ImFormatStringToTempBufferV crashes if a null is passed for a %s parameter

[codefrog2002]

Version: 1.90
Branch: master

Many printf implementations will accept a nullptr when passed to a %s format.
That is, printf("%s", nullptr); will emit "(null)" on many implementations of printf.
I recognize this is undefined behavior, but it's easy to handle it gracefully.

ImFormatStringToTempBufferV instead crashes during its attempt to optimize around the heavy ImFormatStringV.

See this block of code - ImGui.cpp, void ImFormatStringToTempBufferV

    if (fmt[0] == '%' && fmt[1] == 's' && fmt[2] == 0)
    {
        const char* buf = va_arg(args, const char*); // Skip formatting when using "%s"
        
        if (buf == nullptr) buf = "(null)";  // << I added this to prevent crash from calling strlen(nullptr)
        
        *out_buf = buf;
        if (out_buf_end) { *out_buf_end = buf + strlen(buf); }
    }

And as of the newest version (1.90) there's a subsequent code block (looking for "%.*s") with the same issue.

    else if (fmt[0] == '%' && fmt[1] == '.' && fmt[2] == '*' && fmt[3] == 's' && fmt[4] == 0)
    {
        int buf_len = va_arg(args, int); // Skip formatting when using "%.*s"
        const char* buf = va_arg(args, const char*);
        if (buf == nullptr) buf = "(null)"; // Added  to prevent nullptr + length 
        *out_buf = buf;
        *out_buf_end = buf + buf_len; // Disallow not passing 'out_buf_end' here. User is expected to use it.
    }

[ocornut]

I have pushed your proposed change as 7bb0a52, thanks!

Note that the %..*s code path needs to account for clamping buf_len (I quickly tested msvc sprintf behavior and it did the same).