Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 1 Jan 2013 19:03:58 +0100
From: magnum <>
Subject: Re: New self-test for maximum length

On 31 Dec, 2012, at 12:32 , magnum <> wrote:
> A few other formats actually got problems from the test, that would not occur IRL. They were due to optmizations and triggered because crypt_all() is not called before get_key() (I could call it but do not want to because it slows down GPU formats). sapB and sapG was among them. But Clipperz really should not have problems with it.

I re-worked the max-length test a little. It now does call crypt_all() but when used with a many-keys format, it does not really add a call but uses one that would happen anyway. So no slow down for GPU formats.

Anyway, the test now better reflects real-life use and should produce no false negatives. For some reason, I now get a few new problems:

Benchmarking: dynamic_1003 md5(md5($p).md5($p)) [32/64 64x2 (MD5_Body)]... FAILED (get_hash[1](1))
Benchmarking: MS SQL SHA-1 [128/128 SSE2 intrinsics 8x]... FAILED (get_hash[0](1))
Benchmarking: Oracle 11g SHA-1 [128/128 SSE2 intrinsics 8x]... FAILED (get_hash[0](1))
Benchmarking: HMAC MD5 [128/128 SSE2 intrinsics 12x]... FAILED (cmp_all(2))
Benchmarking: HMAC SHA-1 [128/128 SSE2 intrinsics 8x]... FAILED (cmp_all(2))
5 out of 199 tests have FAILED

I'll look into these, it still might be some unpredicted side-effect of the self-test.


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.