64bit arguments passing with 68k .library ABI about amissl

[Futaura]

@chainq Compiler magic :-)

Basically, for GCC/OS3, it is implicit that the register specified is the first of the pair if the type is 64-bit. I had to double-check this by disassembling as I couldn't remember, but I can see that GCC is correctly passing the high bits in d0 and the low bits in d1, for OPENSSL_init_ssl(). If you look at the LP2() macro itself, it puts the uinit64_t macro parameter into a variable of that type in register d0 (and d1 implicitly). I don't think GCC has a way to explicitly specify d0-d1, as you would do in VBCC for example.

[chainq]

@Futaura thanks for the answers. Things are more clear now I think. We're going to add support to specify register pairs for 64bit values in Amiga-style .library calls in Free Pascal, so the AmiSSL headers can be completed.

Actually, meanwhile I learned, that one needs this for some of the double precision math libraries as well. We didn't support this, because Free Pascal provides its own soft-float facilities instead for FPU-less systems, so supporting the old and fairly limited math libs wasn't a priority. So yeah, that could then work too. Closing the ticket, thanks.