Io: Account for free space on the host device

[unknownbrackets]

See #13520. If there's less free space on the actual device, return that instead of simulated free space based on usage. This will tell games about the space problems properly.

Also changed the comment slightly - I'm pretty sure they released 2 GB memory sticks before that Harry Potter game was released. I don't think it was an imagination problem, just a testing problem.

-[Unknown]

[shenweip]

Oh,thanks. But this will potentially break #12761 again.
For example, if the realFreeSpace is smaller than simulatedFreeSpace, usually will get different value in the two free space check on Android, even if the contents of the memory stick have not changed.

[unknownbrackets]

How long does it need to be static? If we keep the free space constant from game start, but don't maintain it between save states, would that probably work?

Free space should take into account sector size, ideally. The byte sum of files definitely isn't what free space is reduced by on a real PSP - it uses FAT, so files are allocated in chunks of sectors.

-[Unknown]

[shenweip]

If don't maintain it between save states, I think it might not work between save states. But I think it's hard to avoid this, because even if we use simulated freespace, we can't guarantee that users won't change the content of savedata between save states.

Keeping the free space constant from game start and do maintain it between save states might work, but I worry that still report that disc is free even if the disc is full.

[unknownbrackets]

I wonder if it's better to have a game id check then, as much as I hate them... I think it's important to report low free space when we can.

Maybe it'd be good to check how this game responds to the memory stick events of removal and reinsertion. Maybe we could act like the card was reinserted when a major space change occurs. But I think some PSP games just kicked you out if you did that, or didn't let you save again...

-[Unknown]

[hrydgard]

Fixing game bugs/misbehaviors like this is definitely fair game for compat.ini checks...

[shenweip]

I'm fine with a compat.ini check, maybe this case is uncommon(I hope so, I don't hope we have to maintain a long list for this).

[unknownbrackets]

I don't think I have this game, but does that work? Not really tested.

-[Unknown]

[shenweip]

Tested simply on Android, works.