From http://bugs.winehq.org/show_bug.cgi?id=21276

"I believe this is a bug in the python test code, not in Wine.  Wine's
UuidCreate function follows RFC 4122, section 4.4.  That is, it generates a Uuid from a pseudorandom number generator, not from a MAC address.  Thus, the error message from the test program is correct:  it doesn't appear to be a MAC address.  On the other hand, it's not an error:  this is expected.

The python code should be checking the version field of the generated Uuid first.  If it's 1, then the generated Uuid *may* be expected to contain a MAC address.  From RFC 4122, section 4.1.6:

  For UUID version 1, the node field consists of an IEEE 802 MAC
  address, usually the host address.  For systems with multiple IEEE
  802 addresses, any available one can be used.  The lowest addressed
  octet (octet number 10) contains the global/local bit and the
  unicast/multicast bit, and is the first octet of the address
  transmitted on an 802.3 LAN.

However, the test for the most significant bit not being set is also invalid on systems with no MAC address.  From the python code:
       self.assertTrue(node < (1 << 48), message)
The most significant bit of a MAC address is set when the MAC address is a multicast address.  For systems with no MAC address, a multicast address is supposed to be used.  Also from RFC 4122, section 4.1.6:

  For systems with no IEEE address, a randomly or pseudo-randomly
  generated value may be used; see Section 4.5.  The multicast bit must
  be set in such addresses, in order that they will never conflict with
  addresses obtained from network cards.

This analysis may of use to the Python folks, but it certainly isn't a Wine bug."

More info from the wine bug:
Whoops, I didn't notice that Python uses UuidCreateSequential.  From MSDN:

  For security reasons, UuidCreate was modified so that it no longer uses a machine's MAC address to generate UUIDs. UuidCreateSequential was
introduced to allow creation of UUIDs using the MAC address of a machine's Ethernet card.

I'm reopening this, as Wine's implementation of UuidCreateSequential calls UuidCreate, whereas it should use the MAC address-based method.  Setting component to rpcrt4, too.

My comment about the multicast bit occasionally being set (when the machine on which UuidCreateSequential is called has no MAC address) still stands, so there's still a Python bug.

The multicast bit is the least significant bit of the first octet.

Could you check if the patch for http://bugs.python.org/issue3581
fixes your problem?

From Juan Lang, who originally pointed out the bug:

My only comment is that the message is now misleading:
-        individual_group_bit = (node >> 40L) & 1
-        universal_local_bit = (node >> 40L) & 2
        message = "%012x doesn't look like a real MAC address" % node
-        self.assertEqual(individual_group_bit, 0, message)
-        self.assertEqual(universal_local_bit, 0, message)
        self.assertNotEqual(node, 0, message)
        self.assertNotEqual(node, 0xffffffffffffL, message)

The test no longer checks for the uuid being a MAC address or not.
Ideally it would be reworded to be less confusing, or perhaps the code
could check the various "flavors" of uuids, and check that a generated
uuid conforms to one of them.

Fixed in issue 3581. Thanks for the report and the comments!