There is a good reason to use zero-IV in CBC-MAC; if there is fixed no-Zero IV then an attacker can modify the $IV$ and $P_1$ so that the first block of the plaintext can become the advantage of the attacker.
The first step of the CBC-MAC tag calculation is performed as $$C_1 = E_k(IV \oplus P_1)$$
Now let $P_1'$ be the target first block that the attacker wants to change without changing the CBC-MAC tag. To achieve this simple forgery, they need to keep the same value for the first ciphertext $C_1$ so that the rest will result in the same MAC tag.
$$IV \oplus P_1 = IV' \oplus P_1'$$
Since the attacker knows $P_1$(‡), $IV$ and $P_1'$ than it is easy to find the $IV'$ $$IV' = IV \oplus P_1 \oplus P_1'$$ I.e. the attacker just need a simple x-or operation and transmit.
In this attack, the message size is not changed, only the first block of the plaintext and the IV is changed to forge the MAC. It is possible to combine two messages to forge another one, however, this changes the message size.
This makes a simple forgery on the non-fixed IV in the CBC-MAC. The CBC-MAC is already defined with zero-IV to mitigate this attack and the combining attack.
(‡) The security of the MAC schemes doesn't require the message to be encrypted. Attackers can still modify if there is encryption, this is very easy in CBC mode.