No vis information after any map change

[kingisalive]

Hi!

I'm trying to learn a bit by modding gm_conctruct. I'm using vmf downloaded from facepunch forums.

As long as I don't change anything, the map compiles and works in Garry's Mod. As soon I change anything: add an entity, change a property of an existing entity, add a brush etc. I get a bunch of compile errors, and global lighting doesn't work ingame. Is there anything I'm doing wrong within Hammer?

Here is log for unchanged map: http://pastebin.com/raw.php?i=ZvRg7wPn
Here is log for a map with added NPC: http://pastebin.com/raw.php?i=GxDerTv9

I'm using Hammer from Source SDK Base 2013 Multiplayer.
Here is my Hammer config: http://pastebin.com/raw.php?i=Fuy5VErJ
Here is GarrysMod\garrysmod\gameinfo.txt: http://pastebin.com/raw.php?i=mGW2ZdxF

[cz_squishy]

You have some leaks

Code: Select all

entity info_node (-5171.00 -3344.00 272.00) leaked

static prop models/props_rooftop/antenna02a.mdl outside the map (-2808.00, -2528.00, 3008.00)

[kingisalive]

And now I feel like a little fucktard. Thanks. I don't how I've missed it.

I've fixed one leak by making a skybox wall thicker, because after loading Pointfile the red line was going through a middle of a solid wall. Is that an OK fix?

[LordDz]

Well is the "solid" wall a func_detail or a brush entity? They don't block vis. You should probably look in further why the pointfile moved through the wall, so you're not compiling vis at unnecessary places.

[kingisalive]

It is a brush. Dunno, after the dimesion change Hammer gives no more leak errors. It was a single brush horizontal, simplest ever, roof placed on four vertical walls, with no empty place between them.

As I've said, changing its width fixed the reported leak.

[kingisalive]

Now I have a leak which appeared somewhat randomly, after I've edited things in another part of the map. After loading pointfile it looks like this:





It originates in an info_node. I'm clueless as to how to fix a leak, that leaks not into the void.

[Lhorkan]

It might be possible that the line of the point file is wrong. To test if you don't actually have a gap to the void somewhere, enclose your map with the cordon tool. If it does compile without a leak, you'll know for certain there's a gap somewhere.

Leaks can happen without there being any holes in a map, but only if there are brush entities whose origin has been placed outside the map bounds. Since an info_node is being leaked, this seems unlikely to be the case here. However, for future reference, you can easily check by selecting the entire map and going around the border of the map in all 2D viewports, checking if there are any white circles (the gizmo for the origin) outside the bounds.

Edit: Also, on a semi-unrelated note, I'm seeing that you enclosed the map in a single giant cube. This seems like the obvious way to build a map, but it's very bad practice in the dated Source engine. Read about it here. If you have a good reason on why the skybox has to be huge and open, at least fill the large connected bits of air with func_visclusters. It will reduce VVIS compile time drastically.

[LordDz]

Copy the coordinates from the leak, use view -> go to coordinates and paste them there, disable all func_details, props and brushes and see if you can detect a hole somewhere.

[kingisalive]

Copy the coordinates from the leak, use view -> go to coordinates and paste them there, disable all func_details, props and brushes and see if you can detect a hole somewhere.

My bad, I should've shown the leak's origin as well. The line starts exactly where the center of the info_node is. Here is another node giving similar leak information. The leak originates in the bottom center of the info_node and goes through the map's roof, which is a one piece world brush.

Some pictures of the reported entity leaking:




The place of the leak through the map's roof:

[jimonions]

Don't know if this is your problem but sometimes the pointfile is a lie, it doesn't point to exactly where the leak would be. In my case I disable all detail visgroups and manually search for it. I think this will also happen if one of your entities has illegal values.

[kingisalive]

It might be possible that the line of the point file is wrong. To test if you don't actually have a gap to the void somewhere, enclose your map with the cordon tool. If it does compile without a leak, you'll know for certain there's a gap somewhere.

After cordoning the map, and compiling with enabled cordon the compile went through. So yeah. What's even more amazing with the cordon disabled, the map started to compile properly as well. I seriously don't know what to say.

Edit: Also, on a semi-unrelated note, I'm seeing that you enclosed the map in a single giant cube. This seems like the obvious way to build a map, but it's very bad practice in the dated Source engine. Read about it here. If you have a good reason on why the skybox has to be huge and open, at least fill the large connected bits of air with func_visclusters. It will reduce VVIS compile time drastically.

We're planning on doing a new map from ground up. We'll keep this in mind and use cordon tool first, then place skybox later. Thanks for the tip, it sounds really helpful, will definitely try this out.

[Knights]

I'm facepalming to death how messy your map is looking. You really need to make sure you enclose your maps and stay on the grid at least for that part (the skybox and ground for example).

[kingisalive]

I'm facepalming to death how messy your map is looking. You really need to make sure you enclose your maps and stay on the grid at least for that part (the skybox and ground for example).

I have to work with what I was given. It's a map used for observing how people behave in virtual reality. I was handed that map, and asked to add some things. For the time being I have to add some features to the old one, while we create new map from zero.