[ ZDBSP] UDMF maps keep doing this to me...

[Ceeb]

Here you go. It's map 4, obviously. The player start is directly in front of the culprit area. (By the way, I get the same exact problem when running this stripped-down version of it. )

[Project Shadowcat]

There's HOM everywhere even in ZDoom. Though I'm certain you can make the area a little less complex, I think you broke ZDBSP dude.
ED: You're using GZDoom. I say strip all of the sectors regarding light and stick an ambient light Thing or a few there at the window.

[Sonnyboy]

Yes, I get the same HOM.

I think you overdid the light-bleeding window.

[Ceeb]

Well, I do complex shit like that all over the place in some maps, and those are absolutely fine.

[Project Shadowcat]

Well, I do complex shit like that all over the place in some maps, and those are absolutely fine.

"Shit happens?" Randy and Graf have both said that program isn't perfect.

[Sonnyboy]

Try drawing over the sectors again. It could be human error, a bleeding sector.

I've done that more times than I can remember, and redoing the sector seems to fix it.

[Project Shadowcat]

Try drawing over the sectors again. It could be human error, a bleeding sector.

I checked it in DB2, all of the sectors are fine.

[Ceeb]

I guess I just need to tone down the light diffusion a bit.

[Sonnyboy]

Perhaps not...

I was able to get rid of a lot of it by deleting some unneccesary vertices.
Curious...

[Ceeb]

Well, I did away with the stairs and made it a regular hallway, kept the light diffusion, now it works.

Thanks for trying to help, y'all.

[Enjay]

I'm not to familiar with DB2 error checking and so on but, yeah, it certainly gave the area a clean bill of health. I then copied the map to ZdoomHexen format and the problem vanished. I was also able to check the map using deePsea in that format and although the map did have errors, none of them should obviously have been affecting that particular area.

Given that the area worked in ZdoomHexen format and not in UDMF, it might well be a bug either with UDMF or Zdoom's implementation of it so I think it would be worth flagging up for checking by Graf/Randy.

[Graf Zahl]

I think you broke ZDBSP dude.

That would be my guess, too.

[Kappes Buur]

Well, I did away with the stairs and made it a regular hallway, kept the light diffusion, now it works.

Thanks for trying to help, y'all.

You could have kept the stairs. Just get rid of unnecessary detail in the lighting aspect and all is fine.

---

Especially, when the detail isn't really visible

---

It's the tiny, intricate sectors that screw things up, ... most of the time.
Sonnyboy was on the right track.

[Graf Zahl]

I activated the internal node builder's debug output and here's where it seems to lose it:

