I had a look and I don't know a way to fix it while using compressed textures. It is 16 bit colour banding due to the conversion to DDS. You can see it in the editor too if you switch on compressed textures.

I have a few too many things to do, so can't really investigate further at the moment. I'm not sure but maybe it could be better to go for a plain colour as the best compromise within the current system.

The bug is already fixed so I don't think it's worth me spending time trying to figure out how to display it on the hosting pages.

Best thing is for me to finish the updates and we can release version F, problem will be solved.

It is possible, but not easy, for me to find out.

I think you should be able to just stop the host and restart it one minute later, but let me know if that would cause a problem.

No, these ex-commercial accounts cannot be gifted. LFS licenses are not transferable.

I don't know how they have escaped from the original owners, but I have reverted the licenses to demo status.

Apparently that is no longer the case. I mean, you don't appear to be online now.

Maybe the server had been running for 49.7 days and got hit by the known bug at that point, but has now been restarted?

Quote from bayanofmansorofisky :

Auto assign layer : tris generated from Extrude or Lathe should be automatically assigned to the active layer (if layer 0, the main one is turned off).. or layer could be determined by shift+rmb prior to function.

Can you explain this? I tried extrude or trace and it did use the currently selected layer, as I would expect so I'm not sure what you would like to work differently.


Related to this, I thought of a possibility specific to extrude (not lathe). The trace line is often along existing triangle edges. I think it could detect this and use all the properties of the triangle that shares the edge (separately for each line in the trace). Smoothing group, mapping, layer.

Yes, I have been thinking of this too, actually have already asked Eric about it.

The only issue I have, and it's not a big problem, just which point to use as the origin of the automatically created "break off" object. I think it should be one of the points belong to the selected triangles, or somehow related to them.

Maybe a yellow button if there is one, or a point that is low down in the selection. You might think of something like a point that is in the centre of the selection, but this could risk losing precision. Maybe the nearest point to the centre, or the lowest point?

Just thought I'd ask this in case anyone has a good idea for that. It's not that big a deal because you can adjust it after breaking off, but a good choice of object centre might be helpful, I guess.

Yes, these DXT formats are interesting, as a special kind of compression that can be used directly by the GPU without decompression. Well, that's not quite true - I suppose they are decompressed at the moment they are read into the pixel shader, but not in advance.

Because the compression is only in blocks of 4x4 pixels, which is a completely different type of compression compared with JPG or PNG (which are also very different from each other).

The result of this is that they are even faster than fully decompressed textures because of the memory bandwidth reduction.

Quote from r3zp3k7 :

Allow JPEG image formats for mod textures. Keep DDS conversation only on PNG textures (although I would also add PNG conversation to JPEG for non-ALP/non-transparent textures).

I've attached two files, both converted into DDS end up in the same 683KB size. You can check quality/size differences on websites like this one (was the 1st one I saw in the search console): https://www.diffchecker.com/image-compare/

PNG - 1355 KB
JPEG - 113 KB

Even if textures could download as jpg they would have to be converted to dds on arrival, as jpg takes a lot of CPU to extract each time, and doesn't contain mip maps etc, so there's quite a bit of processing to get them ready for realtime use. Basically for in-game use it must be a dds and must be saved locally that way, so it can be loaded and sent directly to the GPU without any conversion.

For your download comparison, the best thing to compare is a jpg vs a dds *after* it is compressed in a .7z file. DDS look a lot bigger than a JPG on disc but the DDS often compress well in a 7z file, so the download saving might not be as much as you think.

I don't mind being proved wrong - I'm just stating my current understanding of it, without doing any detailed analysis.

Quote from Aleksandr_124rus :

...maybe I need to do something differently but either I don't understand how it works, or it doesn't have a significant effect.

The NCL can only have an effect when used within the same smoothing group. In that case it will give a higher weighting to the triangles with the higher number. So it affects only the normals at the boundary between two normal contribution levels, within the same smoothing group. But across smoothing group boundaries, separate triangles do not contribute to the same normal anyway so then it has no effect.

But I wondered about the slight shading errors, and had a look in game and in editor. I tried "flat" view mode, because flat shading shows a good representation of the triangle normals, that produce the raw data for the vertex normals.

As you can see in flat shaded view, there is a sort of dent here (attached image). Because of how LFS works, that dent is the cause for the undesirable normals at that location.

