Mask
#12613 posted by SleepwalkR on 2013/03/25 14:29:55
TrenchBroom generally tries to hide the fact that vertices often don't have integer coordinates, but that approach is flawed and I'm working on removing it.
Elusive Brush Corruption
#12614 posted by Spiney on 2013/03/25 16:00:02
I noticed stuff like tetrahedrons or brushes with a lot of vertices lend themselves well to map corruption in TB. For rocks or terrain just mashing deformed cubes together is more work but seems more stable at the moment.
Spiney
my terrain is completely cube-based. I didn't consider making it any other way.
A lot of the problems have come from my work method of basically making the level in an open room-by-room fashion. I could have avoided a lot of problems if I had blocked the layout first and then added all the details second. Maybe... I dunno, I'm still learning.
Maybe Necros Could Chime In
#12616 posted by SleepwalkR on 2013/03/25 16:07:51
He has the most experience with mapping with TB.
State Of The Map...
Ok, I chopped out *all* the terrain brushes, blocked in those areas to stop them leaking, fixed all the pointfile problems and got a build that vis'd... I have a couple of HOM's (that don't leak for some reason) and that's it.
So my todo is simply to try and remake the terrain without getting any leaks. I have a couple of ideas on how I might achieve this, I think the problem was the void-facing terrain were misshapen rather than nicely boxed and uniform.
It's a bit of a nightmare but I have definitely learned a lot in the last couple days.
#12618 posted by JneeraZ on 2013/03/25 17:15:46
Quake is ooooold and rickety. Expect problems. :) But it's fun...
Fifth
#12619 posted by SleepwalkR on 2013/03/25 17:51:25
If you can pull it off and learn some best practices, make sure to write them up on the QuakeWiki or here, please!
Best Practice?
The whole process has been a big ol' practice! That's what the map was about. I have keep a lot of the different stages of the map, so I can easily do a mapping retrospective.
As for the writing I don't really have a desire to write it up on the wiki, but I really want to have a go at making some youtube videos on the process I took, lessons learned etc.
I'll start pestering my film-maker friend until he caves in.
Pro-tips:-
-box out your map/layout
-use trenchbroom 1.05 from the start
-get your hands on TreeQBSP from http://user.tninet.se/~xir870k/
-become best friends with the "pointfile" command and "r_showtris"
TB 1.05
*does* seem to have better vertex accuracy for more standard shaped geometry, after I did all my pointfile detective-work I selected all the brushes and did a correct vertices & snap vertices... seemed to sort out the last remaining HOM's on the map.
Only trouble now is I have no terrain in my map after I brutally removed them all. I'm going to do it bit-by-bit and *hopefully* the map will have it's finalised geometry! Then I can get back to the fun stuff, testing, item placement etc. :)
Well
#12622 posted by SleepwalkR on 2013/03/25 18:17:58
I was talking about what to avoid specifically for TB. Apparently there are some things which will lead to bad brushes etc. and I'd like to know how that happened.
SleepwalkR
The map design videos would be TB-centric.
In terms of how to avoid bad brushes, who knows? I never took off grid snapping and I almost never went below the standard 16 units. I think some brushes just "become" malformed if you play with the vertices too much or you make shapes that are just too crazy!
Even some of the (very plain) curved walls which were grid aligned had become very very slightly nudged off the 16 unit grid. It wasn't until I did the correct/snap vertice trick that it was fixed.
Again, the tools used for pointfiling and treeq really helped me to get to a water-tight map. Though I think whatever you fixed between version 1 and 1.05 obviously helped a lot. I just hope I can redo the terrain, that *IS* after all one of the big selling points of TB ;)
Turns Out TB Was A Trojan Horse All Along
#12624 posted by negke on 2013/03/25 18:57:16
A QuArK in disguise!!!
Here's Some Advice...
I was using the co-ordinates and the pointfile to find leaks (as is the standard procedure), and to locate them easier in TB I was dropping in a monster_boss entity and entering the co-ordinates in the "origin" key, then when the entity moved I would delete and replace it with a shambler (again with the same co-ordinates), and then with a rottfish. Only last night at 3am when I was looking for the leaks I had forgot to take the shambler out the map when I compiled the fixed map... It scared the bejesus out of me! I decided that was a good time to stop mapping and sleep. :D
I Think
#12626 posted by SleepwalkR on 2013/03/25 20:45:21
those problems are mostly the result of imprecisions when loading the map. I'm working on an improved version of the vertex generation algorithm that will do a better job of avoiding degenerate edges and faces.
