Refactoring of fileio.s

[mlund]

Subroutines in fileio.s overlaps with C functions:

1. mega65_io_enable()
2. lcopy()
3. Copying filename to 0x0100 could be done directly from C. Is it safe to override 0x0100 for temporary storage for setname?

Using C for the above, fileio.s becomes smaller; could be inline; and reduce compiler specific code.

Ping @ki-bo.

Notes

• Here's a snippet showing the GNU/LLVM extended assembler syntax for input:

size_t read512(uint8_t* buffer)
{
    volatile size_t len;
    asm volatile("lda 0x1a\n"
                 "sta 0xd640\n"
                 "nop\n"
                 "stx %0\n"
                 "sty %1\n"
                 "lda 0x80\n"
                 "tsb 0xD689\n"
                 : "+r"(*(&len + 0)), "+r"(*(&len + 1))
                 :
                 : "a", "x", "y");
    lcopy(0xFFD6E00, (uint32_t)buffer, 512);
    return len;
}

[ki-bo]

Copying the filename to 0x100 will often work, but be aware that this is by default the stack page, so you may overwrite the stack depending on where the stack pointer is and how long the filename will be.

[mlund]

Thanks! Mega65libc has always been hardcoded to $0100 and I wonder if there are better alternatives, e.g. by dynamic allocation. If I recall correctly the hypervisor has restrictions as to where it can be placed(?).