[tahoe-dev] [tahoe-lafs] #738: failure in block hash tree

[tahoe-lafs]

#738: failure in block hash tree
-------------------------------+--------------------------------------------
     Reporter:  midnightmagic  |        Type:  defect       
       Status:  new            |    Priority:  critical     
    Milestone:  1.5.0          |   Component:  code-encoding
      Version:  1.4.1          |    Keywords:  integrity    
Launchpad_bug:                 |  
-------------------------------+--------------------------------------------

Comment(by warner):

 [http://allmydata.org/trac/pycryptopp/ticket/24 pycryptopp#24] opened.

 midnightmagic and I were able to narrow this down to a failure in
 pycryptopp, in which hashing a 128-byte string in two chunks of size
 (33,95) gets the wrong value on !NetBSD. The block data hasher uses a tag
 (including netstring padding) of length 33. I suspect that the actual
 problem is with any block size BS such that {{{(33+BS)%128==0}}}.

 This smells a lot like [http://allmydata.org/trac/pycryptopp/ticket/17
 pycryptopp#17], which was an ARM-specific alignment issue that corrupted
 AES output on certain chunk sizes. I haven't looked deeply at the SHA256
 code yet, but I suspect the same sort of bug, this time affecting i386.

-- 
Ticket URL: <http://allmydata.org/trac/tahoe/ticket/738#comment:9>
tahoe-lafs <http://allmydata.org>
secure decentralized file storage grid