I forgot to mention flat shaded mode. That is another thing that Eric uses a lot. I think that by moving vertices around a few mm here or there, the flat shaded model can display more consistent lighting and that will affect the vertex normals in a good way. I don't know if you can use the same thing in blender (I know blender supports non-triangle polygons so I don't know).

EDIT: I just remembered, sometimes triangle rotations can also have a quite a noticeable effect on the normals. I mean rotating two triangles within a quad (select 1 triangle, SHIFT+click the adjacent triangle). The LFS importer doesn't always make the best choice of triangle rotation when creating two triangles from a quad in the imported obj file.

Last edited by Scawen, Sat, 30 Mar 2024 14:39.

Quote from Aleksandr_124rus :

the shading algorithm always sets its own direction of normals. And it doesn't always look good. And I don't know how to deal with it. In blender the shading looks different than in LFS Editor, and in LFS game it is even more different.

I haven't seen the specific case you are talking about but there are some tricks you can use.

First to understand:

- the normals at a vertex are based on the directions of the connected triangles, within a smoothing group.
- the normal contributions are higher from the larger triangles meeting at that point.

You can adjust the normals in two main ways:

- extra triangles may sometimes be needed
- use "normal contribution level" (ncl)

I have a feeling that a lot of people don't know about normal contribution levels, although Eric uses them a lot. You can use it to increase the importance of some triangles, relative to others, regarding their contribution to normals.

Actually, triangles with ncl of zero will not affect the normal at all if there are triangles with a higher contribution level also connected to that point.

You can view and assign the ncl in a similar manner to smoothing groups.

I always imagined that type of central deadzone could be set in the controller's software, though I don't think I've ever tried. It seems like a memory from the distant past. Is that not the case, or at least not convenient?

EDIT: Maybe that's only in the Logitech software, but not e.g. Microsoft? Just speculating.
https://www.lfs.net/forum/post/1721869#post1721869

Last edited by Scawen, Sat, 30 Mar 2024 10:41.

And not only the interface, also the setup file format. Other things are already shoehorned in there. The 7th gear ratio is in a completely different position in the file, where it was queezed in to support the BF1 (I think).

So it's not an easy thing to add at this point, although it's an obvious thing to support. The larger setup file requires version support in the setup file reader, an adjustment to multiplayer packets and ability to read old MPRs with old style setups in them, etc. "Changing a number" is far from the reality. Better to do it at the same time as some other additions to the setup file.

Well that's just for interest and to explain why it's way too difficult to do right now.

My comment earlier was not in opposition to the suggestion, but about the statement that "the majority" of modern vehicles have 7 or more gears.

That statement seemed very unlikely, and I did ask if somehow this change in technology had passed me by unnoticed. I prefer suggestions to be based on reality rather than fiction.

EDIT: I think what I mean to say is I would prefer to be persuaded by calm reasoning rather than exaggeration.

Last edited by Scawen, Sat, 30 Mar 2024 14:44.

Sure but I have my doubts that 7,8,9 speed gearboxes are fitted to the majority of modern vehicles. I'm happy to be proved wrong though.
Until then I'll just go on believing that 5 and 6 speed are the usual thing.

Some of these suggestions are now done in Editor E7, including the new line guide as described.

Also a step towards being able to transfer selected triangles between subobjects. You can now merge a subobject into another subobject. So you could break off triangles from one subobject, into a temporary subobject, then merge that one into another subobject. I realise that sounds laborious but in some situations it could be more convenient than merging everything back to main object and breaking off again. Also, the eventual function to transfer triangles might well use an intermediate subobject invisibly, using this new function as part of the operation.

So I think it's an important new function and seems to work well as far as I have tested, but would like to hear if it works as expected for you.


