Use Go's ModSqrt and Jacobi instead of our own #249
Conversation
|
Looks good |
|
Have you double-checked that the Go standard versions are "just as constant-time" as ours? |
|
Yes. They use the same algorithm. As far as I can see, both are NOT constant time. Is that what you expect? Also: While developing this, I went through several steps. My first (incorrectly implemented) attempt caused test failures, so then I rewrote it to calculate Go's answer and our answer side by side and panic on difference. Our test suite passed in that mode .Then I switched it to using Go's results and it continued passing, and that showed me how I'd screwed up the first time. I am confident that our test suite covers this code. |
|
It doesn't surprise me that they're both not constant-time, but I just couldn't remember for sure. Fixing that will presumably be part of the much bigger challenge of making big.Int support constant time in general; see this Golang proposal thread I started a while ago on this topic: golang/go#20654 |
Closes #231.