Fix #13327 FN mismatchingContainers for struct member and c_str()

[chrchr-github]

Co-authored-by: Paul Fultz II pfultz2@yahoo.com

[pfultz2]

This is problematic if there is divergent lifetimes, that is why we check the size is 1 or no duplicates for iterator kinds. Why is there multiple lifetimes? Do we have a lifetime for t.s.a and a both?

[chrchr-github]

This is problematic if there is divergent lifetimes, that is why we check the size is 1 or no duplicates for iterator kinds. Why is there multiple lifetimes? Do we have a lifetime for t.s.a and a both?

In the s.a example, the lifetime tokvalues are s and .. But I think it makes sense to choose the dot, unless there can be multiple dots.

[pfultz2]

Its the . in s.a I assume? Ideally we should remove the lifetimes that refer to a sub expression of another lifetime from the vector. Something like this function could be called here(its not tested though):

std::vector<ValueFlow::Value> pruneLifetimes(std::vector<ValueFlow::Value> lifetimes)
{
    std::vector<ValueFlow::Value> result;
    auto start = lifetimes.begin();
    while(start != lifetimes.end())
    {
        const Token* tok1 = start->tokvalue;
        auto it = std::partition(start, lifetimes.end(), [&](const ValueFlow::Value& v) { 
            const Token* tok2 = v.tokvalue;
            return astHasToken(tok1, tok2) or astHasToken(tok2, tok1); 
        });
        auto root = std::min_element(start, it, [](const ValueFlow::Value& x, const ValueFlow::Value& y) {
            return astHasToken(x.tokvalue, y.tokvalue);
        });
        result.push_back(*root);
        start = it;
    }
    return result;
}

[chrchr-github]

Not entirely sure what it does, but it seems to work 👍

[pfultz2]

First it groups the lifetimes together using std::partition with the lifetimes that refer to the same token or token of a subexpression. Then it finds the lifetime in that group that refers to the "highest" parent using std::min_element and adds that to the vector. Hopefully that makes sense.