Bad arguments (in C++ code) to BlindTransaction() can cause memory corruption

https://github.com/ElementsProject/elements/blob/4e023af58b697f63d0612abd712e1f8310719217/src/blind.cpp#L407-L416

If the issuance blinding key or token blinding key is specified for the input when calling BlindTransaction, but the transaction input itself have null value in issuance.nInflationKeys/issuance.nAmount, then num_blind_attempts will be incremented, while the size of value_blindptrs and asset_blindptrs will not be increased with another 32 bytes of random data.

(the same issue can happen with continue at https://github.com/ElementsProject/elements/blob/4e023af58b697f63d0612abd712e1f8310719217/src/blind.cpp#L434-L435)

If num_blind_attempts is larger than the size *_blindptrs arrays, secp256k1_pedersen_blind_generator_blind_sum that is called later can go over the size of the arrays, because there's no check that the sizes are equal to num_blind_attempts + num_known_input_blinds. This can cause crash. secp256k1_pedersen_blind_generator_blind_sum will also write to the last blinder, and this can cause memory corruption.

Of course this is only possible by the calling C++ code being incorrect, supplying wrong transaction data, etc. But this raises the question - are these continue statements actually needed ? Wouldn't it be better to replace them with returning failure ? It seems that in both cases, either the transaction structure is wrong, or the arguments issuance_blinding_privkey/token_blinding_privkey are wrong.

src/blind.cpp の 407-416 行および 434-435 行から開始し、BlindTransaction の呼び出し元と value_blindptrs および asset_blindptrs の構築を追跡します。無効な発行またはブラインディングキーの引数がどのように処理されるかを確認し、個数が一致しない場合のカバレッジを追加します。完了の条件は、不正な入力によって blind_sum が配列の範囲外を読み書きできないことです。