[Game] MineClone2 [0.83.1]

[Wuzzy]

There were always two ways of writing the name — with and without the space.

This is false. I always wrote it with space. If I didn't, that was a typo. But I have left the project many years ago. This post was just a quick note of mine, nothing else. What you do with the project is up to the project maintainers. I just wanted to make sure if this was a mistake or not.

[Mr. Rar]

Was the name of the game changed to "MineClone2" (without space before the "2")?

Because the ContentDB + repo description + readme uses "MineClone2" instead of the original name "MineClone 2".

I just ask if this was done in error. The original name of the game is "MineClone 2" (with space). Just a friendly reminder.

Whatever game name you pick, I strongly suggest to be consistent everywhere.

I made an issue about this some time ago: https://git.minetest.land/MineClone2/Mi ... ssues/2966

I'm more sympathetic to removing the space because it is weird to write MineClone 2 0.83.0. The 2 is actually part of the name and not the version.

[ancientmariner]

Hotfix Release

Hotfix release (0.83.1) is on ContentDB to fix a few asset issues.

[Codiac]

Hi, while poking around the code I found this comment in function settlements.place_schematics --Note, block substitution isn't matching node names exactly; so nodes that are to be substituted that have the same prefixes cause bugs.

You can overcome this by including the double quotes in the match, additionally you can match block variations (top/bottom/etc) using regular expressions.

Here is an example.

+ Spoiler

Code: Select all

elseif platform_material == "mcl_core:sand" or platform_material == "mcl_core:redsand" then
	schem_lua = schem_lua:gsub("mcl_core:tree", "mcl_core:sandstonecarved")
	schem_lua = schem_lua:gsub('"mcl_fences:fence"', '"mcl_fences:birch_fence"')

	schem_lua = schem_lua:gsub("mcl_core:wood", "mcl_core:sandstonesmooth")
	schem_lua = schem_lua:gsub('"mcl_stairs:slab_wood([^"]*)"', '"mcl_stairs:slab_sandstonesmooth2%1"')
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_wood([^"]*)"', '"mcl_stairs:stair_sandstonesmooth2%1"')

	-- stonebrick has mossy and cracked, what should they match to?
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_stonebrick"', '"mcl_stairs:stair_redsandstone"')
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_stonebrick_([^"]+)"', '"mcl_stairs:stair_redsandstone_%1"')
	schem_lua = schem_lua:gsub('"mcl_core:stonebrick"', '"mcl_core:redsandstone"')
	schem_lua = schem_lua:gsub('"mcl_core:stonebrick_([^"]+)"', '"mcl_core:redsandstone_%1"')

	schem_lua = schem_lua:gsub("mcl_core:cobble", "mcl_core:sandstone")
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_cobble([^"]*)"', '"mcl_stairs:stair_sandstone%1"')
	schem_lua = schem_lua:gsub('"mcl_stairs:slab_cobble([^"]*)"', '"mcl_stairs:slab_sandstone%1"')
	schem_lua = schem_lua:gsub('"mcl_walls:cobble([^"]*)"', '"mcl_walls:sandstone%1"')

	schem_lua = schem_lua:gsub('"mcl_stairs:slab_brick_block([^"]*)"', '"mcl_core:redsandstonesmooth2%1"')
	schem_lua = schem_lua:gsub('"mcl_core:brick_block"', '"mcl_core:redsandstonesmooth2"')

	schem_lua = schem_lua:gsub('"mesecons_pressureplates:pressure_plate_wood_([^"]+)"', '"mesecons_pressureplates:pressure_plate_birchwood_%1"')
end

I've tested this on seed 15912704083336939069 which has a dessert village pretty close to the spawn point. It looks OK, but probably need a better selection of replacement materials. Aesthetics aren't my strong point ;)

[Codiac]

Hi again, in Minecraft houses with beds do not contain job sites and vice-versa, I think this is to force villagers out of their houses more often. Is this kind of change desired? Obviously it requires more schematics. I have a template I'm using for job sites, so a bunch of them have a similar shape but custom bits to fit the jobs. Plain but functional.

