crypto/secp256k1: fix undefined behavior in BitCurve.Add (#22621)
This commit changes the behavior of BitCurve.Add to be more inline with btcd. It fixes two different bugs: 1) When adding a point at infinity to another point, the other point should be returned. While this is undefined behavior, it is better to be more inline with the go standard library. Thus (0,0) + (a, b) = (a,b) 2) Adding the same point to itself produced the point at infinity. This is incorrect, now doubleJacobian is used to correctly calculate it. Thus (a,b) + (a,b) == 2* (a,b) and not (0,0) anymore. The change also adds a differential fuzzer for Add, testing it against btcd. Co-authored-by: Felix Lange <fjl@twurst.com>
This commit is contained in:
committed by
GitHub
parent
d836ad141e
commit
0703ef62d3
@ -102,6 +102,7 @@ compile_fuzzer tests/fuzzers/stacktrie Fuzz fuzzStackTrie
|
||||
compile_fuzzer tests/fuzzers/difficulty Fuzz fuzzDifficulty
|
||||
compile_fuzzer tests/fuzzers/abi Fuzz fuzzAbi
|
||||
compile_fuzzer tests/fuzzers/les Fuzz fuzzLes
|
||||
compile_fuzzer tests/fuzzers/secp265k1 Fuzz fuzzSecp256k1
|
||||
compile_fuzzer tests/fuzzers/vflux FuzzClientPool fuzzClientPool
|
||||
|
||||
compile_fuzzer tests/fuzzers/bls12381 FuzzG1Add fuzz_g1_add
|
||||
|
Reference in New Issue
Block a user