[dot-asm]

No description provided.

[Yudis-bit]

Roundtrip serialization fails for the point at infinity: blst_p1_serialize_eip2537 emits 128 zero bytes when Z is zero per EIP-2537, but blst_p1_deserialize_eip2537 rejects it with BLST_POINT_NOT_ON_CURVE.

This happens because the deserializer calls internal POINTonE1_affine_on_curve(&ret), where affine infinity (0, 0) fails y^2 == x^3 + 4 (0 == 4). Shouldn't this call blst_p1_affine_on_curve instead (or check vec_is_zero), like the other affine helpers do?

Same in blst_p2_deserialize_eip2537 for POINTonE2_affine_on_curve.

--- a/src/e1.c
+++ b/src/e1.c
@@ -387,3 +387,3 @@ BLST_ERROR blst_p1_deserialize_eip2537(POINTonE1_affine *out,

-    if (!POINTonE1_affine_on_curve(&ret))
+    if (!blst_p1_affine_on_curve(&ret))
         return BLST_POINT_NOT_ON_CURVE;
--- a/src/e2.c
+++ b/src/e2.c
@@ -456,3 +456,3 @@ BLST_ERROR blst_p2_deserialize_eip2537(POINTonE2_affine *out,

-    if (!POINTonE2_affine_on_curve(&ret))
+    if (!blst_p2_affine_on_curve(&ret))
         return BLST_POINT_NOT_ON_CURVE;

[dot-asm]

Roundtrip serialization fails for the point at infinity

Fixed. Thanks!