In my testing of this change I noticed it's quite easy to get a village with a big variation between beds and jobs, so you can end up with an empty village or way to many villagers for jobs. Probably the site planning would need to be adjusted once this was in place.

On a similar vein, is there a reason to limit beds to red? I have a tweak that allows any color bed, it seems to work OK and allows a bit more variation between buildings.

Also, finally, my git account has been verified, so I can open issues and MRs now! :D

[ancientmariner]

Hi again, in Minecraft houses with beds do not contain job sites and vice-versa, I think this is to force villagers out of their houses more often. Is this kind of change desired? Obviously it requires more schematics. I have a template I'm using for job sites, so a bunch of them have a similar shape but custom bits to fit the jobs. Plain but functional.

In my testing of this change I noticed it's quite easy to get a village with a big variation between beds and jobs, so you can end up with an empty village or way to many villagers for jobs. Probably the site planning would need to be adjusted once this was in place.

On a similar vein, is there a reason to limit beds to red? I have a tweak that allows any color bed, it seems to work OK and allows a bit more variation between buildings.

Also, finally, my git account has been verified, so I can open issues and MRs now! :D

I have no idea. It was done a while a go.

I have a number of issues with the schematics (uses walls and window sills which break the villager pathfinding is a biggie imho. A zombie siege wouldn't be safe in that situation and villagers often fail to get home at night.).

Building improvements with job sites could be cool. We would weigh them up against what we have and seriously consider options.

Red beds used to be limited because I think the old villager logic only gave beds to villagers if they are red. I have since changed the logic so villagers will pick up any bed but the village generation doesn't reflect that.

Village generation does have a big issue right now. It can freeze for about 30s. It performs very poorly, especially when creating platforms and finding the ground. It needs to be rewritten IMHO, but it's not at the top of the priority list right now. People are free to work on it if they are passionate about it though.

Glad your git account got verified :).

[ancientmariner]

Hi, while poking around the code I found this comment in function settlements.place_schematics --Note, block substitution isn't matching node names exactly; so nodes that are to be substituted that have the same prefixes cause bugs.

You can overcome this by including the double quotes in the match, additionally you can match block variations (top/bottom/etc) using regular expressions.

Here is an example.

+ Spoiler

Code: Select all

elseif platform_material == "mcl_core:sand" or platform_material == "mcl_core:redsand" then
	schem_lua = schem_lua:gsub("mcl_core:tree", "mcl_core:sandstonecarved")
	schem_lua = schem_lua:gsub('"mcl_fences:fence"', '"mcl_fences:birch_fence"')

	schem_lua = schem_lua:gsub("mcl_core:wood", "mcl_core:sandstonesmooth")
	schem_lua = schem_lua:gsub('"mcl_stairs:slab_wood([^"]*)"', '"mcl_stairs:slab_sandstonesmooth2%1"')
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_wood([^"]*)"', '"mcl_stairs:stair_sandstonesmooth2%1"')

	-- stonebrick has mossy and cracked, what should they match to?
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_stonebrick"', '"mcl_stairs:stair_redsandstone"')
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_stonebrick_([^"]+)"', '"mcl_stairs:stair_redsandstone_%1"')
	schem_lua = schem_lua:gsub('"mcl_core:stonebrick"', '"mcl_core:redsandstone"')
	schem_lua = schem_lua:gsub('"mcl_core:stonebrick_([^"]+)"', '"mcl_core:redsandstone_%1"')

