Dear Anatolij,
In message 20100629142343.7ecadfaa@wker you wrote:
'memcpy_align32wrap' isn't really good since the fixed memcpy also fixes 16-bit accesses, too.
we discussed this a bit more internally, and could not find a good name yet, either. Guess I'll stick with the current one until someone comes up with a really good idea.
BTW, shouldn't we fix print_buffer() also? do_mem_md() doesn't use memcpy() and the issue with corrupted dumps still remains here. I'm testing a patch to fix it. Will submit it soon.
I though about this, too. But then I decided against it. "md" is intended to to exactly what the user requests - as is, we can easily demonstrate the issue. I consider this an extremely useful debug feature. If I command the memory to be read in units of 32 bits I really mean that. A "fix" ins print_buffer() whould hush up the symptoms, so the problem would be much harder to detect.
Let's keep this as is, please: if you need a memory dump, you can either align your start address, or use "md.b".
Best regards,
Wolfgang Denk