On Feb 22, 2006, at 4:21 PM, Reed Sheridan wrote:
On 2/22/06, Kon Lovett <address@hidden > wrote: I will make them static.
There's still a conflicting typedef of SHA<digits>_CTX. The one in libcrypto happens to be a few bytes bigger than Aaron Gifford's, so my original example works, but I think that if you were to load the sha2 egg first, and then libcrypto, you would have problems with accessing unallocated memory. There may be other conflicts as well: I get the impression that Gifford mostly copied his code (or at least the interface) from openssl, changing just enough to make it incompatible.
I don't understand. The symbols SHA<digits>_CTX are not visible to the linker. Is the Gifford C source textually included outside of sha2.scm?
I went ahead and implemented sha256:digest with libcrypto. It's a screenful of code, easier to deal with than groveling through Gifford's code. (It also appears to be faster, and I have more faith in the stability and security of code from de Raadt & co than a random hacker anyway. libcrypto would be a much better candidate for message digests than the assortment of standalone files that we have now if nobody minded depending on it, which, obviously, some people do). So this isn't my problem anymore, and I've spent enough time on it, but this issue should be documented if nobody wants to make sure it's fixed.
Yes.
|