Bug 155138 - app-crypt/mhash-0.9.8.1 version bump request
Bug#: 155138 Product:  Gentoo Linux Version: unspecified Platform: All
OS/Version: Linux Status: RESOLVED Severity: enhancement Priority: P2
Resolution: FIXED Assigned To: robbat2@gentoo.org Reported By: eva@gentoo.org
Component: Ebuilds
URL:  http://sourceforge.net/project/showfiles.php?group_id=4286
Summary: app-crypt/mhash-0.9.8.1 version bump request
Keywords:  
Status Whiteboard: 
Opened: 2006-11-14 07:00 0000
Description:   Opened: 2006-11-14 07:00 0000 
Latest version in portage is 0.9.3 which is hard masked. I have bumped this
ebuild localy to test aide-0.12. 

I tested mhash-0.9.7.1 with aide-0.10 (to cover some algorithms on a large
range of input) and it works well.

------- Comment #1 From Robin Johnson 2006-11-15 00:33:33 0000 ------- 
On x86 (athlon-xp), and amd64 (core 2 duo), the testsuite passes fine.
However, on ppc64-32ul, I get two test failures:
...
testing SHA1 ...
testing HAVAL256   -- TEST FAILED: HAVAL256 "abcdefghijklmnopqrstuvwxyz"
got "4002E86400000002FFE8A8D04000ABB00000000000000000000000000000093C"
but expected "72FAD4BDE1DA8C8332FB60561A780E7F504F21547B98686824FC33FC796AFA76"
FAIL: hash_test.sh
...
Checking fragmentation capabilities of HAVAL256: Failed
...

I've filed a bug upstream in the meantime:
https://sourceforge.net/tracker/index.php?func=detail&aid=1596844&group_id=4286&atid=104286

------- Comment #2 From Robin Johnson 2007-01-04 00:38:17 0000 ------- 
no response from upstream on that ppc64 breakage.

------- Comment #3 From Gilles Dartiguelongue 2007-01-11 12:46:28 0000 ------- 
This other bug on ultrasparc looks similar :

http://sourceforge.net/tracker/index.php?func=detail&aid=1525585&group_id=4286&atid=104286

Robbin, did you try compiling with just -O2 in your CFLAGS ?

------- Comment #4 From Robin Johnson 2007-01-11 18:52:56 0000 ------- 
My normal CFLAGS on ppc64-32ul are:
CFLAGS="-Os -mtune=970 -mcpu=970 -mabi=altivec -maltivec -pipe
-Wstrict-aliasing"
The error (4002E86400000002FFEB15C04000ABB00000000000000000000000000000093C) 
is identical when I use:
CFLAGS="-O3"
CFLAGS="-Os"
CFLAGS="-O2"
CFLAGS="-O1"

If I use
CFLAGS="-O0", I still get an incorrect result, but a different one:
00000004000000004001F04C4002E86400000002FFAAB5C04000ABB000000000

Given that you say sparc has a similar failure, the source of the error MUST be
bit operations that don't take into account endianness of the machine.

------- Comment #5 From Samuli Suominen 2007-03-20 17:06:26 0000 ------- 
mhash-0.9.8.1 is released, needs to be tested.

------- Comment #6 From Robin Johnson 2007-03-28 03:38:37 0000 ------- 
in cvs now.
passes on my all my hardware.