UPDATE: the code below DOES work to dereference the pointer. I had incorrectly inserted some lines at the entry point that had the effect of overriding the memory location f1_ptr. The important part is that to defererence the pointer when it's stored in a memory location is: mov r15,qword[f1_ptr] / mov rdx,qword[r15+0]. Move memory to r15, then move r15 to rdx. That does it. But as Peter Cordes explains below, memory locations are not thread safe, so best to use registers for pointers at least.
****End of Update****
I am using ctypes to pass a pointer to an array of pointers; each pointer points to the start of a string in a list of names. In the Windows ABI, the pointer is passed as the first parameter in rcx.
On entry to the program, I ordinarily put pointers into memory variables because I can't keep them in the low registers like rcx and rdx; in this case, it's stored as mov [f1_ptr],rcx. But later in the program, when I move from memory to register it doesn't work. In other work with simple pointers (not pointers to an array of pointers), I have no problem.
Based on the answer to an earlier question (Python ctypes how to read a byte from a character array passed to NASM), I found that IF I store rcx in another register on entry (e.g., r15), I can freely use that with no problem downstream in the program. For example, to access the second byte of the second name string:
xor rax,rax
mov rdx,qword[r15+8]
movsx eax,BYTE[rdx+1]
jmp label_900
If instead I mov r15,[f1_ptr] downstream in the program, that doesn't work. To emulate the code above:
xor rax,rax
mov r15,qword[f1_ptr]
mov rdx,qword[r15+8]
movsx eax,BYTE[rdx+1]
jmp label_900
but it not only doesn't work, it crashes.
So the question is: rcx is stored in memory on entry to the program. Later I read it back from memory into r15 and dereference it the same way. Why doesn't it work the same way?
The full code, minus the code segments shown above, is at the link I posted above.