	schem_lua = schem_lua:gsub("mcl_core:cobble", "mcl_core:sandstone")
	schem_lua = schem_lua:gsub('"mcl_stairs:stair_cobble([^"]*)"', '"mcl_stairs:stair_sandstone%1"')
	schem_lua = schem_lua:gsub('"mcl_stairs:slab_cobble([^"]*)"', '"mcl_stairs:slab_sandstone%1"')
	schem_lua = schem_lua:gsub('"mcl_walls:cobble([^"]*)"', '"mcl_walls:sandstone%1"')

	schem_lua = schem_lua:gsub('"mcl_stairs:slab_brick_block([^"]*)"', '"mcl_core:redsandstonesmooth2%1"')
	schem_lua = schem_lua:gsub('"mcl_core:brick_block"', '"mcl_core:redsandstonesmooth2"')

	schem_lua = schem_lua:gsub('"mesecons_pressureplates:pressure_plate_wood_([^"]+)"', '"mesecons_pressureplates:pressure_plate_birchwood_%1"')
end

I've tested this on seed 15912704083336939069 which has a dessert village pretty close to the spawn point. It looks OK, but probably need a better selection of replacement materials. Aesthetics aren't my strong point ;)

Aesthetics aren't my strong point neither. I think it would be good to optimise or rework the village generation logic and make it performant before tacking the look and feel side. My previous comment I just posted touches upon this.

[Codiac]

Hi, FYI I deleted my git account. I specifically told it to hide my email address and it used it to get some data off the web somewhere and put my real name in the commits, something I deliberately did not enter. I tried to disable that but it proved more difficult that it was worth so I just deleted the whole thing.

[IGPslot]

Hello, I've got a problem that feels to me a bit nonstandard, so I'm not sure whether to open an issue on Github.

I've had a power failure while playing, and after restarting PC my world looks completely black. SFX works the same, so is GUI, but I cannot see blocks, entities nor even the wireframe.

I created new world just to check, and the new one functions and looks completely normal, so the problem pertains only to the world I was playing during the blackout.

[Blockhead]

Hello, I've got a problem that feels to me a bit nonstandard, so I'm not sure whether to open an issue on Github.

I've had a power failure while playing, and after restarting PC my world looks completely black. SFX works the same, so is GUI, but I cannot see blocks, entities nor even the wireframe.

I created new world just to check, and the new one functions and looks completely normal, so the problem pertains only to the world I was playing during the blackout.

This may or may not relate to MineClone2 specifically. I think the best place for advice is probably a new thread in the "Problems" subforum. There you should give more details like the output of debug.txt (which is also the contents of the terminal when Minetest runs), what game, mods, operating system and so on.

[IGPslot]

Hello, I've got a problem that feels to me a bit nonstandard, so I'm not sure whether to open an issue on Github.

I've had a power failure while playing, and after restarting PC my world looks completely black. SFX works the same, so is GUI, but I cannot see blocks, entities nor even the wireframe.

I created new world just to check, and the new one functions and looks completely normal, so the problem pertains only to the world I was playing during the blackout.

This may or may not relate to MineClone2 specifically. I think the best place for advice is probably a new thread in the "Problems" subforum. There you should give more details like the output of debug.txt (which is also the contents of the terminal when Minetest runs), what game, mods, operating system and so on.

Thank you so much for the info, will do.

[ancientmariner]

Hi, FYI I deleted my git account. I specifically told it to hide my email address and it used it to get some data off the web somewhere and put my real name in the commits, something I deliberately did not enter. I tried to disable that but it proved more difficult that it was worth so I just deleted the whole thing.

Your commit info is stored locally, you usually set it with git config. You have a global configuration which is used for everything on your git, and one for the repo you're on, so you can set that to a different username and password and it'll have that info in your commits.

Our git instance is a Gitea instance, a self-hosted open source instance. I very much doubt there are any uglies in there so I'd double check your git config on your local machine.

Once you've commited locally, there should be a way to see the metadata set on it (on intellij I can check email and username for the commit, so there should be a command for it).

