Andrew Haley <aph-gcc@xxxxxxxxxxxxxxxxxxx> wrote: > > It would work on gcc, because gcc allows members of a union to be > written as one type and read as a different incompatible type. Other > C compilers allow this too. However, it's is not guaranteed by the > language standards, so it doesn't much help you with truly portable > code. Right. It is also a bad idea to use 'register', as some compilers incorrectly use it to change representation - or they used to. The safest and most portable code involves setting a variable to a known value, reading each byte of the variable as an unsigned char, and using AI to deduce the endianness. For integers in C99 or on any system with a fairly conventional binary representation (don't ask), you can generate a translation table directly by storing (say) (0<<24)+(1<<16)+(2<<8)+3 and using the result as unsigned char [4]. Doing the same for IEEE 754 formats is left as an exercise for the reader :-) I have done it often enough, but it is tedious. If you are faced with doing this for a system with integers that don't match C99's model, or with floating-point that doesn't have a base of 2 or 16, real numbers that are not floating-point or anything similar, then please contact me by Email before I retire. That isn't entirely a joke :-) Regards, Nick Maclaren, University of Cambridge Computing Service, New Museums Site, Pembroke Street, Cambridge CB2 3QH, England. Email: nmm1@xxxxxxxxx Tel.: +44 1223 334761 Fax: +44 1223 334679
