Why does OP_CheckMultiSig not fail on an invalid public key?

[Coding Enthusiast]

Basically this test vector: https://github.com/bitcoin/bitcoin/blob/b6a5dc90bfd4640cf9f914e59bf8e21cd265b51e/src/test/data/tx_valid.json#L185-L188

It is spending a P2SH output with the redeem script of it being:
OP_2 <invalid_pub><Pub><Pub> OP_3 OP_CheckMultiSig
The first push which is supposed to be a public key is an invalid public key (it is a DER encoded signature) and despite that OP_CheckMultiSig passes successfully!
Is this another bug at protocol level? Also is it the same for SegWit scripts?

[gmaxwell]

Is this another bug at protocol level?

Why would you consider this a bug?  Yes, it could have been implemented differently, but the definition of CMS is that signature validation passes N of M times.

Obvious ways of implementing it-- including the way it was originally implemented-- wouldn't provide for any way to determine exactly why a signature failed to validate only that it did.

[Coding Enthusiast]

Why would you consider this a bug?  Yes, it could have been implemented differently, but the definition of CMS is that signature validation passes N of M times.

I guess because I usually expect certain things to be strict specially when it comes to evaluating scripts.
This is not the first time I'm surprised by such rules though.