Player: Test if saves created before the LSD fixes are still loadable

[Ghabry]

When attempting to load saves written by EasyRPG Player the newer versions report multiple issues in the old saves, e.g. when loading a Vampires Dawn save I get:

Reading Primitive of incorrect size 5 (expected 1) at E39
Warning: Corrupted Chunk 0x3d (size: 5, pos: 0xe39): music_stopping : Read 9 bytes! Reseting...
Warning: Corrupted Chunk 0x22 (size: 1, pos: 0x2ccd): parallax_vert : Read 2 bytes! Reseting...
Warning: Corrupted Chunk 0x23 (size: 1, pos: 0x2cd0): parallax_horz_auto : Read 2 bytes! Reseting...

This is all not really critical but we should check, if there are any old savegames that break before pushing new APKs to Android e.g.

[fmatthew5876]

Once #1502 is merged, old save games loaded in new Player will make all events walking in place, including the player.

Changing maps would for fix events, but player animation_type may be stuck on continuous forever.

[Ghabry]

That's problematic. Considering that the Player animation_type can't be altered by event commands: How about changing Player animation_type to the correct value upon loading?

[fmatthew5876]

I added a compatibility hack to #1502. We'll force animation_type to non-continuous for player and vehicles on any loaded savegame.

[Ghabry]

Old savegames will load with incorrect pan, the hero is at the top-left of the screen and the camera won't fix itself until the map is changed.

[fmatthew5876]

Old savegames will load with incorrect pan, the hero is at the top-left of the screen and the camera won't fix itself until the map is changed.

This will only affect save games with non-default panning offsets. The defaults are represented by missing chunks.

[Ghabry]

Save01.zip

This is a save for TestGame-2000 created with Player 0.5.4.
In master it will load with wrong Pan but in RPG_RT it is correct.

[Ghabry]

Save05 is an old save and after loading the hero walks veeeery slow.

Off-saves.zip

Save10 is a new save created with master

But in RPG_RT both saves result in a fast hero

[carstene1ns]

Another broken thing currently with old savegames: player position is messed up.
game: standstill girl savegame: sg-panorama.zip

This has been fixed by #1554.

[fmatthew5876]

Ok I understand this now. In that save game we have this:

map position_x 3584 (14)
map position_Y 3328 (13)
player_x 23
player_y 20
pan_current_x 0
pan_current_y 0
pan_finish_x 0
pan_finish_y 0

In master we recompute the map position from the panning offsets every time. This is wrong. In the PR, we load the map offsets from the save file and respect them.

The panning offsets are still wrong here, we just don't see them take effect. This is a looping map so I can't test, but I imagine as you reach some of the edges of the map, the pan would start to lock back into (0,0).

EDIT: I tested further (and it was hell, I had to hexedit Map0069.lmu otherwise the game wouldn't play..)

Change the map to not looping horizontal. In RPG_RT the panning offsets are cleared when loading the game, you can see it by saving again. In Player (PR), if you walk right everything works, but if you walk left, the screen will stay stuck until you reach the left edge. That's because we don't clear the panning offsets.

Cherry mentioned somewhere the RPG_RT has a bug with loading panning offsets. It clears them in some cases. Maybe this is one of them.

EDIT2:

Yes I've confirmed RPG_RT loads the map position and then squashes the panning offsets. Load Save01.lsd and then immediately save to Save02.lsd.

In Save01.lsd, the panning chunks are (0,0), in Save02.lsd they will be missing (default 9,7).

We could emulate this behavior, but then we reproduce RPG_RT bugs where it doesn't correctly resume scrolling.

TLDR version:

Old save games with (0,0) panning chunks will work fine in looping maps. They will only break when you start moving in non-looping maps.

[carstene1ns]

All known cases seem to have been fixed.