When i==arg.size(), arg[i] is out of bounds.

[ibc]

https://github.com/philsquared/Catch/blob/master/single_include/catch.hpp#L4435

        void parseIntoTokens( std::string const& arg, std::vector<Token>& tokens ) {
            for( std::size_t i = 0; i <= arg.size(); ++i ) {
                char c = arg[i];
                if( c == '"' )
                    inQuotes = !inQuotes;
                mode = handleMode( i, c, arg, tokens );
            }
        }

Shouldn't it be for( std::size_t i = 0; i < arg.size(); ++i )?

[philsquared]

That does look very suspicious. It shouldn't be out of bounds as the character at size() should be a null - although I suspect that the standard allows for the implementation to throw if asked for it (would need to check that).
I suspect that making it < is the right thing to do. The reason I don't sound 100% certain is that there does seem to be handling for null elsewhere and I'd need to check it's not relying on that (and if it is probably change it).
So I'll take a closer look soon.
Have you actually had an issue with it, or did you just spot it manually?

[ibc]

I just found it by using a code coverage/linter stuff, never experimented an issue related to it.

[horenmar]

C++03 does not guarantee there being \0 at str[str.size()], in fact I think it does not even guarantee that it is inside the string's bounds.

Granted, I don't think there is an implementation that does not have it there, but it's better to be safe than sorry.

[philsquared]

Agreed. I'll change it once I've checked for side-effects

[ibc]

Great.