:
    hand = list(''.join(x) for x in grouper(2, item))
    if len(hand) != 2:
      raise Exception('hands must contain two cards of 2 chars each!')
    hands.append(hand)
  matchup(hands)

if __name__ == '__main__':
  main(sys.argv[1:])

To reproduce the crash, on an amd64 Ubuntu 10.4 LTS system:
python preflop_matchup.py 9h9s adjd jc7c

Correcting the offending line solves the problem. This may already be fixed upstream, as the Ubuntu package is 1 revision out of date.

[Victor Vargas (kamus)]

checking latest release of poker-eval (137.0) that line now is replaced by: " _curElem = (int **) malloc(num_sets * sizeof(intptr_t)); \ ", , please could you check if this issue is still occuring under latest release included in Maverick (or use that code just for testing under Lucid)? Thanks.

[Scott Wolchok (scott-wolchok)]

Yes, s/sizeof(int*)/sizeof(intptr_t)/ in the patch seems to work just as well. I'd argue it's bad style, but that's a different debate.

[Launchpad Janitor (janitor)]

[Expired for poker-eval (Ubuntu) because there has been no activity for 60 days.]