[mine_eru]

Hi, FYI I deleted my git account. I specifically told it to hide my email address and it used it to get some data off the web somewhere and put my real name in the commits, something I deliberately did not enter. I tried to disable that but it proved more difficult that it was worth so I just deleted the whole thing.

Your commit info is stored locally, you usually set it with git config. You have a global configuration which is used for everything on your git, and one for the repo you're on, so you can set that to a different username and password and it'll have that info in your commits.

Our git instance is a Gitea instance, a self-hosted open source instance. I very much doubt there are any uglies in there so I'd double check your git config on your local machine.

Once you've commited locally, there should be a way to see the metadata set on it (on intellij I can check email and username for the commit, so there should be a command for it).

Personally I have some issues with Intellij and Git, many times I can do something at private repo with user ptoection, but Intellij sent my change as unknown user. So I stop using Intellij and Git integration and use integration in Eclipse or VSCode. Of course you can use terminal :)

[ancientmariner]

Hi, FYI I deleted my git account. I specifically told it to hide my email address and it used it to get some data off the web somewhere and put my real name in the commits, something I deliberately did not enter. I tried to disable that but it proved more difficult that it was worth so I just deleted the whole thing.

Your commit info is stored locally, you usually set it with git config. You have a global configuration which is used for everything on your git, and one for the repo you're on, so you can set that to a different username and password and it'll have that info in your commits.

Our git instance is a Gitea instance, a self-hosted open source instance. I very much doubt there are any uglies in there so I'd double check your git config on your local machine.

Once you've commited locally, there should be a way to see the metadata set on it (on intellij I can check email and username for the commit, so there should be a command for it).

Personally I have some issues with Intellij and Git, many times I can do something at private repo with user ptoection, but Intellij sent my change as unknown user. So I stop using Intellij and Git integration and use integration in Eclipse or VSCode. Of course you can use terminal :)

I only use command line for git for usual operations. GUI seems to abstract a lot away and people often end up making errors that they wouldn't if they learnt through the command prompt.

IntelliJ integration with git is great to see history and changes, and more information on commits though.

The point was not to recommend it, the point was that if git integration through intellij has that information, then it is accessible. If it's sending stuff as an unknown user, the git config is incorrect. This would happen via command line or any integration...

[erlehmann]

I got back into MT after hearing there's finally support for shadows and other pretty effects […]

Offtopic, but you could have have had dynamic shadows in Minetest for years by simply using the Irrlicht library feature:

1. First, enable the stencil buffer – this is a flag on createDevice().
2. Call the addShadowVolumeSceneNode() function for your mesh node.
3. Add some light node to the scene. The only light node I could find in Minetest 5.4.1 is attached to the camera for reasons unknown to me (could this be the reason for some lighting bugs related to the removed bump mapping feature?).

The shadows generated by Irrlicht are an extrusion of the mesh outline. They can only be a single color, but are reasonably fast and have crisp edges and handle self-shadowing for free (basically you get shadows like in DOOM 3). I suspect it does not become as computationally expensive as shadow maps do when you add more lights to the scene, but I have not tested that.

This means shadows of leaves and colored glass will probably not work. You will also notice that it seems to look wrong for many models. The reason for that is probably that the outline optimization in Irrlicht fails to handle models that are not entirely closed (i.e. if you would fill them with fluid, it would leak out somewhere). You can just disable that optimization, then it looks correct.

I have suggested this a long time ago, but so far it was rejected every time to even try – it seems no one even wanted to look at that part of the Irrlicht documentation. I could not figure it out myself, but Fleckenstein was more persistent and implemented it some time back – turns out I did forget to enable the stencil buffer.

I also made a comic after sfan5 wrote on IRC that rendering shadows like this will never be done due to being “legacy OpenGL” (lol).

[MirceaKitsune]

Nice. If I'm being honest, I've been wondering if given the simple graphics Minetest still uses we need Irrlicht and can't just use OpenGL (also Vulkan as an option) directly as a backend: Irrlicht has a ton of features out of which I believe we only use a few, if we used bits like its hardware lighting it would make sense but with so much simplicity I don't know. Of course there's probably a good reason for it and too big of a change to make now.