An edge is considered degenerate if it is very short (shorter than 0.5 units) and a face is considered degenerate if it is a triangle and contains such an edge. This should help avoid such problems, however I can't predict whether QBSP will do the right thing, too. But I think everything will be much better if I can switch TB to use only integer plane points.
Negke
#12627 posted by SleepwalkR on 2013/03/25 20:45:34
You're not too far off there, sadly.
#12628 posted by Spiney on 2013/03/25 21:31:29
those problems are mostly the result of imprecisions when loading the map. I'm working on an improved version of the vertex generation algorithm that will do a better job of avoiding degenerate edges and faces.
That might explain why some problematic brushes (one that caused a crash for instance) look completely different when reloading the map?
In terms of how to avoid bad brushes, who knows? I never took off grid snapping and I almost never went below the standard 16 units. I think some brushes just "become" malformed if you play with the vertices too much or you make shapes that are just too crazy!
Pretty much my experience.
One thing I think can be problematic is adding new vertices, in certain cases those can create strange n-gons that almost certainly will get screwed up after a while. Also dragging planes on a non-grid-aligned plane tends to end up badly a lot of times (especially in conjunction with above case). I guess that kind of makes sense when you think about the floating point issue...
Spiney
#12629 posted by SleepwalkR on 2013/03/25 22:27:12
That might explain why some problematic brushes (one that caused a crash for instance) look completely different when reloading the map?
It might ;-). I can't say without seeing an example.
It Almost Sounds Like...
#12630 posted by quaketree on 2013/03/25 22:33:45
you're running into the same type of problems that plagued Quoole 0.99 (creeping brushes). I hope that you can figure it out and fix it.
Possibly
#12631 posted by SleepwalkR on 2013/03/25 22:35:55
I promise that I will find a solution, even if that means that vertex editing has to be dumbed down a bit.
Dumbed Down Crushes Creativity
#12632 posted by sock on 2013/03/25 22:56:09
That would be a great shame because this is one of the best features of the editor. It would be a shame to fall into the trap of GTK 1.5 and force mappers to only create limited brush shapes because of vertex problems. Personally I prefer the ability to push brushwork to the limit and then have tools that look for problems afterward.
I personally use SDRadiant 1.8.3 and when I get issues with brushes I take the brushes and cut them. This forces the editor to re-generate all the planes and this usually fixes most of the problems. One other solution is I use an editor plugin called Bob's Tools. This goes through the whole map looking for potential brush plane errors and re-creates any brushes with problems. When it is finished it highlights what has changed (selected) so I can decide if to save/carry on.
Yes
#12633 posted by negke on 2013/03/25 23:07:12
I liked how Quest would allow me to make invalid-shaped brushes which I could sort out afterwards, whereas in GTK 1.5 I always have to go out of my way to create properly angled (rotated) or abstract stuff, because its validation system is so overly rigid and limiting.
#12634 posted by necros on 2013/03/25 23:40:58
hi, was very busy today and didn't have a chance to check func.
i'd really need to see how the map is made to be able to tell, but a lot of the time, i found just snapping the vertices seemed to always fix my problems. i never really did anything special to get it to work.
sorry i can't be of more help.
One Should Always Snap The Vertices
#12635 posted by negke on 2013/03/25 23:48:50
But it's not obvious to new mappers. When TB was released, I actually expected it to do this automatically - making perfectly valid, grid-snapped trisoup terrain/brushes without additional user input. Easier said than done apparently. I suppose a notification popup ("Do you want to snap/validate brushes?") before saving wouldn't be the smoothest solution, either.
Snapping Verts
#12636 posted by SleepwalkR on 2013/03/25 23:52:48
This is actually what I meant by dumb down. This would solve all problems, but there may be situations where the user wants off grid vertices?
Wanting Offgrid Verts.
most of the time I'd say you don't want off-grid verts, but I know for certain in my map there are a couple of instances where I *do* have them (on purpose). But you'd be unlikely to tell if you looked in the editor as they are clipped along the same plane as another brush that *is* grid snapped properly.
I think I'll stick with v1.05 and just run through the motions of checking at every stage.
|