fix(crypto): reject infinity in point address derivation

FE
5d ago
+65-0

What does this PR do?

Reject the point at infinity in Hash.computeAddress(ECPoint) with IllegalArgumentException. Document the uncompressed public-key input format of the byte[] overload.

Why are these changes required?

The infinity encoding is {0x00}. Previously, removing its first byte caused address derivation to hash empty input and return a well-formed address for an invalid public key.

This PR has been tested by:

  • 39 tests passed across HashTest, ECKeyTest, and BouncyCastleTest, including 3 new regression tests for infinity rejection, known public-key addresses, and legacy byte-array behavior.
  • Passed :framework:checkstyleMain :framework:checkstyleTest :plugins:checkstyleMain, check_math_usage.sh, and git diff --check.
  • Full test suite and manual node testing were not run.

Follow up

None.

Extra details

Transaction/block signature verification and TVM address recovery use the unchanged byte[] overload. Existing ECKey construction already rejects infinity, so normal getAddress() behavior is preserved. Direct callers passing infinity to the ECPoint overload now receive an exception; null-input behavior is unchanged.

Summary by cubic

Rejects the point at infinity in Hash.computeAddress(ECPoint) so invalid public keys no longer produce a well-formed address, and documents the input contract of the byte[] overload.

  • The infinity encoding {0x00} previously lost its first byte and hashed empty input, returning a valid-looking address for an invalid public key.
  • The ECPoint overload now throws IllegalArgumentException for infinity; the byte[] overload keeps its legacy behavior, which consensus paths such as signature recovery rely on.
  • Adds regression tests covering infinity rejection, known public-key addresses, and legacy byte-array behavior.