Here, in the function: int CodeCompletion::DoAllMethodsImpl(), we have such code:

wxString type = token->m_BaseType;
            if ((type.Last() == _T('&') || type.Last() == _T('*')) && type[type.Len() - 2] == _T(' '))
            {
                type[type.Len() - 2] = type.Last();
                type.RemoveLast();
            }

But both three tokens have the same "base type". Do shall we use "m_FullType" instead?

Found another problem:

If I have another class:

class A

{

public:

A(int x, int y){}

~A(){}

};

And declare a member function in TestClass:

void Test4(const A& a = A(1, 2));

Insert implementation will result in:

void TestClass::Test4(const A& a, 2))

{

}

Hi, thanks for this report, I believe this is another bug, and I will look into it. Thanks.

wxString Token::GetStrippedArgs() const
{
    // the argument should have the format (xxxx = y, ....) or just an empty string
    // if it is empty, we just return an empty string
    if (m_Args.IsEmpty())
        return wxEmptyString;

    wxString args;
    args.Alloc(m_Args.Len() + 1);
    bool skipDefaultValue = false;
    for (size_t i = 0; i < m_Args.Len(); ++i)
    {
        const wxChar ch = m_Args[i];
        if (ch == _T('\n'))
            continue;
        else if (ch == _T('='))
        {
            skipDefaultValue = true;
            args.Trim();
        }
        else if (ch == _T(','))
            skipDefaultValue = false;

        if (!skipDefaultValue)
            args << ch;
    }

    if (args.Last() != _T(')'))
        args << _T(')');

    return args;
}
This function works badly when m_Args="(const A& a = A(1, 2))".

Now, I see skipping parentheses is quite complex. See:

void Tokenizer::ReadParentheses(wxString& str)

I add a new bug report about your second bug in this thread.

https://developer.berlios.de/bugs/?func=detailbug&bug_id=18600&group_id=5358

Here, I will focus on the original bug.