Land on Slope

There are many things that got fixed over my holidays, and many things that remain to be done before I can claim the 3 rooms "done". I had to pick one for the last week-end ... it seemed wise to pick something that I'd likely not have enough focus to work on during the evenings of a regular week. "Land on slopes" seemed the right one to pick.

Il y a pas mal de choses qui restent à arranger avant de pouvoir cocher la case "démo avec les trois salles de test", et ça malgré toutes les choses qui ont pu être corrigées pendant les vacances. Alors pour ce week-end, pourquoi pas s'attaquer à l'atterrissage sur des pentes ?  

So far, only the "walk" (behaviour) controller is aware of slopes. For the rest of the code, the world is all made of square tiles, and the properties of a tiles are homogeneous over that square. The result is that if you try to land from a jump on sloped ground, you're very likely to end up floating over the ground until you start walking.

Jusque là, le seul bout de code qui s'intéresse au pentes, c'est celui qui contrôle le comportement "marcher". Pour le reste du moteur, une pente, ce sont des pavés carrés comme tout le reste. Du coup, si vous faites un p'tit saut pendant que vous étiez sur une pente, il y a de fortes chances que vous finissiez en flottant au-dessus du sol ^^". 

J'ai passé la journée d'hier à faire des va-et-viens entre le débuggueur et le compilateur pour essayer de comprendre ce qui se passait, mais c'est sans espoir tant qu'on ne saura pas tomber à travers les tiles qui constituent les pentes... A cogiter.

I gave it a first try yesterday, opening the debugger, seeing what happens when I'm entering such a tile, designing a patch, compiling it, trying it, discover that it wouldn't work, refine it, and repeating the cycle. Over and over.

It wouldn't work. First because the slope tiles couldn't be fallen through. There's no need to try working around it: they *must* be made fall-through. Second because slopes are currently complemented by walk-through-but-don't-fall-through blocks that ensure smooth walk, but interfere with falling. I had that discussed with the Undisbeliever in the past, as it was a difference between our implementations, and the solution will be to replace them with a 'sloped' ground that actually is square.

But the ultimate reason of my failure is that I was trying to provide a solution for 'fall on the sloped ground' as if we'd have the last row of tiles all made of F_SLOPE tiles. The reality is much more diverse: there are so many 'corner' cases that we can hardly call them 'corners' at all.

Mais même avec cette nouvelle propriété et même avec des blocs carrés qui disent que "oui, en vrai, on est des pentes, mais des pentes plates. haha", ça ne marche pas encore. La solution que j'avais prévue suppose que le sol sur lequel on veut atterrir est constitué d'une brave rangée de tiles de type "pente". Mais comme vous pouvez le voir sur mon petit croquis, il y a pas mal d'autres combinaisons qu'il faudra aussi prendre en compte. 

J'ai une idée pour remplacer ça ... il faudra que je la teste. Elle n'est pas sans me rappeler (ce que j'ai compris) du moteur de Sonic, d'ailleurs: considérer toute la colonne de tiles par-dessus la nouvelle position souhaitée pour le "hot spot" et calculer la "hauteur du sol" dans cette colonne.  

I have a replacement design sketched, which I'll give a try in the afternoon. Amusingly, it looks a lot like what (I've understood) happens in Sonic the Hedgehog engine: consider the whole column of tiles on top of the desired 'new hot spot position' and figure out the 'ground height' in that column. Then make sure the move planned so far doesn't work past the ground.

Do that every time. If done right, it doesn't matter how many sloped tiles were encountered when checking that we cando() the move.

Well, that was the plan, but for some reason, it is not yet quite working. And for some (possibly other) reason, it managed to break walk-on-slopes despite my care to avoid so.

edit: fixed