Re: OpenCL plans & details

Bitweasil wrote:Update:
Found a major bug in how I'm handing odd password lengths, to the point that odd password lengths are being done "totally wrong."
I need to fix this before I throw out a beta...
Further update:
I'm an idiot.
Before copying a small number of bytes into a uint64_t pointer, one may wish to set the memory at the pointer to zero first.
Values hanging out in the high order bytes you aren't copying into matter. A LOT.
The good news: My generate code is actually solid. Awesome.
The annoying news: I just spent 90 minutes tracking down a bug in my "table verify" code due to not having the memset in place. GRAGH.
Better than the few hours I spent trying to figure out weird data corruption when opening binary files in Windows with the "r" or "w" flags instead of "rb" or "wb"...
Anyway, I'm more confident in my code integrity now, and that it's doing the correct thing, so as soon as I finish adding NTLM support to the OpenCL cracking tools and verifying those, I should be ready to throw out a beta. It's less useful with MD5 only right now...
