In 5.6.0 I did make a change to the BlockTransformation interface, so that classes (such as AES) that implement it no longer assume that input and output are aligned. In 5.5.x and earlier this alignment assumption was supposed to be enforced by the Mode classes, but apparently (as Brian's work shows) there was a bug that manifested itself on ARM. I made that change in part because I thought the alignment enforcement code was too complicated, but I wasn't specifically aware of this bug at that point.
I'll add a note to the change history about the modification to the BlockTransformation interface, as well as this bug. -------------------------------------------------- From: "Zooko Wilcox-O'Hearn" <[email protected]> Sent: Thursday, June 04, 2009 3:06 PM To: "Crypto++ Users" <[email protected]>; "Brian Warner" <[email protected]> Subject: Re: [tahoe-dev] pycryptopp vs ARM > > Dear Wei Dai: > > It looks like you specifically fixed this bug, judging by the name of > the macro: "CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS". But there is no > mention of this issue in the Readme.txt about release 5.6.0, so I'm > guessing that you forgot to mention it or decided not to mention it > in the release notes. Is that right? > > Thanks, > > Zooko > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the "Crypto++ Users" Google Group. To unsubscribe, send an email to [email protected]. More information about Crypto++ and this group is available at http://www.cryptopp.com. -~----------~----~----~----~------~----~------~--~---
