Oops. That program isn't supposed to crash.
Okay. That fixed it.
Crashed again . . . but this time it looks like it found an inconsistency first. Ah, there's the problem! Okay. Back to the test data.
IT WORKS! And that one works also! Hey, I bet these outputs would look a lot better if I added vertex display. Ooooh, pretty. Constellations. (note: once I've finished the next step of the algorithm, I'm planning to post a few of these outputs on my LJ - they *are* pretty.) Time for a real-world case.
Ugh. That's a new way of breaking. More debug output!
. . .
More debug output!
Hmm . . . it almost looks like it's folding inside itself . . . Wait, I've been working on the assumption that loops have the solid part on the *outside*. What if it's getting the solid part on the *inside*?
*proceeds to sketch on paper for several minutes*
Well, if *that* happens, this error correction code would just be making things worse . . . so the obvious solution . . . is to detect which one of those is the case, and reverse the error correction code to push it past the singularity point!
I wonder what game is running on the MAME box right now.