Miro Kropáček

So what would you suggest in this case? `static_assert` worked as expected (and I'm grateful for that) but I wasn't sure how to fix it. When "real" sizeof() is 8,...

> any current generation platform is stuck with such an old gcc It's not any current generation at all. ;) It's a 30 years old hardware. > the typical manifestation...

As discussed in the linked issues, this is tied to `STRUCTURE_SIZE_BOUNDARY` setting in gcc which most platforms have set to 8 (1-byte = no padding) while some platforms have this...

Yeah, that is what I meant by the need of some kind of private tracking: the kernel would need to track which one was the first, second etc. This way...

> Although i've ported that to my gcc 4.6.4 branch, i think it won't be easy to do the same for newer gcc. Look at his gcc-7-mint-fastcall branch (linked above),...

> That means you have trouble compiling code where cdecl calling convention is required, like when calling the USERDEF function in AES, calling SLB functions etc. That's possible but since...

My ScummVM doesn't use SDL & friends, so that's not a problem but yes, I see what you mean. Every lib, even mintlib, would need to be carefully checked where/whether...

For me it causes trouble when installing software: I have m68k-atari-mint-gcc and m68k-atari-mintelf-gcc installed side by side and to avoid changing --prefix paths on many places I change just `CC`...

Btw specifying `--with-sysroot=${INSTALL_DIR}/${TARGET}/sys-root` is usually enough. `--with-build-sysroot` is useful, too but I use it only for native builds.

Oh and my report is from WSL's ubuntu.