Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 11 Jul 2012 19:29:27 +0200
From: Frank Dittrich <>
Subject: Re: Test Suite vs. GPU formats

On 07/11/2012 06:48 PM, Solar Designer wrote:
> magnum, Jim, myrice, Samuele, all -
> magnum ran the Test Suite on bull today, on lots of formats (both CPU
> and GPU).  xsha512-cuda was stuck for more than 8 hours until I killed
> it.  Some others similarly did not complete soon enough and I had to
> kill them.  IIRC, this applied to all of myrice's formats, plus some of
> Samuele's.  In general, if a fast and saltless hash takes more than a
> minute to pass through TS, the implementation is seriously broken and
> needs to be fixed.  For slow/salted hashes, we may allow a bit more
> time (like up to 10 minutes for those that are both slow and salted),
> but not a lot more.

May be the test suite should (optionally) use individual session names /
pot file names per format, and delete those individual files only in
case of success, not in case of an error.
Then, you could try to investigate what went wrong for which format and
optionally even restore the sessions in case someone had to kill these

> My guess is that the broken formats test many more candidate password
> slots than we have lines in the test wordlist, although even that is
> not enough of an explanation for the saltless formats - there must be
> some other shortcoming as well.  This needs to be investigated.

We could check if adding --mkpc=5000 (or whatever number of candidate
passwords is in the file helps.
If this helps, the formats might need to be adjusted nevertheless.
But at least it gives a clue where to start looking.

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.