[wasphin]

What problem does this PR solve?

Problem Summary:

The RVV Base64 decoder accepts incomplete groups and strips more padding than the scalar decoder.

What is changed and the side effects?

Changed:

• Require nonempty padded inputs to have a length divisible by four.
• Remove at most two trailing padding characters so extra padding is rejected.
• Add common malformed-length and padding cases.

Side effects:

• Performance effects: Constant-time validation before vector decoding.
• Breaking backward compatibility: Malformed inputs now follow the scalar decoder behavior.

---

Check List:

• Changes compile locally.
• Added focused test coverage.
• Please follow Contributor Covenant Code of Conduct.

[Copilot]

🟢 Approval recommended

The narrowly scoped validation change has focused regression coverage and no identified blocking issues; tests were not executed during this review.

No open issues 🎉

What changed in this PR

This PR aligns the RVV Base64 decoder’s input validation with the scalar decoder, rejecting malformed lengths and excess padding while preserving empty-input handling.

Changes:

• Require nonempty inputs to have a length divisible by four and strip at most two trailing padding characters.
• Add regression cases for malformed inputs and valid boundary cases.

File	Description
test/​base64_unittest.cc	Adds malformed-length, padding, and valid-input regression cases.
src/​butil/​third_party/​modp_b64/​modp_b64_rvv.cc	Enforces scalar-compatible length and padding validation.

_{Review effort: Balanced}