Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 22 Jun 2012 17:22:39 +0400
From: Solar Designer <>
Subject: salts to use when benchmarking (was: SSH thread-safety)

magnum, Jim, Dhiru -

I don't recall us fixing the bug below, so I guess it's still present?

magnum, Jim - can one (or both) of you look into it ASAP and fix it?
I think it is very important to have it fixed before 1.7.9-jumbo-6,
especially if it's been introduced after -jumbo-5 (I don't know whether
this is the case or not).

If we don't fix this bug, then at least the SSH format's comment about
the use of two different keys in the benchmark is wrong (and needs to be
changed).  But I'd rather have the bug fixed.



On Tue, May 15, 2012 at 12:52:38PM +0400, Solar Designer wrote:
> magnum, Jim, Dhiru -
> On Tue, May 15, 2012 at 12:55:36PM +0530, Dhiru Kholia wrote:
> > Clean 1.7.9 switches between top two self-test vectors (other vectors
> > are tried twice or thrice only). In comparison, jumbo uses only top
> > self-test vector ( (other vectors are tried once). Tested with MD5
> > crypt.
> You mean salts, not entire test vectors.  It is as expected that clean
> 1.7.9 uses just the first two salts when benchmarking.  Jumbo's use of
> only one salt is a bug.  My guess is that it's been introduced along
> with support for -1000 and -1001 for BENCHMARK_LENGTH.  Can one of you
> look into this, please?
> Thanks,
> Alexander

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.