probe: stlink: Fix 1-byte transfers

Internally the 1-byte transfers are handled in 3 phases:
1. read/write 8-bit chunks from every unaligned addresses until the
   first aligned address is reached,
2. read/write 32-bit chunks from all aligned addresses,
3. read/write 8-bit chunks from the remaining unaligned addresses.

Size of the first unaligned read/write is set to the result of address
(4-byte) alignment check and can be either 1, 2, or 3 bytes (the value
of `unaligned_count` calculated as `addr & 0x3`). This is incorrect and
every transfer with the requested size smaller than `unaligned_count` is
terminated with the following error:

  Unhandled exception in handle_message (b'm'): result size (3) != requested size (1) [gdbserver]

Skip the first unaligned transfer if the requested size is smaller than
the result of the address alignment check. Handle the whole request in
the second, unaligned read/write.

pyocd/probe/stlink_probe.py

@@ -329,8 +329,8 @@ def read_memory_block8(self, addr: int, size: int, **attrs: Any) -> Sequence[int
         res = []
 
         # read leading unaligned bytes
-        unaligned_count = addr & 3
-        if (size > 0) and (unaligned_count > 0):
+        unaligned_count = 3 & (4 - (addr & 3))
+        if (size > unaligned_count > 0):
             res += self._link.read_mem8(addr, unaligned_count, self._apsel, csw)
             size -= unaligned_count
             addr += unaligned_count
@@ -355,8 +355,8 @@ def write_memory_block8(self, addr: int, data: Sequence[int], **attrs: Any) -> N
         idx = 0
 
         # write leading unaligned bytes
-        unaligned_count = addr & 3
-        if (size > 0) and (unaligned_count > 0):
+        unaligned_count = 3 & (4 - (addr & 3))
+        if (size > unaligned_count > 0):
             self._link.write_mem8(addr, data[:unaligned_count], self._apsel, csw)
             size -= unaligned_count
             addr += unaligned_count