Zircon Engine - a "DarkPlaces Mark V" project mod for Quake

March 8: Next Version Closer (Download: Zircon Beta #139)

All the hardest work is done. Need to test everything thoroughly and review all the changes. Some of the test cases require complex setups.

March 7: UI Work Continues (Download: Zircon Beta #139)

I have been polishing up the "devinfo" UI during the refinement process. And adding new information panels. I have a hard line drawn, but everything before that line needs to be exactly how I would want it.

March 5: (Download: Zircon Beta #139)

Update: An unsupported #148x is available. Probably the final UI adjustments in the deep UI code. I added context menu support and "tool tip" support (which is to say text will hover to say what it is, if indicated -- at least usually that is what it is used for). These just happened to be "in the neighborhood" -- I already worked on hard UI issues, and thought about handling these 2 additional UI issues before closing this type of coding out.

What seems likely now is I do refinement and testing and get out a Windows / Linux / Mac regular release version over the next few days.

March 4: Text Filtering ✅ (Download: Zircon Beta #139)

Update: An unsupported #147x is available. "devinfo" entities and such can be filtered to show only rows containing specific text (like "door" to show only "func_door" as one example).

March 4: Text Filtering (Download: Zircon Beta #139)

Working on "devinfo" having a textbox to filter results.

Should help narrow down looking for information. I have 2000 Q3 shaders and finding a specific one needs to be easy. QuakeC can have a lot of fields, and QuakeC can have a lot of global variables. Etc.

March 3: UI Wrap Up For Now (Download: Zircon Beta #139)

Managed to get all the "hard things" done. Will spend the next few days reviewing everything.

I expected a lot of unwanted surprises for the UI work I did today. And they happened -- but the end result is great. 😊😊

March 3: "Devinfo" Continued (Download: Zircon Beta #139)

FTE: RenegadeC has made a Tremor RPG page: Tremor RPG page.

Update: An unsupported #146x is available. It is a work-in-progress version that needs some polishing and has not gone through all the testing yet because some features require complex test cases that are hard to setup.

Recent:

• Solved shader uniqueness issue by having the stringlist do a multi-column sort (now an easy standardized engine function) and then walking the list checking for duplicates.
• Added TextBox double-click (select word at mouse cursor) and triple-click (select all text)
• Added List multi-row selection capability and ability to copy that data to clipboard with CTRL-C.

Going to close out this type of work for now after completing the following:

• Extend some controls be movable and closable with little handles on them or whatever ends up making sense.
• Make sure multi-select can be disabled.
• Make sure everything in "devinfo" works the way intended.
• Engine testing including triple-checking "Global transport" again (some things shifted).
• Test changelevel2 nosave.

March 2: Shaders .. (Download: Zircon Beta #139)

The shaderlist screenshot below takes 22 seconds to load in my debug environment (Visual Studio) and 6 seconds to load compiled with mingw64.

Rewrote everything to parse each scripts/something.shader file once and then use that data to fill in information. Now is nearly instant in the debug environment so it will be instant when compiled for release.

However, it finds 16 more shaders (these are duplicates). To do this proper, it must officially recognize the correct one that the engine uses and not the duplicate.

So have to rewrite that a bit more.

Update: There is no good way to de-duplicate the shader list. The speed and efficiency require writing out an unsorted, multicolumn list processing each scripts/myshader.shader file one at a time. That process cannot control the parse order. So I will have to de-duplicate it later and what I have to work with is a multicolumn column list (no direct sort method). Solution: Write a multicolumn sort (now done) and step through the list and eliminate duplicates.

March 2: WIP Shader Text Viewer (Download: Zircon Beta #139)

This is a work in progress shot, that little window that shows the shader text -- I want to give it a "grip" so it can be moved and the "X" button needs to be "internalized" as a native part of the control.

Pictured: "devinfo" Zircon Beta console command. It is getting another upgrade.

The dungeon project has a ton of pieces and sometimes it can become very difficult to locate things, I am taking a little time to make the engine expose more information quickly to reduce the "can't find [x]" problem. Shader text is one of those things that can be hard to find and if you look carefully, you will notice the at the bottom of the shader list a couple of entries near the bottom ==> it shows the .pk3 file they are in (if they are in a pk3 file).

March 1: Menu QC "Commands" (Download: Zircon Beta #139)

DarkPlaces has no way for Menu QC to do registercommand and, in fact, has no way to a command to access to Menu QC. CSQC can run commands through float CSQC_ConsoleCommand (string strMessage) provided the command is known (by doing registercommand("dothis") in CSQC_Init).

