[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 2 Jul 2012 00:04:11 -0400
From: "Robert B. Harris" <rs904c@...scape.net>
To: <john-dev@...ts.openwall.com>
Subject: RE: Significant items concerning the MS Windows (cygwin) build of john-1.7.9-jumbo-6
Alex,
Sorry it has taken me so long to reply, that was a lot of work to test.
> I guess you also needed to install more components of Cygwin for this to
work? Or did you make a "full" > > > install of it right away? IIRC, they
exclude a lot of stuff by default.
############## I did a full install of cygwin, so I don't know which
components install these DLLs.
> -The self test for Django PBKDF2-HMAC-SHA-256 hangs in the OMP buid.
> It works fine in the non-OMP build.
>> This did not happen for me. (And I did test would-be -jumbo-6 on Win32,
including an OpenMP build, before releasing.)
>> The hang sounds like the known Cygwin bug that I worked around last year
by binary-patching cygwin1.dll.
>>Can you try again with the binary-patched cygwin1.dll as included in
john179j5w.zip?
######### I didn't know about the patched cygwin1.dll. Can you point to me
some more information about that?
Bottom Line up front: The format Django does not actually hang in a windows
compiled OMP version, but the performance is very poor.
See below.
The following is an intrinsic build of john, running the format Django:
C:\#dev\john\wip\john-1.7.9-jumbo-6\run>john --test -format=Django
Benchmarking: Django PBKDF2-HMAC-SHA-256 (x10000) [32/32]... DONE
Raw: 28.5 c/s real, 28.8 c/s virtual
The following is an intrinsic and omp build of john, running the format
Django.
(It also is using the patched cygwin1.dll)
C:\#dev\john\wip\john-1.7.9-jumbo-6\run>john-omp --test -format=Django
Benchmarking: Django PBKDF2-HMAC-SHA-256 (x10000) [32/32]... (8xOMP) DONE
Raw: 0.6 c/s real, 2.8 c/s virtual
It didn't seem to matter if I was using the new cygwin1.dll or the patched
one.
The following is an intrinsic and omp build of john, running the format
Django.
(It is using the regular cygwin1.dll, which is not patched)
C:\#dev\john\wip\john-1.7.9-jumbo-6\run>john-omp --test -format=Django
Benchmarking: Django PBKDF2-HMAC-SHA-256 (x10000) [32/32]... (8xOMP) DONE
Raw: 0.7 c/s real, 2.6 c/s virtual
C:\#dev\john\wip\john-1.7.9-jumbo-6\run>john-omp-nonintrinsic.exe --test
-format=Django
Benchmarking: Django PBKDF2-HMAC-SHA-256 (x10000) [32/32]... (8xOMP) DONE
Raw: 0.6 c/s real, 2.8 c/s virtual
C:\#dev\john\wip\john-1.7.9-jumbo-6\run>john-nonintrinsic.exe --test
-format=Django
Benchmarking: Django PBKDF2-HMAC-SHA-256 (x10000) [32/32]... DONE
Raw: 29.2 c/s real, 29.2 c/s virtual
> form=des guesses: 26 time: 0:00:00:01 :
Expected
> count(s) (1500) [!!!FAILED!!!]
>
> .pot CHK:des guesses: 1500 time: 0:00:00:00 [PASSED]
>
>
>
> form=netlmv2 guesses: 926 time: 0:00:00:01 :
Expected
> count(s) (1500) [!!!FAILED!!!]
>
> .pot CHK:netlmv2 guesses: 1500 time: 0:00:00:01 [PASSED]
This is puzzling, especially "des". Non-OpenMP build, right?
############## Yes, this was the Non-OpenMP build. It does look like these
are new problems in jumbo-6, see test suite results below.
Is this reproducible? With the exact same guess counts?
Can you see if the issue is new with j6 or if we had it in j5 as well?
For "des", if we had it in 1.7.9 proper as well?
Here are my Test Suite results for 1.7.9-jumbo5 ran on Windows
----------------------------------------------------------------------------
---
- JtR-TestSuite (jtrts). Version 1.12.5, June 29, 2012. By, Jim Fougeron &
others
- Testing: John the Ripper password cracker, ver: 1.7.9-jumbo-5_omp
[win32-cygwin-x86-sse2i]
----------------------------------------------------------------------------
----
form=crc32_dups_dupe_rem guesses: 0 unk unk : Expected count(s)
(50) [!!!FAILED!!!]
.pot CHK:crc32_dups_dupe_rem guesses: 0 unk unk [PASSED]
form=crc32_dups2_dupe_rem guesses: 0 unk unk : Expected count(s)
(50) [!!!FAILED!!!]
.pot CHK:crc32_dups2_dupe_rem guesses: 0 unk unk [PASSED]
form=hmacMD5 guesses: 504 time: 0:00:00:01 : Expected
count(s) (1500) [!!!FAILED!!!]
.pot CHK:hmacMD5 guesses: 168 time: 0:00:00:00 : Expected
count(s) (1500) [!!!FAILED!!!]
form=sapG-utf8 guesses: 0 unk unk : Expected count(s)
(1500) [!!!FAILED!!!]
.pot CHK:sapG-utf8 guesses: 0 unk unk [PASSED]
form=sapB guesses: 0 unk unk : Expected count(s)
(1500) [!!!FAILED!!!]
.pot CHK:sapB guesses: 0 unk unk [PASSED]
----------------------------------------------------------------------------
---
- JtR-TestSuite (jtrts). Version 1.12.5, June 29, 2012. By, Jim Fougeron &
others
- Testing: John the Ripper password cracker, ver: 1.7.9-jumbo-5
[win32-cygwin-x86-sse2i]
----------------------------------------------------------------------------
----
form=crc32_dups_dupe_rem guesses: 0 unk unk : Expected count(s)
(50) [!!!FAILED!!!]
.pot CHK:crc32_dups_dupe_rem guesses: 0 unk unk [PASSED]
form=crc32_dups2_dupe_rem guesses: 0 unk unk : Expected count(s)
(50) [!!!FAILED!!!]
.pot CHK:crc32_dups2_dupe_rem guesses: 0 unk unk [PASSED]
form=hmacMD5 guesses: 504 time: 0:00:00:01 : Expected
count(s) (1500) [!!!FAILED!!!]
.pot CHK:hmacMD5 guesses: 168 time: 0:00:00:00 : Expected
count(s) (1500) [!!!FAILED!!!]
form=sapG-utf8 guesses: 0 unk unk : Expected count(s)
(1500) [!!!FAILED!!!]
.pot CHK:sapG-utf8 guesses: 0 unk unk [PASSED]
form=sapB guesses: 0 unk unk : Expected count(s)
(1500) [!!!FAILED!!!]
.pot CHK:sapB guesses: 0 unk unk [PASSED]
-Robert B. Harris from VA
View attachment "Test-Suite-Results-OMP (for 1.7.9-jumbo5).txt" of type "text/plain" (22524 bytes)
View attachment "Test-Suite-Results (for 1.7.9-jumbo5).txt" of type "text/plain" (22521 bytes)
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.
