Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 13 Jun 2012 15:27:42 -0500
From: "jfoug" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: RE: JtR test suite: 2 errors with generic build on 32bit linux.

>This is a false alarm. The cause is the new wordlist dupe suppression.
>This format has FMT_NOT_EXACT and it is supposed to be tested *with*
>dupes in the password file. A normal format should not present another
>crack of the same hash from dupe input, but this should.

The 'dupes' on the dictionary file, was the quick and easy way to test
duplication.  Instead of finding 30 duplicate crc values for each of the 50
hashes, I simply built the dictionary, by making it 50 entries long, then
appended those same 50 words to the end of that password file 30 times.
Then when run against an input file with the expected 50 CRC's, it should
find 1500.  The code to do the pot checking sends the words through unique,
so the .pot check for this test is only 50, with all duplicates removed.

Thus, when the wordlist crunching code works (which, btw, I think SHOULD
only be done when a command line option is used), the input dictionary gets
ripped down to 50 words, from 1500, and thus the TS run only finds 50, but
that is fewer than it expected.

Magnum and I talked about this offlist last night.  Likely we can add an
option to the TS on this line, and do a -mem=1 forcing the wordlist to not
dupe check.

Jim.

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.