[frodeseverin]

You can keep items on death with the following:

Code: Select all

# Normally, players drop all their items when they die. Enable this
# setting, so players always keep their inventory on death.
mcl_keepInventory (Keep inventory on death) bool false

Thanks for the help. I edited the line and it works perfectly!

Being a NOOB in Lua and Minetest settings at large I had some troubble setting the functionality for players to keep their inventory on death on my server running in Docker with Docker image from https://docs.linuxserver.io/images/docker-minetest. This is a necessity for my children gaining interest in playing the game at all.

I ended up doing the following

1. Editing minetest.conf in the Mineclone2 game root directory to read:

   Code: Select all

   # MCL2-specific stuff
   # Inventory
   # Normally, players drop all their items when they die. Enable this
   # setting, so players always keep their inventory on death.
   mcl_keepInventory (Keep inventory on death) bool false

2. Editing main-config/minetest.conf to include the following section:

   Code: Select all

   #
   # Keep inventory
   #
   # General Minetest setting
   keepInventory = true
   #
   # MCL2 Specific setting
   # Normally, players drop all their items when they die. Enable this
   # setting, so players always keep their inventory on death.
   mcl_keepInventory = true

The code mcl_keepInventory (Keep inventory on death) bool false seems to me to be a variable definition, and is missing from minetest.conf in the Mineclone2 game root directory. I suppose this is an error, and worthy of an issue in the issue tracker. Also there should be an explanation on how to use this in the main-config/minetest.conf file, and wether or not this functionality is dependent on setting both keepInventory = true and mcl_keepInventory = true.

[erlehmann]

Nice. If I'm being honest, I've been wondering if given the simple graphics Minetest still uses we need Irrlicht and can't just use OpenGL (also Vulkan as an option) directly as a backend: Irrlicht has a ton of features out of which I believe we only use a few, if we used bits like its hardware lighting it would make sense but with so much simplicity I don't know. Of course there's probably a good reason for it and too big of a change to make now.

You can do that – and I believe that the goal of Irrlichtmt (the “lets delete everything out of Irrlicht we do not need and also everything we do not understand and also some things we might need and then also delete the failing test cases” fork of irrlicht) is trying to do that.

I think it is a bad idea though, since:

• It tightly couples the renderer to hardware requirements (sfan5 keeps telling me that Minetest will stop working on my Hardware REAL SOON NOW due to some renderer refactoring requiring a supercomputer GPU, but so far it still seems to work on my Thinkpad T60).
• Irrlicht has solved a bunch of things in clever ways that Minetest did not manage to do in many years.

Here are several examples for why “doing raw OpenGL instead of using a scene graph that is independent of the rendering target” may lead to worse results than just using a library:

• AFAIK OpenGL does not mandate support for more than 8 lights per scene. I am sure that a lot of hardware can do more than that, but for hardware that can not do this, Irrlicht automatically prioritizes lights based on the distance from the camera. Irrlicht also has also an optional light manager that can turn lights off and on during rendering so you can have a limit of 8 lights per node or per zone (for local lighting that is not supposed to spill over to other rooms or so)!
• The built-in Irrlicht particles are much more performant than the Minetest billboard particles, likely due to batching drawcalls (it is possible lag out players with weak GPUs using particle effects pretty easily). AFAIK it was never used in Minetest, because it can not do the rain-style particles. Why not render rain differently than other things or contribute to Irrlicht so it can be done? I do not know – I am not aware of any patch going to Irrlicht (which is still developed) before I called sfan5 out on not contributing to upstream Irrlicht.
• Irrlicht can measure which scene nodes are entirely occluded (i.e. rendering the stuff does not change pixels on the screen). I doubt that Minetest uses that feature, since IIRC it was possible to occlude complex geometry with a single node that occupies the full screen and still get a huge FPS drop. Edit: In fact, I wonder right now if you can get huge rendering speedups using Irrlichts addOcclusionQuery() functionality.