Now the reason it sucks that Menu QC cannot run commands is that Menu QC is *ALWAYS* running. It is the only permanently running progs.dat. In fact is also always drawing (except when the loading screen is up, but Zircon has m_drawloading) -- although an m_draw function usually has a first statement like "if (!menu_state) return;" but m_draw is always available to 2D draw.

So you might have a command that should always be available in the console and Menu QC is always running so it would be great to have Menu QC be able to run commands. But it can't.

There is a workaround ... "menu_cmd".

menu_cmd is the single exclusive command that Menu QC can run. Due to this, it should be possible to run commands in Menu QC through a "backdoor" -- like: alias my_command "menu_cmd my_command". menu_cmd routes the command and the arguments to GameCommand

• alias my_command "menu_cmd my_command" ...

// menu.qc
void GameCommand (string s) // menu_cmd
{
	print (strcat("Menu GameCommand: ", s, "\n"));

	float argc = tokenize_console(s);
	string cmd = argv(0);

	if (cmd == "my_command") {
		// Do something
	}
}

February 27: Rework (Download: Zircon Beta #139)

Briefly tried to allow entities to exclude themselves from dirt mapping with an entity key (this was a q3map2 modification). Coded about 80% of it before realizing that surfaces in the light compile stage do not know what entity they came from. There is already a q3map_noDirty shader keyword, but I wanted entity control because it is easier to work with. In the past, I made surfaceparm nomarks into an entity keyword ("_nodecals 1") so I thought this would be similarly easy. (I have some arches that -dirty argument in the -light phase causes a very bad looking line, so I want to exclude just those).

Also took a step back and realized that the lighting model I am using is too dark and the sizing model I am using is too narrow. I want multiple players to be able to traverse halls easily. The too dark thing matters because I want to have wide latitude to lighting approaches available.

Multiplayer: I would prefer to use the q3 standard lightmap that embeds in a bsp (size 128 lightmap) because the data is small (therefore the map is smaller) and the map compiler times are much faster. If I use larger lightmap sizes, the shadows are vast improved but there are external lightmaps that are large in size and fairly numerous in quantity (a map I compiled had 46 lightmaps of 0.7 MB. I'll have to decide which one to go with -- with DarkPlaces sticking with the size 128 lightmap might be ok because if you turn on real-time lighting, it will look better than the compiled lightmaps.

Extra note: DarkPlaces real-time lighting only works with "light" entities. q3map_surfacelight is great, but does not work with DarkPlaces real-time lighting -- DarkPlaces reads the light entities from the .bsp entities segment (_keeplights 1 must be in worldspawn during map compile, Q3 tools want to remove the lights which is opposite behavior of Quake 1).

Random: in the q3map2 BSP phase, as best I can tell -- always use -maxarea. This cuts up triangles in a GPU friendly way, and I noticed the triangle count is substantially reduced (15% to 20%). I tested it expecting a modest reduction in triangles, but 15% to 20% is very significant.

Random 2: q3map2 likes to warn about entities in solids -- for decal projection this is essentially the norm. The end point often needs to be in the ground or inside a wall. TODO: I want to check to see if decal projections end points could just be removed in q3map2.

Engine: made a way to print model information in the console. Should put in it devinfo. Should put shader script text in devinfo. Should make a textbox to filter in devinfo and pressing CTRL-C should copy information.

February 26: Engine - Not Always Enough (Download: Zircon Beta #139)

Engine coding is not always enough.

I am able to procedurally generate maps now, q3map2 has "func_group" (these are like func_wall except they are part of the world). When the Quake engine is used, those were converted to part of the world and they are no longer entities. But they are entities to q3map2.

I create a lot func_group entities (A LOT OF FUNC_GROUP). I make complex map parts into a func_group to keep them glued together as one piece if I need to move them.

I was working on a relatively simple modification to "tessellate" the floor (break it into blocks of 384x384) so decals project better. Then I can also remove unneeded floor.

Now nothing works right. And it seems there are no error messages.

I broke a limit in q3map2. Opened my custom q3map2 source code, increased the entities limit from 0x800 -- I don't know what 0x800 is off hand --- to 0x8000 // Baker: raised to 32768 old: 0x800 // 2048 entities.

While I was in there, q3map2 sometimes prints "entity 382 in world" type messages. No classname prints. No origin prints. So I made that more helpful by printing both the classname and the origin. I will likely occasionally make the warnings or notifications I actually see when compiling a map more helpful.

By adding the classname and the origin, it will reduce the human time to identify the offending entity greatly. Also the fairly useless "leaked" warning .. I expect to make that a bit improved and say the origin and maybe classname.

Fun History: LadyHavoc and Vortex both did a significant amount of work in the q3map2 source code. If you work with the q3map2 source code, you come across their comments in the code.

