[tahoe-dev] [pycryptopp] #24: SHA256 failure on NetBSD with multiple segments

[pycryptopp]

#24: SHA256 failure on NetBSD with multiple segments
---------------------+------------------------------------------------------
Reporter:  warner    |           Owner:       
    Type:  defect    |          Status:  new  
Priority:  critical  |         Version:  0.5.1
Keywords:            |   Launchpad_bug:       
---------------------+------------------------------------------------------

Comment(by zooko):

 Good work, guys!

 I had a quick look at
 [source:pycryptopp/hash/sha256module.cpp at 20090528001548-92b7f-
 962955aa4bd6122250415514932018f0e0d7cc29 the current sha256module.cpp] and
 I didn't see any obvious bugs.  Do you?  I also ran the new
 {{{pycryptopp.test.test_sha256.SHA256.test_chunksize}}} on my linux
 workstation yukyuk both normally and under valgrind and it passed.  (Not
 surprising since the automatic tests on the yukyuk builder also pass.)

 A-ha!  I see that the new {{{test_chunksize}}} fails on Black Dew's Debian
 buildbot!  This is exciting because something else that Black Dew's Debian
 buildbot and MM's NetBSD machine have in common is that they fail
 http://allmydata.org/trac/tahoe/ticket/737 .  Hopefully there is a common
 cause of #24 (== http://allmydata.org/trac/tahoe/ticket/738 ).

 The next step is probably to see if Crypto++'s own test suite exercises
 this case, and if it doesn't make it do so, and then run Crypto++'s own
 test suite on one of the machines where this fails.

 Also, Black Dew could try running it under valgrind, like this:

 {{{
 valgrind --suppressions=/usr/lib/valgrind/python.supp python ./setup.py
 test -s pycryptopp.test.test_sha256.SHA256.test_chunksize
 }}}

 And, in parallel: set up a buildbot for pycryptopp on MM's NetBSD machine
 (in addition to the one he already has for Tahoe-LAFS).

-- 
Ticket URL: <http://allmydata.org/trac/pycryptopp/ticket/24#comment:1>
pycryptopp <http://allmydata.org/trac/pycryptopp>
Python bindings for the Crypto++ library