From: Tor L. <tm...@ik...> - 2000-11-10 22:22:13
|
Greg Chicares writes: > Currently sizeof(long double) is 12, and I for one wouldn't > want it changed to 8 bytes. If mingw needs that for complete > ABI compatibility, then is the -fnative-struct step toward > partial compatibility worthwhile? Good point. For gtk interoperability, -fnative-struct is essential, but sizeof (long double) isn't. > Is it needed if, instead of using the ms compiler at all, > everything is instead built with mingw/gcc? We, the people on this list, obviously prefer to use gcc (mingw). But the point is that there are lots of people who, for one or another reason, want or have to use MSVC for their code. They still would benefit from being able to use gcc-compiled generic libraries. > Even if so, why not just use the native_struct attribute in the gtk > headers? > struct foo{char a;} __attribute__ ((native_struct)); Mainly because I didn't want to clutter the gtk headers with strange stuff that would annoy the core (non-Win32) developers. Although, at that time I was thinking of having to mark each problematic bitfield with some __attribute__ which would have been much uglier. I didn't know of the native_struct __attribute__... That __attribute__ for the whole struct doesn't look so bad, it would be just one macro call after the struct declarations in question. The risk is just that you don't find all the structs that need to be marked... I prefer the global -fnative-struct. --tml |