February 25: Mount Grindmore (Download: Zircon Beta #139)

I finished off a major checkpoint today. There are more to go.

Today is a day for beer. Tomorrow is it back to the grind.

February 24: Experimental #144x (Download: Zircon Beta #139)

Update: An unsupported #144x is available. I was able to more deeply review and test some new functionality and double checked most things, but haven't entirely put it through all the right testing.

geri43 at Github submitted a DarkPlaces patch for better support for "v_flipped 1" and that is in 144x.

February 23: No Movement = Timestop (Download: Zircon Beta #139)

There is a modified DarkPlaces 2014 called SuperQot available for download that makes it so time stops when you aren't moving.

February 23: Flood Fill (Download: Zircon Beta #139)

Flood fill is annoying. Here is my version of flood fill, which is non-recursive. Non-recursive is important otherwise you could stack overflow if the size gets large.

Something like this might become a QuakeC function in the future. I tried to get Google AI to write the code, it came up with something gross and you can never really trust AI code -- although most time it is helpful. I'm sure somehow out there, there is an equally streamlined and simple flood fill function -- but I couldn't find one with Google.

void Baker_Flood_Fill (int ch_fill, int ch_target, int x, int y, char *s_modify, int width, int height)
{
	int		numcells = width*height;
	int		cell_size = sizeof(point2d_s);
	point2d_s	*my_queue = (point2d_s *)calloc(numcells, cell_size);
	int		queue_count = 0;
	int		iters = 0;
	int		ox = 0, oy = 0;
	point2d_s	q = {0,0}; // point2d_s has int col, int row;

	my_queue[queue_count].col = x, my_queue[queue_count].row = y;
	queue_count ++;

	while (queue_count > 0) {
		q.col = my_queue[queue_count - 1].col, q.row = my_queue[queue_count - 1].row; 
		queue_count --;

		s_modify[q.row * width + q.col] = ch_fill;
				
		ox = q.col + 0, oy = q.row - 1; // LOOK UP
		if (in_range_beyond (0, ox, width) && in_range_beyond (0, oy, height) && s_modify[oy * width + ox] == ch_target) {
			my_queue[queue_count].col = ox;
			my_queue[queue_count].row = oy; 
			queue_count ++;
		}

		ox = q.col + 0, oy = q.row + 1; // LOOK DOWN
		if (in_range_beyond (0, ox, width) && in_range_beyond (0, oy, height) && s_modify[oy * width + ox] == ch_target) {
			my_queue[queue_count].col = ox;
			my_queue[queue_count].row = oy; 
			queue_count ++;
		}

		ox = q.col - 1, oy = q.row + 0; // LOOK LEFT
		if (in_range_beyond (0, ox, width) && in_range_beyond (0, oy, height) && s_modify[oy * width + ox] == ch_target) {
			my_queue[queue_count].col = ox;
			my_queue[queue_count].row = oy; 
			queue_count ++;
		}
		ox = q.col + 1, oy = q.row + 0; // LOOK RIGHT
		if (in_range_beyond (0, ox, width) && in_range_beyond (0, oy, height) && s_modify[oy * width + ox] == ch_target) {
			my_queue[queue_count].col = ox;
			my_queue[queue_count].row = oy; 
			queue_count ++;
		}
	} // while

	freenull_ (my_queue);
} // FloodFill

February 22: Created Map In QuakeC (Download: Zircon Beta #139)

Update: An unsupported #143x is available with the new QuakeC functions.

I'd post a pic, except because I had to change J.A.C.K. to use a different graphics card after a BIOS update, ShareX (the free video capture tool I use) cannot seem to capture the J.A.C.K. window any more --- haha.

This creates a .map with a single block of brick. The new QuakeC function map_entity_brush_add_cubeoid worked the first time. This small function creates a new .map, adds a world entity (#0) and then adds a brush with ambientcg/Bricks018 as the texture.

// Standard QuakeC ..
#define int float // standard QuakeC does not have int, only float, but makes code more legible
void Map_Make ()
{
	int hmap = map_create ();

	int entnum = map_entity_add(hmap, "classname", "worldspawn");
	printvarf (entnum);

	int brushnum;
	
	// Sides = 63
	int sides = TEXSIDE_NORTH_1 | TEXSIDE_SOUTH_2 | TEXSIDE_EAST_4 | TEXSIDE_WEST_8 | TEXSIDE_CEIL_16 | TEXSIDE_FLOOR_32;

	vector textures_scale = { 1/8.0, 1/8.0};
	// Returns brush # for entity (0 or greater) or -1 on failure
	// Texture scale if not specified is 0.125 which is 1/8 scale (high resolution texture size friendly like 1024 x 1024).
	brushnum = map_entity_brush_add_cubeoid (hmap,
		WORLD_ENTNUM_0, 
		[-256,-256,-256], 
		[ 256, 256, 256],
		"ambientcg/Bricks018",  // textures/ambientcg/Bricks018.jpg
		sides, 
		textures_scale // 1/8.0 == 0.125 which is one eighth scale suitable for a 1024 x 1024 high resolution texture.
	);

	map_save_as (hmap, "maps/sample_block.map"); // Disallows path not named "maps/" and any extension not ".map"
	zpr ("Sample block written");
}

