if go fast is broken, the go slow it shall be. OK
On 2023 Mar 16 (Thu) at 15:28:56 +0100 (+0100), Theo Buehler wrote: :The AES-CBC implementation is broken on T-4 machines. I reported this :upstream: : :https://github.com/openssl/openssl/issues/20523 : :From grep I suspect that this also affects other AES modes, and also :DES and Camellia. : :Until someone fixes this, I think we should disable the assembly :optimizations. This makes regress on a T4-2 LDOM stable where with :assembly optimizations it explodes about 2 out of 3 runs. : :It looks like this bug has been around for at least a decade. So maybe :this is overkill, but when the choice is between reliability and speed :(if that's the right word for anything related to sparc64), it seems :clear that reliability should trump, especially when it comes to a :dependency of a mailserver and a backup utility. : -- "It runs like _x, where _x is something unsavory" -- Prof. Romas Aleliunas, CS 435
