[tahoe-dev] [pycryptopp] #3: serialize ecdsa keys without the fluff

[pycryptopp]

#3: serialize ecdsa keys without the fluff
-----------------------------+----------------------------------------------
     Reporter:  zooko        |      Owner:  zooko   
         Type:  enhancement  |     Status:  reopened
     Priority:  major        |    Version:  0.4.0   
   Resolution:               |   Keywords:          
Launchpad_bug:               |  
-----------------------------+----------------------------------------------

Comment(by zooko):

 Here's a patch which is the previous patch minus the feature of embedding
 C++ objects directly into Python objects.  (If you are interested, observe
 how there are more conditions in this patch which can lead to
 {{{PyErr_NoMemory()}}}, while doing so, please see if there are some
 conditions in which allocation can fail but I forgot to test for it.
 Those can be exploitable bugs in some cases, and I found another one just
 now while working on this patch.)

 This patch is probably buggy as all hell.  I'm too sleepy to test it.  By
 the way, I've been observing how running the pycryptopp unit tests from
 trunk pycryptopp reliably give me a segfault in
 {{{pycryptopp.test.test_rsa.SignAndVerify.test_serialize_and_deserialize_signing_key_and_test}}}
 on yukyuk (linux-amd64) but not on ootles (MacOSX-amd64), and how the
 reliable segfault disappears when I try to run it under valgrind.  Also I
 occasionally see segfaults from other programs on yukyuk -- I can see one
 from okular in my syslog from earlier today.  memtest86 always passes on
 yukyuk.  I'll probably experiment with different linux kernels, or pouring
 cold water over the CPU, or something tomorrow.

-- 
Ticket URL: <http://allmydata.org/trac/pycryptopp/ticket/3#comment:10>
pycryptopp <http://allmydata.org/trac/pycryptopp>
Python bindings for the Crypto++ library