February 22: Wrecked By Bios Update (Download: Zircon Beta #139)

Yesterday, my HP Victus laptop decided to do a bios update and shut itself down without asking me and install the BIOS update. After that, the J.A.C.K. map editor 3D window no longer rendered, making J.A.C.K. useless and reinstall did not fix.

I ended up solving the issue by going (Windows 11) Settings > Display > Graphics and adding J.A.C.K. and specifying it force use of my ATI Radeon instead of the NVIDIA that machine has (10 years ago a computer having both ATI and NVIDIA would be crazy, but this laptop model has that).

The above sounds "simple", but for about 30 minutes it looked like I was totally up a creek.

Then I had the thought that maybe I could force a different video driver in the Control Panel and tried that as a random act of desperation, but it worked. As I understand it, J.A.C.K. uses Qt4 to render (TrenchBroom uses Qt6).

Short version: There was nothing "simple" about it. Only luck and intuition saved the day. By sharing this information, maybe someone else may find this helpful in the future.

February 22: Map QuakeC API add brush (Download: Zircon Beta #139)

I have been using a nearly empty .map created in J.A.C.K. and then filling with map pieces from other .map files and pasting them into the .map. But that method requires knowing the size of the map in advance and if I come up with an idea that requires a larger "shell" I have to make a new .map in J.A.C.K.

I don't want to do that. So adding a function to add a cube with specified mins and maxs and then with flags indicating what side or sides to texture (the other sides get common/caulk).

float map_entity_brush_add_cubeoid (float hmap, float entnum, vector brushmins, vector brushmaxs, string texture, float texture_flags, optional vector texture_size); // Return value is brush number

Texture_flags will be something like NORTH_1, SOUTH_2, EAST_4, WEST_8, CEIL_16, FLOOR_32 to indicate which side gets textured or send 63 for all of them to get textured.

(*) common/caulk is the "pink" Q3 map texture that indicates that the surface will be stripped from the BSP for rendering (it is solid for collision) and likewise that surface will not get a lightmapped. Good use of caulk reduces the size of a BSP and reduces the number of triangles rendered.

February 21: buf_create (Download: Zircon Beta #139)

Unsupported Zircon Beta Release #142x is available for Windows -- hammered out maze_generate QuakeC function and made other adjustments including "buf_create Quake function shall never return 0 (to avoid confusion with uninitialized global variables which also have a value of zero)".
--------------------
The next version of Zircon Beta will never return "0" for buf_create. buf_create creates a stringlist array handle and DarkPlaces has it return 0 the first time it is called.

So you have a variable float my_stringlist_number = buf_create();

And it gets a value of 0.

Q: How does one distinguish an uninitialized variable with a value of zero from a handle to a valid stringlist array that has the value of 0?

A: You can't.

The next Zircon Beta will never return "0" for buf_create and avoiding this problem.

February 20: Zircon Beta Release 141x (Download: Zircon Beta #139)

Unsupported Zircon Beta Release #141x is available for Windows in the downloads section with changes indicated in the description. It is probably ok, but has not gone through appropriate testing -- it has maze generation QuakeC functionality and it is complicated to setup test cases.

February 19: Mazegen API (Download: Zircon Beta #139)

Implementing maze generation API into QuakeC. Probably should implement maze solve API into QuakeC while I am doing this.

Update: Nearly done with that. About as fun as walking across broken glass, but that is going to be "the norm" with almost every aspect of hammering out the "dungeon project". I expect dozens more of these types of challenges.

February 2026 - The Dungeon Project (Download: Zircon Beta #139)

One year ago, attempted to complete "The Dungeon Project" and it led to much engine coding and no small amount of QuakeC work (the game logic in progs.dat), but was not able to get any true traction -- all efforts led to engine coding side quests instead.

Some 2025 Side Quests Spawned From The Dungeon Project:

• Zircon Beta particle system overhaul (Timewarp example).
• QuakeC particle system experiments (Dust1 and Dust2 and Dust 0)
• QuakeC Scripted particles (pic)
• Water experiments (pic)
• Camera experiments (pic)
• Several new Zircon .map pre-processor capabilities

Steam Integration In Zircon Beta (OBJECT N)