blst_scalar_from_uint32 appears to read past a uint32

[friedrichsenm]

When using blst_scalar_from_uint32, the scalar gets additional bytes added to it that aren't from the integer.

This looks like it is because the function expects an array of 8 integers, but in the Rust bindings (which I'm using), I'm only passing in a single u32, so the C code is reading extra bytes past the integer and filling the scalar with them.

[friedrichsenm]

Thinking about the Rust API a bit, more it's possible that this is meant to take a pointer to a slice/array of uint32 in which case this is completely expected behavior. The name of the API suggests to me that it expects a single uint32 so maybe some documentation to clarify how it should be used would be helpful.

[dot-asm]

Originally the Rust bindings targeted a specific application, which is why they can feel counterintuitive in a more general case. Yes, the call in question expects a properly sized array of unit32-s primarily for deserialization purposes. Either way, passing a single 32-bit value to make a scalar would be a waste since both input and output need to be properly sized. The question should rather be is there a better way? Scalar is just a byte array (starting with the least-significant byte) and you can specify how many bits to use in the point multiplication. As opposed to passing full-blown scalar with just few significant bits that is. One can also do it en-masse in multi-scalar multiplication. Customarily it's about sort of blinding, and what one does is to fill a byte array of suitable size with random data and have blst treat it as a contiguous array of scalars...