Well, GDI+ is for sure easier to use for C++ programmers. Instead of "selecting" stuff back and forth you could just use C++ objects, working with various picture formats is very easy, and overall model is well though out. There are few quirks though - if you're trying to load anything but bitmap picture from resources you'll be forced to use GlobalAlloc, CreateStreamOnHGlobal and whatnot.

But if you try to use that some code to load bitmap from resources it will fail?

Furthermore, if I load PNG file, I get correct info about colors and palette (8bit, indexed), but if I load 256 color bitmap it gets implicitly converted to 32bit?

Drawing text is so different that I found it impossible to display some text to look the same as with GDI. Yes, I've read the Knowledge base article explaining why and how to get similar result, but it doesn't help - I can't seem to get the font of exactly the same size as before.