std::max is no perfect solution. It breaks trying to compare int to long, instead of just letting the compiler promote the type as needed.

I need unistd.h for sleep(). Most online references say that’s the place to look for it. So, does that function even exist in a VC++ world? In some other header?

@AhmedAEK - I’m running cmake within MSVC because that’s the only environment I can find with a debugger that works on Windows/ARM. How gladly I would have used VSCode with msys2, but I could find no debugger there that works! Windows/ARM seems to be a forgotten stepchild.

I'm tempted to just turn off STRICT. But first, is there a good way that I could code this more gracefully so that the debugger would display it more gracefully? Can a debugger display the data hidden in a handle at all? Or is that data hidden behind operating system magic?

Here’s a small example file that fails to load: easyupload.io/w0rsdk

Thanks that’s partway there. I am surprised at your remark: seems to me, if relative pathnames are that unreliable, something is BROKEN! But anyway, results: I create a BMP file from scratch with Paint, it loads just fine. I take one I was already using and re-save it (with or without making a change) and it does not load. Both files are 240 dpi. The successful one is 64 X 64 pixels, the unsuccessful one is 74 X 152 pixels. Any more ideas?

So how does this all relate to BITMAPINFO (device-independent bitmap) and is there a good way to read BMP files into that memory format? Or is that a more involved approach?

How did you draw that conclusion about the architecture? And why am I seeing mingw32 (mingw64) on the kit description? That is just as confusing as it can be!