There are a lot of cool features in Irrlicht, most of which were removed in the Irrlichtmt fork (either because they were not used or because it was not known if they were used or could ever be used) – which obviously makes it easier to argue that the rest (i.e. what is done) can be pruned as well, since it means there is infrastructure that can never be used. It also makes stuff like adding stencil shadows much more than a few lines of code, since one first would have to put that code back into Irrlichtmt again (as I understand this, it means if sfan5 disagrees with that feature, bad luck!).

[erlehmann]

if we used bits like its hardware lighting it would make sense but with so much simplicity I don't know

AFAIK Minetest has its own lighting system. You can make the world use smooth shading (Goraud?) if you place a light source, but a) you need to disable the triangle merge algorithm to avoid a rendering error b) it looks too weird for my blocky taste. IIRC It is merely a few lines of code, Fleckenstein played around with it when we were exploring the shadow volumes.

[MirceaKitsune]

Thanks for clarifying: In this case yeah Irrlicht is the best solution, especially if IrrlichtMT is going to remove everything that's unneeded for optimization. I take it we still back-port updates made to vanilla Irrlicht? Part of me still wishes we had hardware lighting but that's always been hard to get right with our voxel lighting and technically unneeded.

Just hope we can get shadows working with point lights such as torches and it's all awesome, currently only the sun casts shadows but I heard there were plans to lift this limitation. Specular and normal mapping is probably something we'll never use sadly, would love them but some may consider it too much trouble for simple blocky graphics... if yes I presume they could be supported with our lighting granted we transmit light direction from voxel data. The feature I really wanted it support for colored lights, like green / blue torches for magical fires and so on... maybe it will happen someday and light nodes can include their own color.

[erlehmann]

Thanks for clarifying: In this case yeah Irrlicht is the best solution, especially if IrrlichtMT is going to remove everything that's unneeded for optimization.

I think you misunderstand. IrrlichtMT removes stuff based on if it was used, not based on if it could be useful. Right now you will not find the code to render the crisp shadows in it, since it was not used. It was, along with other stuff, deleted in a huge PR that was approved without even reviewing it properly. It basically means that all the cool Irrlicht features not currently used are probably not accessible unless you do the work to bring them into Irrlichtmt.

Basically, at the fork in the road where Minetest could either aim to use Irrlicht correctly or write a new renderer, the latter path was chosen, with the ultimate goal of folding a gutted Irrlicht library into Minetest. I think that this will show results eventually, but at the expense of tight coupling, higher hardware requirements than strictly necessary and keeping bad home-grown solutions like the particle system etc. pp. … you better ask a core developer about how and why exactly this choice came to be, I suggest v-rob. (If anyone claims Irrlicht is not actively developed, disregard their opinion – Irrlicht does releases far too rarely, but it is not a dead project.)

I take it we still back-port updates made to vanilla Irrlicht? Part of me still wishes we had hardware lighting but that's always been hard to get right with our voxel lighting and technically unneeded.

Just hope we can get shadows working with point lights such as torches and it's all awesome, currently only the sun casts shadows but I heard there were plans to lift this limitation.

I think you can do that using the shadow volume functionality. Once you have shadows working, just add more lights using addLightSceneNode(). IIRC Fleckenstein did that when we tested it.

Specular and normal mapping is probably something we'll never use sadly, would love them but some may consider it too much trouble for simple blocky graphics... if yes I presume they could be supported with our lighting granted we transmit light direction from voxel data. The feature I really wanted it support for colored lights, like green / blue torches for magical fires and so on... maybe it will happen someday and light nodes can include their own color.

I suggest to start here for your journey: https://irrlicht.sourceforge.io/docu/example011.html