Editor Patch E7:

Drag selection box:

Function to make drag box from points now available in points mode
Scale / rotate / shear axes are now hidden while making a drag box
Grey button shows drag info to avoid left buttons moving around

Typing individual values to align points:

ALT + click r/f/u numbers to set all selected points to same value
- relative values compared with last selected point not preserved

New modeller line guide:

Set up a trace then click "guide" to create a line guide
- the guide is visible in most editor modes and all views
- options to hide/show the guide or draw it open/closed

Merging subobjects:

You can now merge a subobject fully into another subobject
- this is a step towards moving a selection between subobjects
- triangles can now be transferred via an intermediate subobject

https://www.lfs.net/forum/thread/106792

Editor Patch E7:

Drag selection box:

Function to make drag box from points now available in points mode
Scale / rotate / shear axes are now hidden while making a drag box
Grey button shows drag info to avoid left buttons moving around

Typing individual values to align points:

ALT + click r/f/u numbers to set all selected points to same value
- relative values compared with last selected point not preserved

New modeller line guide:

Set up a trace then click "guide" to create a line guide
- the guide is visible in most editor modes and all views
- options to hide/show the guide or draw it open/closed

Merging subobjects:

You can now merge a subobject fully into another subobject
- this is a step towards moving a selection between subobjects
- triangles can now be transferred via an intermediate subobject

https://www.lfs.net/forum/thread/106792

Quote from dh7938726 :

As majority of modern vehicles carry 7 to 8 or even 9 speeds

I'd like to see the stats that bring you to this assertion!

In your third screenshot, if you press the pedals, do the axes (blue and yellow) move?

If they do then you just need to assign them to the correct input.

EDIT: I see your axes are set to "con botones" - it should be 'separate'.

Have you tried to setup according to the guide?
https://en.lfsmanual.net/wiki/Options#Controls

Hi, I see there were some threads in the 2023 forum section so I hope it's OK to create a thread here. If not, someone can delete it.

My question is about directions, I have a vague memory that last year one of the rounds (that I didn't enter) was run in both directions around the track. Is that happening this year, or maybe not now that the two races are of unequal length?

Thanks!

Thanks for that detailed description of the guide lines based on trace. I wonder if it could actually be initialised from a trace. I'm trying to keep the interface simple to implement. So you would make a trace then click a button to keep it as reference line data, until cleared.
[EDIT: I have done this now, works well - see attached image]

About the centering, I've thought more a bit today. I think a good way would be if I enabled ALT+click on the number (to the right of r/f/u buttons). Then the dialog that is initialised would be similar to the existing dialog but with some different text. And the outcome of that would be that all the selected points acquired that value.

So...

1) You could use that to achieve what you are saying, like this:
ALT+click the number. Enter 0.

2) You could make all points match the green point, like this:
ALT+click the number. Press enter.

3) Set all to some other value:
ALT+click the number. Enter a new number.


[I'm also looking back at the other suggestions]

Last edited by Scawen, Thu, 28 Mar 2024 18:31.

It's really not on topic as it's a completely different interface.

By the way, my own Xbox controller reports separate axes through Direct Input.

The problem you report is really a Microsoft bug.

I can't imagine that it would be easy to implement the new interface, maybe even impossible with the old compiler used for the public version. Anyway, it's a really big task and I'm 99% certain this will never be implemented into the old version of LFS.

Nice to see a dragster approved. I have an improvement suggestion.
In version 23, I believe the engine rotation direction is wrong. For nearly all RWD cars, when you blip the throttle, the left side should lift (car rocks clockwise from driver's viewpoint as engine revs increase).
This is confirmed at least in this video.
https://www.youtube.com/watch?v=ywDgpkWas8k

To fix: 'Object Positions' -> Rotation (from front) -> clockwise.

I want to have a day on the editor tomorrow.

Quote from bayanofmansorofisky :

Static Line : Draw a line either from selecting multiple points (as trace function) or hold and drag mouse. Similar to select box. But only its job is to give a guide line thats helpful to see tris edge, heading or pitch when in 2D view.

I am a bit tired now, but I'll ask you to make this clearer for me to understand.

Is it really something that behaves like the 2D drag box, only it is separate from that and doesn't select points? Can you write a bit more description and a couple of use cases so I can visualise it better?

Quote from bayanofmansorofisky :

Snap to : edit the existing button Center, but have an option to snap selected points to one of the axis, 0 the value of X, Y or Z
currently only available for X

I think I thought of something like this before, but more related to typing in a number for X, Y or Z. But instead of preserving all the points value relative to the green selected point, it would set all of those points to that exact value that you typed in.

If that makes sense, would it achieve what you are suggesting? I think your description is what I've said but you type in '0'.

Thanks, I decided not to add that one as it is not available in the new physics version.