me wrote:speaking of visiting the menu restoring things to default, check out that chest with a switch near the end of the Cave to the Sealed Gate. flicking it breaks a bridge on you. (Japanese humor?)
flicking it again restores the bridge piece (so it's a 2-way switch), as does visiting the Atma Weapon room (++). entering the menu or a battle will also do so (inexplicably), but unlike the other ways, they don't change the position of the switch. by this, i mean that if you hit the switch again after menu/battle, the bridge piece will stay complete.
woo, i finally tackled the menu/battle wrongful restoring of the bridge piece. it was much simpler than i expected; change CB/386C to 11h. that's the value that CB/386E uses when destroying the bridge piece. the value previously at CB/386C was ABh, which is what CB/3882 uses to restore the bridge piece, and afaik the original map value.
then there's the skull switch nearest to the chest. it regenerates the bridge piece for no good reason (and like menu/battle, has the chest switch position problem). it doesn't mess with the chest graphic, however.
haven't done this yet. i'm thinking that when that skull switch rewrites the big chunk of map to move the far right island and its bridge, it also overwrites the chest-togglable bridge piece which technically isn't part of the island's vertical bridge (as it's a horizontal bridge piece connecting to the stationary right land mass).
and looking closely, what it rewrites the piece with is actually a horizontally flipped version of the original piece.
i figured this out when ABh was nowhere in the data set.
to remedy the reappearance, i'd have to write a duplicate CB/3867 with an IF-test to check $1EBE, Bit 7 at the beginning, then add a call to that new function at the end of: CB/32A9, CB/33C9, and CB/3506. and i don't know where the free space is for that.
i also noticed another minor problem relating to this island's moving. it shifts two vertical squares total. when Function CB/3506 does the second shift, it inexplicably operates on a 12x5 chunk at 109,11 , instead of a 13x5 chunk at 109,10 , which is what the first half of the function does, as well as Function CB/32A9 (which refreshes the map upon return to it). so you'll see weird lava patterns (namely, duplicate "swirly circles", for lack of a better phrase) a few rows above the righthand island. they'll go away if you enter+exit battle or the menu. CB/3506 really needs to be expanded by 10 bytes to remedy this...
the needed space could be freed by replacing CB/3472 thru CB/34FD with a 4-byte call to CB/333B, saving 136 bytes (!!). it'll take me a bit to figure out how to edit the data in CB/3506, as i don't understand the format, or know exactly which tile IDs are needed.
