JPLambert
#1863 posted by aguirRe on 2004/05/03 11:27:47
Have you fullvised the map to see if the problems persist? Many maps expose these problems when not vised, so the engine must render too much.
Try also enabling the r_speeds cvar in GLQuake and check out the epolys value, this indicates how many entity polys that are visible. If too many, you might have to consider reducing visibility.
My suspicion is that this is your basic problem. You can also run Win/TyrQuake, enabling r_draworder and see how much of the map that suddenly becomes visible.
Try also a Google search of "packet overflow" quake and see what turns up ...
Engine Refers To A Rebuild
#1864 posted by HeadThump on 2004/05/03 11:40:29
the Quake executable. You wont get packet overflows with DarkPlaces because entity management is handled by a higher bit extension then Quake was originally built in. So you could have several thousand models in a single level if you so desired.
AguirRe
#1865 posted by JPL on 2004/05/03 11:45:55
Hello,
Yes, I made the try running full process (BSP + VIS + LIGHT)... and the problem remains....
So, considering all previous advices, I will try to enable r_speed cvar in FitzQuake in order to check out epolys value, but, what is the maximum value value I must reach.. some says 800, others 2000 (If I remember correctly, Vondur said that in a previous post...)... but what's the good one according to your experience???
And, if I need to reduce visibility, how can I do ??
Thanks a lot..
Bye
There Used To Be
#1866 posted by aguirRe on 2004/05/03 13:30:54
certain recommended values to aim for (especially in DM maps), but I doubt they're still valid today.
What I meant was to use these numbers and methods to quantify your current situation, which seem to choke the engine. When you see the numbers go down, you know you're making progress.
Reducing visibility can be done by using vis blockers, i.e. donut designs, walls etc. Think of a large open arena with many small objects (faces or entities) as the worst possible scenario for vis or the engine.
Look at many older Q1 maps and you'll see that most of them have a layout and entity amount/placement that prevents the arena situation.
I believe one of the challenges in Q1 mapping is to make interesting and fun maps while still keeping them playable in most engines.
AguirRe
#1867 posted by JPL on 2004/05/04 02:18:54
Hello,
I did it last night, I mean trying r_speeds value... after a full process (bsp + fullvis + light) there seems to be some very big value... As you felt, in the main big area, the r_speeds number increases up to 10000 (more or less..) I achieved to reduce it down to 1500 by an additional wall that cut the visibility....
But still remaining the problem of packet_overflow and sound disapperance... so I tried removing all monsters... packet overflow error disappear, but sound is still "inaudible"... and other problem, a big HOM problem appears at the beginning of the map while it doesn't exists before .... (and no errors/warnings occured during compilation...)
It's very discouraging to see these problems incoming while the map is almost achieved...
Anyway, I will try removing back all groups, one by one, to find where is located the problem... with a bit of luck, it can work...
Bye
JPLambert
#1868 posted by aguirRe on 2004/05/04 04:59:20
Send me the zipped map+wad and I'll take a look at it.
OK
#1869 posted by JPL on 2004/05/04 05:03:23
I will do it this evening... I'm at my job office now, and till to 17:00 ...
Thanks again for your help...
AguirRe
#1870 posted by Mike Woodham on 2004/05/04 15:17:07
This is probably one for you:-
I have a sealed i.e. no leaks, map of 2600 brushes. I do get several Cut_Node_Portal erros but the map runs fine in its own right.
I have a second sealed map of 600 brushes, which again runs fine in its own right.
So I import smaller map into larger map and get a leak through a solid brush when building Hull 1.
I have changed the brush concerned, overlapped surrounding brushes etc. The brush is one of many (terrain type triangles) on the 'floor' of the level but I have also added a large 'sealing' brush beneath the terrain to try to avoid the misaligned brush problems (this is not one of my large terrains!), which has been in place long before the import.
The only slightly unusual thing is that at present the imported map is entirely unconnected in any way to the larger map: I have imported it just to see how I can fit it in. I have not exceeded the world boundary - close to the edge but 32 units to spare.
If I delete the imported map (select Group and CTR X) without saving or reloading, I can immediately compile without leaks.
Bit long winded... any pointers? Anyone?
And I still have one more section to import!
Not Exactly Any Help
#1871 posted by aguirRe on 2004/05/04 15:41:38
but maybe some kind of attempt to explain why it's happening. The compiler quickly translates all brush faces in the map file into infinite planes with a normal (perpendicular to the plane) that specifies direction and a distance from origin (0 0 0).
This pair identifies the plane and they're represented by float values (in my compilers 64 bit double precision). One can rather easily suspect that the more of these infinite planes there are in a map, the more risk of planes getting too close (i.e. coplanar) for comfort due to numerical inaccuracies.
