[...] since both are memory corruption things, although I really don't follow the technical side enough to say whether that's appropriate or not.
No, this submission isn't memory corruption.
Sketch is memory corruption because it forces the game to perform a task that's too large for the allocated space in the RAM; kind of like trying to empty a bathtub into a bucket the bucket just isn't big enough so the water overflows and spills on the floor. The sketch glitch is doing the same thing, but instead of water it's trying to create a gigantic monster sprite. The space given to it in the RAM just isn't big enough so it ends up writing monster sprite pixel data to other parts of the memory.
This submission exploits a programming error that allows them to create an event script that will run after the final game over. It doesn't corrupt any memory, it manipulates some values to make the event script work (step count, money, items in inventory, settings in the config menu, battle inputs in the last suicide fight, and the frame that the final game over starts).
Currently it looks something like this: The Battle inputs and frame of the final game over tell the game to go to the address for your money, (game time as well but as I understand it the game time doesn't matter for the event script they use) step count, and the first 5 items in the inventory. These values tell the game to go to addresses in the config menu, and the stuff they manipulated in there tells the game to go to the address for the credits.
Now it doesn't have to go to money, step count, inventory, or the config menu addresses; there's a range of possible addresses they could jump to. They choose these addresses for pragmatic reasons.
Yes, and how this problem is resolved by judges appears to be completely inconsistent from submission to submission and game to game as well. It bothered me in the last published FF6 movie, and it still bothers me.
I might start sounding like a broken record to a few people, but I'll repeat it again for everyone else: a softlock before the actual end screen does not count as beating the game in my opinion, and as long as TASVideos won't have an objective definition as to what counts as beating the game in the rules, I cannot accept this as a speedrun, but only as an interesting glitch exploit demonstration.
If the completing the credits is to be considered having finished the TAS, then yes this doesn't complete the run.
But I think a good philosophical case could be made that reaching the credits is a completed state, if you take examples from other forms of media. For example, do you finish watching a movie once you reach the credits, or do you have to watch the credits to say you have finished the movie? Or how about books, if you have a novel and the author left a personal message at the end of his book after the story, and you choose not to read the author's message, can you say that you finished the book?
And in speed running different communities have elected to follow different views on the matter. Eg. By SDA rules when using real time, time ends when the player loses complete control at the credits, effectively saying that that is the point when the runner has completed the game.
I'm not trying to say it should be one way or another; my point is more that it isn't as cut and dry as it might seem. I mean it's conceivable to have standards based on the kind of exploits involved in a run; which would allow for a more nuanced way of appraising submissions.
BUT ANYWAY moving on to what I was going to post before I got side tracked into thinking about what it means to finish something.
jlun2 wrote:
[...] is it possible to modify the TAS so it won't softlock?
Gyre's got the right idea, instead of jumping to the address of the credits; you would jump to the address of the cutscene where the world gets remodeled by Kefka. That would then leave you at the beginning of the World of ruin with Kutan. So the hypothetical modified TAS would have to do this whole process twice.