Map Bug

[Szymanski]

I've made a small area with crates, light sources etc. No sectors are unclosed and all checks (except texture) give it the all clear.

At the bottom corner of the map the player is blocked, yet theres no sectors/linedefs blocking the way. Is the map being screwed up when building the nodes?, i've never had this problem before.

Adding another sector in this area fixes the problem, but i'd rather know what is breaking and avoid it. Any ideas?

[beezee88]

Hmm, it could be the nodes...

Have you tried all node builders? Which one do you use?

[Szymanski]

Panic over, BSP5.1 fixes it. Was using Deepbsp.

[beezee88]

Thought so. I have noticed that BSPs and NODEs that are integrated into editors, well, don't work. DoomCAD and Deep, to name two.

[deep]

FYI, it's NOT integrated into the editor and never has been. Also there are options that most people don't use (accessible only via the GUI).

Only an old version of DeePbsp had that problem, but I suppose it's possible if a "vertex" is mathematically ambiguous - which can happen. Overall, DeePbsp is superior to BSP 5.1.

Both pgms were derived from the same source, DeePbsp in 1994 with various changes over the years:)

PS: It's possible for a sector to not show any "closed" errors, yet is is unclosed. That would also screw up the results.

[Brad_tilf]

I find that zennode usually gives superior results

[Szymanski]

I'm using 11.66 which is pretty up to date. I'll stick to BSP until any other problems turn up.

[deep]

Well, that's true for Legacy (since he worked specifically on some issues for that port), but not as a general statement. You'd have to compare various large levels to make that claim. And I've never seen any person that did a valid comparison. Only one's I've seen are invalid, since they ran it outside of the GUI using defaults that are really not appropriate.

Overall, for most levels (not Legacy), there's really not much "superior" about any of the standalone nodebuilders (which DeePbsp is) - except for speed or if you think rejects are a big deal (I don't). Some levels go 1 way, others some other way. I've run 100's of objective comparison and play tests over the years:)

I'd be interested in seeing that level to see the area in question. As I said, it's a matter of a line being calculated to be on the "wrong" side or there is an invalid reference that is not obvious.

[The Ultimate DooMer]

If you have a raised sector near to the bottom corner, you could have the linedef connecting them pointing towards the higher sector. Flip the linedef (so it points towards the lower sector) and it should be ok. (age-old bug)

[Szymanski]

Map sent. let me know any probs

[deep]

Thanks. It's the first one I mentioned - ambiguous side resolution causing it to not splitting the bottom left corner of the sector 66 box. It's the same problem Randy mentioned if a "thing" is right on top of a line - which side to pick. Anyway, if you move sector 66 2 units East (to the right) it's fixed.

This will make a good test map for me to look at. Maybe can make a slight change or alternate method.

[Sharessa]

Hmm...I just use the WA nodebuilder...it seems to work perfectly 99% of the time. Unfortunately, the other 1% of the time, something goes HORRIBLY wrong. But thats what backup files are for.

[Amaster]

WA's nodesbuilder isnt very "efficent" in my experience (I dont know if thats the right word). It seemd to increase the likelyhood of VPOs in an area with lots of 2s lines when compared to BSP.

[stphrz]

I've made a small area with crates, light sources etc. No sectors are unclosed and all checks (except texture) give it the all clear.

At the bottom corner of the map the player is blocked, yet theres no sectors/linedefs blocking the way. Is the map being screwed up when building the nodes?, i've never had this problem before.

Adding another sector in this area fixes the problem, but i'd rather know what is breaking and avoid it. Any ideas?

I used to get that sort of problem all the time when I was using WadAuthor's internal node builder. Every map I made it happened at least once. Adding another sector or moving a vertex slightly always fixed it.

[deep]

Ok - DeePBSP is fixed. I made a stupid mistake - in trying to speed things up, I avoid sqrt by looking at the square instead. But brain was dividing - as in x/2 <= 1, but what I really need to think about was sqrt[x] <=1.

Funny I never ran into the problem over all these years. And then all of a sudden the level in this post and then a GLBOOM issue in MAP16 of ETERNAL turns out to be the same thing. Level has to have angled objects and then it has to intersect where the above check makes a difference.

Now I wonder how many LEGACY things this also fixes? Anyway will post later today. Hope nothing else is fuxered by the change:) I'll make a new topic when it's done.