Sharing my thoughts about the changes:
A proper review is some effort when done properly (checking the codestub in different languages - does it try to parse an int while a long is needed? is there a trailing whitespace?).
Are given constraints met? Do testcases cover the most relevant cases?
It can turn into a lot of effort for SDK contributions (e.g. trying to crash the referee, check for proper seeding to reproduce a game under same conditions, …)
Personally I think that honoring the effort was justified. I can’t tell much about XP harvesting and if it was a problem, as I don’t follow the puzzle contributions too actively.
I also disagree on the reward for new contributions: each gives 250XP. It takes me a full weekend to write a game like Onitama, even after learning the SDK (graphics take a lot of time). In the same amount of time I could spam multiple CoC contributions, each rewarding the same about of XP as that single game.
I don’t even care about having or not having the XP itself, it’s just the balancing that feels off to me.
That’s great! I change small details from time to time (e.g. removing trailing spaces in the testcases, thus touching them ), was always wondering why the update is that slow.
Quelques coquilles en français :
Dans Entrée pour le générateur de stub, il y a un x erroné et un e en trop (à moins de traduire string par chaîne).
Un string peut contenir des espaces, à l’inverse d’un word. Donc seulement un string peu_t_ être déclaré_e_ dans une même suite de variables.
Au lieu de stub, on peut utiliser squelette ou modèle pour franciser.
Tout en bas de la fenêtre, il y a un s erroné, c’est Statu_t_.
I don’t know if it’s actually related to this update, but the following contribution has an official solution that doesn’t output anything and has managed somehow to bypass these tests.
Because this solution is also wrong (it’s a faulty approach to the corresponding problem) it’s likely to be modified by the author soon, so I let a copy here. If you superficially modify the tests or this solution right now, the failure will be detected and the update will be rejected, but it might be interesting to know how this got through in the first place.