I have never seen such confusion over something that has been proven. The only completly secure classical encryption method is the one time pad. It has been mathematically proven and is also quite intuitive as well.
As far as quantum key distribution is concerned, it is also theoretically proven to be "uncrackable." The reason for this is, if after comparing a sample of the sifted keys, Alice and Bob can then determine if their mutual (Shannon) information is higher than that of Eve's (by noting the error rate). If it is, they continue, if not, they ditch the key and start over. The simple fact that Eve introduces errors by merely examining the qubits is what makes QKD secure. Too further lower Eve's information about the key, Alice and Bob simply implement well known classical secrecy amplification algorithms. Just check out LANLs preprint server, there are tons of papers that go much more in depth into how it works.
However, "photon guns" that shoot out pulses of exactly one photon are not feasable therefore Eve could use beamsplitting techniques to intercept the key, but I am not all that well versed on this method so I'll go no further. Needless to say, cryptologists are working on methods of cracking QKD, but most of these methods attack the flaws in the technology we are forced to used in implementation, not the physics of QKD itself.
Also, to the guy who said anything that could be decrypted could be cracked: get real. Someone already explained it, but I'll reiterate. The one time pad makes a message secure because if one attempts to decrypt it one could come up with anything of equal length. If you think you can crack the one time pad by all means, have at it. If I were you, I would stick to things in the realm of the possible.