LOS Through Wall

[Griogre]

I've noticed some reproducible issues with seeing through walls. For the screenshots here I used DM preview, but these same things occur with players connect using normal LOS in a game (where I and the player's first noticed the issues). The screens are from the development campaign which I will zip up and send along with the unreleased map pack used to make the maps.

You can only see through the walls when a token is in specific squares. In this case to reproduce just drag a PC token from the combat tracker. When you move the token around after you can see through the wall, you will get FOW where the token could originally see through showing it can't see through in that square.


The LOS01 series is from map 01a Dungeon, Room 01-07 Guardroom.
The LOS11 series is from map 01b Dungeon, Room 01-10 Plank Room (next post).


In the LOS15 images I noticed something different. If you dragged the PC token from the combat tracker an placed it in the SE corner as shown you did not see through the wall initially as shown in 15a. But if you clicked the token again without moving it to unselect it, as shown in 15b and then clicked it again to select it as shown in 15c you could now see through the wall.

[Griogre]

Next Screens

[Zacchaeus]

I'm guessing it has something to do with having the occluders on different layers. Do you get the same issue if you put all the occluders on the one layer?

[Griogre]

It might be, Zacchaeus - I'll have to check but since the layers are geomorphic it would defeat the long term purpose to move them all to the same layer.

Last screens

[Moon Wizard]

Can you please provide the campaign with the issue; as well as the steps to find the map and location on the map in question?

Thanks,
JPG

[TVDinner]

I recognize that one.

See here (https://www.fantasygrounds.com/forums/showthread.php?67311-LoS-bug-see-thru-wall) for a module.

I haven't checked in a few months, but I don't think it was ever fixed.

[Griogre]

Can you please provide the campaign with the issue; as well as the steps to find the map and location on the map in question?

Thanks,
JPG

I sent the campaign and associated map pack to customer support, a bit ago. The images are all in the Sumuta heading. The maps have a number of places where this happens, I only picked two. It's usually where there is a shared wall between room or hallway areas.

[TVDinner]

I've seen it in dual-LoS walls as well, but the second wall stops the error. With the usual way the map backgrounds are drawn, the little error triangle is over black and the players don't see it.

Just a WAG but I'd look at a rounding error for the intersection point or comparing floats for equality, but deleting or moving the non-related LoS points shouldn't change things. I'm happy someone else sees it.

I couldn't just cut out everything from the original module but the one room and have it fail. It really seemed to depend on the other LoS points nearby, or maybe I just needed to sacrifice another chicken or something.

[Moon Wizard]

I've escalated to @cpinder again.

Regards,
JPG

[Griogre]

Just a WAG but I'd look at a rounding error for the intersection point or comparing floats for equality, but deleting or moving the non-related LoS points shouldn't change things. I'm happy someone else sees it.

I wondered about rounding errors as well. It does seem to be some sort of close distance association between points because I haven't seen it in places of maps where there are not any shared walls, ie there is 10' of space between walls. My wild *** guess is something similar to TVDinner's, except I think a sign (reg shift?) may be reversed in a portion of the LOS calculations because of those triangle view slices.

[DCrumb]

From what I see as a guess, is that the intersection point is actually two points, and you would have to zoom in to the intersection to see what those two points actually look like.

[TVDinner]

Ha, apologies, Carl, enjoy. :) I always *loved* programmers who told me what was wrong with my code when they've never even seen it. Doing so must be one of those irresistible things in coding. I would be curious to hear the final verdict though, knock knock.

[Moon Wizard]

I had a talk with @cpinder; and this is a very low-level piece that will require a large rewrite that we may not have time to do for a while.

His theory is that this has to do with exactly parallel blocker lines that are fairly close to each other. As a workaround for now, he suggested trying to insert extra points in the problematic lines or slightly shifting one of the end points.

Can you give that a try?

Thanks,
JPG

[TVDinner]

I putzed with the old test module and got it to error with only three segments. One secret door, vertical; one wall, vertical below secret and on same line; one wall starting at the join of the other two segments and heading east (but not perpendicular anymore). Same procedure as the first module, orc goes above the horizontal wall.

Opening and closing the secret door triggers and fixes the error. Moving either of the end points of the vertical segments horizontally fixes the error. Moving the stray endpoint of the 3rd segment makes no difference. Walking the orc around the whole thing can fix it, but playing around brings it back.

I use the keyboard to place the points by hopping around the grid, so my LoS blockers are for the most part always parallel or at right angles.

New module attached for grins.

I think this is what Carl is talking about, since offsetting either of the two end points works around the error. I can live with that, although I bet at least one of my players will notice the offset and figure out the secret door is present.

Thanks.

[Griogre]

To follow up on this. I re-did the maps to remove areas having shared walls. This appears to have fixed seeing through the walls though I know more after the games this weekend. To follow up on Carl's theory, while the walls, doors, ect are still all parallel and perpendicular rooms and/or corridors no longer have a common wall anywhere in the dungeon.