karmic cloud
Hi, I created a chapter to play through a game using the interactive lesson feature in a study.
For this chapter: https://lichess.org/study/lTNKB0q1/rvrJIXZm Lichess tells me to retry 10. O-O-O even though this is the correct move. I've tried inputting this move in a few different ways and I get the same result. Note that I play through the game by using the "Preview" function. Thank you.
empty wharf
huh indeed
empty wharf
Issue here : https://github.com/lichess-org/lila/issues/12206
cc @coarse sun Possible something related to scalachess?
coarse sun
possibly, let me check
oh, so the castle move is valid but not accepted as solution. It's probably not related to scalachess because the rules are still correct. But I'm not 100% sure.
coarse sun
It seems any interactive study with castlings have this issue 😱
empty wharf
Oh 🤔 i did try to repro using a different move order first, but was not successful
coarse sun
also, it happen only with import pgn. If you input moves by yourself, it works 😄
empty wharf
ah 😄
coarse sun
so it is related to how we read/write pgn maybe from database 🤔
scalachess looks suspicious now
but in the other hand, if I imported the game using lichess.org/import it works, so 🤷♂️
karmic cloud
How would I use the import function to copy it to my study? Or is the best way to add the game at this point to manually input moves? Thanks for looking into this.
empty wharf
If you input moves by yourself, it works
You could do this. However, you could also just wait until it's fixed unless you have a great urge to study right now
karmic cloud
Nope, I can wait. Thanks!
toxic zephyr
it's because of 960 uci notation
lichess always uses that but seems like the imported one doesn't. so the expected move is e1c1 but the actual move is e1a1. which are the same move but the notation is different.
probably indeed scalachess i'd assume
coarse sun
I tried both e1a1 and e1c1, both didn't work
toxic zephyr
you mean on the board?
coarse sun
on interactive study
toxic zephyr
yeah, but what you mean is you tried moving the king to a1 and c1? that doesn't matter. the move will always be e1a1 exactly so that it can compare by notation.
coarse sun
ah, yes, that what I mean
toxic zephyr
it looks like it's also using the wrong notation on https://lichess.org/jkdqipAw (when checking the underlying data) which should be older than the changes so probably it's somewhere while serializing the moves
coarse sun
what is wrong in that game?
toxic zephyr
well, nothing noticeable. but if you check the json, it also uses the wrong format for the moves.
and actually, if you manually make the long castle on the board at the same time where it happens in the game, you get a variation
coarse sun
so there is some different between imported pgn and manually input moves for castle
toxic zephyr
i think it's probably when generating moves from the db data. since i don't think uci is stored there and otherwise the game above shouldn't have the wrong ones.
so probably around here: https://github.com/lichess-org/lila/blob/master/modules/round/src/main/TreeBuilder.scala/#L26
but i guess fundamentally in scalachess
coarse sun
yeah, probably some changes on scalachess. Similar to the enpassant bug.
I will look more into that later today. Around lunch.
toxic zephyr
https://github.com/lichess-org/scalachess/pull/349 should fix it i think
haven't tested it yet