Let's rewrite the checks in message_reader::init()

[Roman-Koshelev]

1. Strengthen the checks (for example, check that the separating symbol is '='
2. We will detect invalid messages as soon as possible

[Roman-Koshelev]

TODO

1. Add checks that tags are numbers
2. Add a check that each tag has a corresponding value (and remove this check from the iterator increment)
3. Add hash sum check

[jamesdbrock]

Reviewing...

[jamesdbrock]

The way this worked before is that we only looked for the \x01 in the first 11 bytes, because the longest possible version prefix string is "8=FIXT.1.1". I think we should keep that behavior, and not look for \x01 in the entire buffer.

[Roman-Koshelev]

I don't understand why to take the maximum existing length. And if tomorrow there is a new version one character longer? I'll return it as it was ...

[jamesdbrock]

The invalid() function performs a mutation, it has no return value. This also returns nothing, but it makes it look like we care about the return value of invalid(). I prefer { invalid(); return; }

[Roman-Koshelev]

I agree

[jamesdbrock]

TODO

1. Add checks that tags are numbers

Yeah that's probably a good idea. I used to think that this was a waste of CPU cycles but I've changed my mind.

2. Add a check that each tag has a corresponding value (and remove this check from the iterator increment)

I think it's already true that if a tag has no corresponding value, then the message will be considered invalid?

3. Add hash sum check

What kind of hash sum check? We already have the FIX checksum.

[Roman-Koshelev]

3. The checksum we do not verify (

[Roman-Koshelev]

Update

[jamesdbrock]

1. The checksum we do not verify (

We have the option to verify the checksum though. We leave it to the library user to decide if they want to verify the checksum. https://jamesdbrock.github.io/hffix/classhffix_1_1message__reader.html#ad3d5b61394355c230321018cdc38a57b