Opened at 2007-11-01T17:08:23Z
Closed at 2007-11-10T00:48:33Z
#199 closed defect (fixed)
pycrypto gives bad output from SHA-256 for certain input lengths
Reported by: | zooko | Owned by: | zooko |
---|---|---|---|
Priority: | blocker | Milestone: | 0.7.0 |
Component: | code | Version: | 0.6.1 |
Keywords: | Cc: | ||
Launchpad Bug: |
Description
The implementation of SHA-256 in pycrypto (which we use in Tahoe <= v0.6.1) gives the wrong output if the input length % 64 is 55.
The plan is to fix this by either applying another patch to our version of pycrypto or switching to a different crypto library. Investigating the latter is why I am working on pycryptopp.
Change History (6)
comment:1 Changed at 2007-11-07T05:32:59Z by zooko
- Component changed from unknown to code
comment:2 Changed at 2007-11-07T05:56:11Z by zooko
- Status changed from new to assigned
comment:3 Changed at 2007-11-07T20:31:14Z by zooko
Next step -- figure out why it doesn't compile on Brian's Mac OS X.
Next step after that -- change uses of pycrypto's SHA-256 to pycryptopp's SHA-256.
comment:4 Changed at 2007-11-08T00:04:47Z by zooko
- Resolution set to fixed
- Status changed from assigned to closed
Fixed by 11412da436be9e37.
(Except possibly for build/porting/packaging errors on some systems -- see the loyal BuildBot? for details.)
comment:5 Changed at 2007-11-09T19:32:21Z by zooko
- Resolution fixed deleted
- Status changed from closed to reopened
The version of pycryptopp that I used had a bug in the SHA-256 wrapper code. This bug is now fixed, so I'm now running tests and will re-apply this patch.
comment:6 Changed at 2007-11-10T00:48:33Z by zooko
- Resolution set to fixed
- Status changed from reopened to closed
closed again by 5807e83f7ba926af
http://allmydata.org/pipermail/tahoe-dev/2007-November/000231.html
Next step: update tahoe README to state that the user has to manually install Crypto++, and update the tahoe automatic dependencies to depend on pycryptopp, and change our calls to pycryto's sha256 to pycryptopp's sha256.