07:28 <dinkelhacker> borneoa_: Thanks for answering in my absence (I have to get a proxy that keeps me logged in).

07:30 <dinkelhacker> borneoa_: That sounds like it could be fixed rather easy. On the other hand I don't know the OpenOcd code base at all. That beeing said, if you think this is a low hanging fruit and could be a good first contribution I'd be willing to give it a try.

07:31 <borneoa_> dinkelhacker: there is a patch pending in gerrit that fix this for aarch64 state only. If the core is in aarch32 the patch doesn't add anything

07:32 <borneoa_> dinkelhacker: https://review.openocd.org/c/openocd/+/7192

07:33 <borneoa_> dinkelhacker: would be great if you could test it and vote it in gerrit.

07:44 <dinkelhacker> I'll take a look!

08:06 <dinkelhacker> borneoa_: Just out of curiosity. The patch constructs the 64 bit value by two reads with mem_ap_read_atomic_u32. Isn't there a way to read a double word in one go?

08:11 <borneoa_> dinkelhacker: I didn't check it. On armv7 and armv8 the way to transfer a CPU register to the debugger is through a memory mapped register that the debugger access with mem_ap_read/write. Maybe the register is a pair of 32 bits? Also not all DAP support 64 bits and there is another patch in Gerrit for 64 access when available.

08:38 <dinkelhacker> borneoa_: You're right. According to the docs the MEM-AP only supports up to 32bit accesses. https://developer.arm.com/documentation/ihi0031/a/The-Memory-Access-Port--MEM-AP-/MEM-AP-functions/Variable-access-size-for-memory-accesses?lang=en