I think it has been covered, but to amplify a bit, the two methods of distance measurement are subject to completely different error problems.

A bike computer that uses wheel revolution data can be extremely accurate if it has accurate wheel circumference data. Any error in this input will be cause a corresponding error in distance traveled, but it will be a consistent error that has almost no trip-to-trip variance. With a little care, you should be able to measure wheel circumference very closely (within a fraction of a percent), but I've always found the wheel circumference data that comes with the computer to be fairly accurate as well.

The GPS odometer has a more subtle error problem - the "jitter" in how it computes position. The computer will compute distance by sampling its position at some sampling interval and adding up all the distance intervals to get total distance travelled. The problem is that GPS position has jitter errors that cause position to "jump around" a few feet in random directions so while you may be traveling in a more/less straight line, the computer sees you making little zig-zags that add distance. On a curvy route, the computer will "cut the corners" and estimate a distance that is less than the distance actually travelled. And finally, GPS signals are subject to electrical interference, poor geometry, and issues of signal blocking (tunnels, tree coverage, buildings, etc.) that may cause momentary glitches with huge position errors, perhaps tens of miles. If the computer thinks you momentarily travelled a mile off your route and returned a few seconds later, it could add two miles to the distance.

All these GPS errors can be somewhat compensated for by averaging and filtering software as well as adjusting the sampling interval based on signal condition. A bicycling GPS may be optimized to minimize these errors for typical bicycle speeds. How effective they are with these techniques, I don't know. The slow speed of bicycles makes the problem harder vs. cars since the positional errors are larger with respect to the distance and speeds of a bicycle.

My experience is that GPS odometers do a "fair job", but can't compare in accuracy to a decently calibrated bicycle computer which uses wheel revolution information.

- Mark