Code: Select all

Seg   186, ld 116 ( -704,-1248)-( -704,-1223) scores 0
Seg   241, ld 158 ( -704,-1248)-( -720,-1259) scores -32454
Seg   256, ld 172 ( -752,-1280)-( -757,-1275) scores -293753
Seg   341, ld 235 ( -976,-1328)-( -839,-1299) scores 0
Seg   343, ld 236 ( -839,-1299)-( -752,-1280) scores 0
Seg   344, ld 237 ( -839,-1299)-( -844,-1296) scores -458298
Seg   346, ld 238 ( -976,-1328)-( -844,-1296) scores -45700
Seg   348, ld 239 ( -844,-1296)-( -757,-1275) scores -113682
Seg   350, ld 240 ( -757,-1275)-( -736,-1270) scores -1109
Seg   352, ld 241 ( -844,-1296)-( -848,-1294) scores -1262964
Seg   354, ld 242 ( -736,-1270)-( -752,-1280) scores -21715
Seg   355, ld 243 ( -757,-1275)-( -762,-1270) scores -309330
Seg   362, ld 247 ( -976,-1328)-( -848,-1294) scores -28216
Seg   364, ld 248 ( -848,-1294)-( -762,-1270) scores -67784
Seg   366, ld 249 ( -762,-1270)-( -720,-1259) scores -1171
Seg   369, ld 251 ( -848,-1294)-( -852,-1292) scores -896411
Seg   371, ld 252 ( -762,-1270)-( -765,-1266) scores -99045
Seg   386, ld 260 ( -976,-1328)-( -852,-1292) scores -18213
Seg   388, ld 261 ( -852,-1292)-( -765,-1266) scores -30403
Seg   390, ld 262 ( -765,-1266)-( -704,-1248) scores -7082
Seg   392, ld 263 ( -852,-1292)-( -864,-1285) scores -358301
Seg   394, ld 264 ( -765,-1266)-( -779,-1252) scores -393698
Seg   408, ld 271 ( -976,-1328)-( -864,-1285) scores -256353
Seg   410, ld 272 ( -864,-1285)-( -779,-1252) scores -235772
Seg   421, ld 278 ( -864,-1285)-( -881,-1276) scores -709541
Seg   423, ld 279 ( -779,-1252)-( -798,-1230) scores -354531
Seg   437, ld 286 ( -976,-1328)-( -881,-1276) scores 0
Seg   441, ld 288 ( -798,-1230)-( -704,-1177) scores 0
set 186, step 0, nosplit 1 has no good splitter (1)
Seg   186, ld 116 ( -704,-1248)-( -704,-1223) scores 0
Seg   241, ld 158 ( -704,-1248)-( -720,-1259) scores -32454
Seg   256, ld 172 ( -752,-1280)-( -757,-1275) scores -293753
Seg   341, ld 235 ( -976,-1328)-( -839,-1299) scores 0
Seg   343, ld 236 ( -839,-1299)-( -752,-1280) scores 0
Seg   344, ld 237 ( -839,-1299)-( -844,-1296) scores -458298
Seg   346, ld 238 ( -976,-1328)-( -844,-1296) scores -45700
Seg   348, ld 239 ( -844,-1296)-( -757,-1275) scores -113682
Seg   350, ld 240 ( -757,-1275)-( -736,-1270) scores -1109
Seg   352, ld 241 ( -844,-1296)-( -848,-1294) scores -1262964
Seg   354, ld 242 ( -736,-1270)-( -752,-1280) scores -21715
Seg   355, ld 243 ( -757,-1275)-( -762,-1270) scores -309330
Seg   362, ld 247 ( -976,-1328)-( -848,-1294) scores -28216
Seg   364, ld 248 ( -848,-1294)-( -762,-1270) scores -67784
Seg   366, ld 249 ( -762,-1270)-( -720,-1259) scores -1171
Seg   369, ld 251 ( -848,-1294)-( -852,-1292) scores -896411
Seg   371, ld 252 ( -762,-1270)-( -765,-1266) scores -99045
Seg   386, ld 260 ( -976,-1328)-( -852,-1292) scores -18213
Seg   388, ld 261 ( -852,-1292)-( -765,-1266) scores -30403
Seg   390, ld 262 ( -765,-1266)-( -704,-1248) scores -7082
Seg   392, ld 263 ( -852,-1292)-( -864,-1285) scores -358301
Seg   394, ld 264 ( -765,-1266)-( -779,-1252) scores -393698
Seg   408, ld 271 ( -976,-1328)-( -864,-1285) scores -256353
Seg   410, ld 272 ( -864,-1285)-( -779,-1252) scores -235772
Seg   421, ld 278 ( -864,-1285)-( -881,-1276) scores -709541
Seg   423, ld 279 ( -779,-1252)-( -798,-1230) scores -354531
Seg   437, ld 286 ( -976,-1328)-( -881,-1276) scores 0
Seg   441, ld 288 ( -798,-1230)-( -704,-1177) scores 0
set 186, step 0, nosplit 0 has no good splitter (1)
 - seg 186(-704,-1223)-(-704,-1248) line 116 front 74305984 back 74306356
 - seg 241(-704,-1248)-(-720,-1259) line 158 front 74303752 back 0
Need to synthesize a splitter for set 186 on seg 241
Subsector from set 186

After that last line this set is never being looked at again even though it still contains lots of lines that most definitely do not form a convex region.

[Graf Zahl]

This seems to be caused by the SplitCost setting. If it's at 8 the nodes for this map are broken.
However, if I set it to 4 they build correctly.

But what I don't know is how this can be handled dynamically on a case-by-case basis if some map does not build anymore with the higher setting.

At least it's good to know that it's not a floating point precision issue because then we would have had a real problem.