I believe that this might be one reason why bigger maps sooner or later start to exhibit weird leaks or clipping errors. I know maps where it appears nigh impossible to get rid of these problems.
In your case, you have two maps that you merge together and both are terrain generated, which means many triangular faces and as I've mentioned before, this appears to be extra provoking for the compiler.
This is all only a theory, though; I can't substantiate it. Still I recommend to inspect the areas of the warnings and try to get rid of them, in my experience the warnings seem valid and should be corrected.
If you can't proceed, send me the map and I'll see what I can do.
AguirRe
#1872 posted by Mike Woodham on 2004/05/04 16:10:45
Thanks.
The build process for the enlarged map was:-
1. small terrain (honestly!)
2. then one self contained non-terrain map
3. then the linking brushes created on-the-fly in the new map
4. then the attempt to bring in the later map
Each map compiled ok on its own.
I think perhaps I will just release the terrain + one section as a stand alone 'quickie'. At least I can cram it with monsters and end up with a Serious Sam style arena fight at the end.
It seems that I am fighting a losing battle here with the limits of the tools ... I know "a bad workman blames his tools".
Trouble is, time is tight and I go in three weeks.
You're welcome to have first look as you have helped a lot over the last few weeks.
Thanks again.
One Thing I Would Try..
#1873 posted by glassman on 2004/05/04 17:02:54
is moving the imported section nearer in than 32 units from the edge. I seem to recall having problems myself with brushes that close.
Tried...
#1874 posted by Mike Woodham on 2004/05/06 14:51:44
... but no good. I am working with very little room left in the 4096 world-space as this has become a very 'flat' map. Ho hum, start map and three small additional maps then.
Two general questions:
How many units from the player can a generated sound be heard? I do not want the player to hear monsters being teleported in so am wondering how far away from the spawn site I need to set the trigger.
Can I switch off the monster-spawn sound from within the monster's .qc file? I cannot find a reference to the spawn .wav (r_tele5.wav?) in the .qc but has anyone tried this? Maybe it's hard coded in the engine?
Tele Sounds
#1875 posted by czg on 2004/05/06 15:11:56
The function that plays ALL the teleport sounds is play_teleport() on line 282 in triggers.qc, so if you wanted to make ALL teleports (monsters, player) silent, you could just remove the contents of that function. (Or at the very least, comment away line 299, you might want to keep line 300 though...)
Czg
#1876 posted by Mike Woodham on 2004/05/06 15:57:05
Tusen takk!
(sum total of my trip to Hemsedal last Christmas)
Heh
#1877 posted by czg on 2004/05/06 16:33:32
heh
Wait A Second...
#1878 posted by metlslime on 2004/05/06 17:56:19
Isn't there a "silent" spawnflag on the teleporter?
.
#1879 posted by necros on 2004/05/06 18:34:52
yes, but that only gets rid of the 'humming' ambient sound that trigger_teleports emit.
Teleport Sounds
#1880 posted by Fern on 2004/05/06 19:25:14
Zerst�rer has a spawnflag for truly silent teleports.
GetFileSpace
#1881 posted by madfox on 2004/05/06 20:50:41
My last map suddenly run out of light.
What can I do?
light.exe
142entity - 130lights
+++ERROR+++
GetFileSpace : overrun
What Light Are You Using?
#1882 posted by necros on 2004/05/06 22:24:06
if you're not, i highly suggest using aguire's light util, because it seems you have hit a limit which aguire's stuff probably removed.
Aguire's Light Tool
#1883 posted by madfox on 2004/05/06 23:33:19
I deleted 16 torches, and the problem was solved. Could be something in the animated light count.
Never had the message before.
If It's My Light
#1884 posted by aguirRe on 2004/05/07 03:02:01
tool, it must be a very old version < 1.14 (Aug 2002). Later versions have another more informative error message.
Q1 Tools Update
#1885 posted by aguirRe on 2004/05/07 06:23:55
at http://user.tninet.se/~xir870k . Main features are new attenuation formula (delay 5) and improved additive minlight in Light, engine matched validation in compilers and a new GLQuake engine especially made for mappers. Please also see readmes for more details.
Any comments are welcome.
Gtkrad 1.5
#1886 posted by glassman on 2004/05/07 13:00:46
I'd like to start using GTKRadiant 1.5 but turning off 'Freelook in camera view' doesn't seem to work in the Preferences. Is it just me having that problem?
AguirRe
#1887 posted by glassman on 2004/05/07 13:02:03
What are the mapper specific features in your version of glquake?
|