[vlack]

Patch Title: vmap assembler secondary fix

Applies cleanly to: rev.2007

What bug does this patch fix: NPCs don't move or fall out of world, they don't attack you, you can shoot through walls, you fall through too many objects...

Detailed Explanation: diff'ed rev. 1134 and rev. 2007, server code of tile handling not changed (X and Y coords), assembler changed. If assembler is broken, revert the change -> it's fixed.

How to reproduce: Select NPC, hide behind a wall and try fireball/shadow bolt, something. Without the patch, you can shoot the NPC, with the patch it will say, target not in line of sight.

Link to thread in bug reports section if available:
NO

Patch:

--- trunk/src/collision_tools/assembler/vmap/TileAssembler.cpp  2009-01-17 15:59:55.000000000 +0100
+++ trunk/src/collision_tools/assembler/vmap/TileAssembler.cpp  2009-01-21 02:18:01.000000000 +0100
@@ -190,7 +190,7 @@
																		char buffer[100];
																		if(iCoordModelMapping->isWorldAreaMap(mapId) && x<65 && y<65)
																		{
-																			   sprintf(buffer, "%03u_%d_%d",mapId,y,x); // Let's flip x and y here
+																			   sprintf(buffer, "%03u_%d_%d",mapId,x,y); // Let's flip x and y here --- OR MAYBE NOT, it's good for the server as X,Y even if it loads them by Y,X - test confirmed it - by: VLack aka. VLsoft
																				dirname = std::string(buffer);
																		}
																		else
@@ -232,7 +232,7 @@
														char fullnamedestnamebuffer[500];
														if(nameCollection.iMainFiles.size() >0)
														{
-															   sprintf(destnamebuffer,"%03u_%i_%i.vmap",pMapId, pYPos, pXPos); // flip it here too
+															   sprintf(destnamebuffer,"%03u_%i_%i.vmap",pMapId, pXPos, pYPos); // flip it here too --- OR MAYBE NOT, it's good for the server as X,Y even if it loads them by Y,X - test confirmed it - by: VLack aka. VLsoft
																std::string checkDoubleStr = std::string(dirfilename);
																checkDoubleStr.append("##");
																checkDoubleStr.append(std::string(destnamebuffer));

Delete the vmaps dir's content on the client, and do the assembling with the assembler patched with this. Copy it to server, test it and comment on it here!!!

3.0.3 enGB client data size should be like this:
DBC: 95,925,888 bytes in 255 files
maps: 1,520,923K bytes in 119 files
vmaps: 1,737,428K bytes in 4249 files

Attached File(s)

•  vmapassembler_rev2007.patch (1.26K)
  Number of downloads: 154

[vlack]

A little note though:
this patch still does not fix the issue, that you won't get removed from your horse when you go into a building. I think Whoami is right, the extractor still misses some data from 3.0.3, this is one of those informations...

[vlack]

Already 12 downloads, and no comments so far? I need those test results for continuous development. It seems to be working properly, but I have to know if someone found something bad, because then I have to re-check my theory and the code.

[lakilak]

it works good!!! good job, found some bug:
in caves/dungeons when some monsters attack you and
when you leave combat but not killing monster, when monster
come to their initial position you can't attack him again.
Neither monster attack you again, but this is happen only on some monsters
not all...
Thanks for the patch btw!!

P.S. sorry on my english.

[vlack]

lakilak, on Jan 21 2009, 06:51 PM, said:

it works good!!! good job, found some bug:
in caves/dungeons when some monsters attack you and
when you leave combat but not killing monster, when monster
come to their initial position you can't attack him again.
Neither monster attack you again, but this is happen only on some monsters
not all...
Thanks for the patch btw!!

P.S. sorry on my english.

Your english is quite readable (much better than many other's).

I think this "leaving-combat and can't attack next time the same mob" problem is not related to vmaps.

To others:
Northrend might got more bugs, Kalimdor and Eastern Kingdom seems to be OK.
If you use this vmap assembler, you had to use the ArcEmu-supplied ad.exe and the vmapextractor_v3.exe to make the buildings directory (these are the ones you can find in the SVN).

[lakilak]

vlack, on Jan 21 2009, 09:07 PM, said:

I think this "leaving-combat and can't attack next time the same mob" problem is not related to vmaps.

I think it is, because:
1. when wall is between his route he stack at wall and can't back to start position
2. is he manage to find a way to his initial position when I try to attack him
have this message: "Target not in line of sight." even if I stand next to him

[vlack]

lakilak, on Jan 21 2009, 09:16 PM, said:

I think it is, because:
1. when wall is between his route he stack at wall and can't back to start position
2. is he manage to find a way to his initial position when I try to attack him
have this message: "Target not in line of sight." even if I stand next to him

You seem to be right! (or at least this makes sense to me)
It's true, that mobs don't have a working path-finding algorithm, they just march back to their start position in a straight line... Maybe with this behavior something gets borked if they cross walls or other normally non-passable objects on their way back

[hummin]

vlack, on Jan 21 2009, 06:20 PM, said:

Already 12 downloads, and no comments so far? I need those test results for continuous development. It seems to be working properly, but I have to know if someone found something bad, because then I have to re-check my theory and the code.

it works well.

thank you very much!

Hummin

[insultant]

Another excellent fix by Guru Vlack.
Thanks for relieving this pain.

[sergio066]

Hello vlack here are my results.

Working

- Hills are working fine

- You cannot attack NPC through the wall means LOS char to NPC working fine

Not Working

- Starting DK zone not working with collision enabled you will fall through Heart of Acherus

- NPC will attack you through the wall if you get close to them Tested in Halls of Lightning and Stone Watch Keep means LOS NPC to char not working ,

I used all your files and of course collision is enabled !!!!!

[vlack]

sergio066, on Jan 22 2009, 05:52 PM, said:

Hello vlack here are my results.

Working

- Hills are working fine

- You cannot attack NPC through the wall means LOS char to NPC working fine

Not Working

- Starting DK zone not working with collision enabled you will fall through Heart of Acherus

- NPC will attack you through the wall if you get close to them Tested in Halls of Lightning and Stone Watch Keep means LOS NPC to char not working ,

I used all your files and of course collision is enabled !!!!!

Thank you for the testing and the report!

I suspect that the "new content" (which were not present in 2.4.3, like Northrend and DK zone) will surely have problems, it's maybe an extractor issue (Whoami told me about a few things, so the extractor needs some improvement too, maybe building walls get missed or viewed as transparent, or something).

First I'll look into "NPC to Player" LOS checks in the core, especially around Stone Watch Keep (as if you had problems there, it's the perfect place to test out things).

So again, thank you for the feedback.

[santopazzo]

You tried to fix maps with landwalker? with new collision dll?

[vlack]

santopazzo, on Jan 27 2009, 09:52 AM, said:

You tried to fix maps with landwalker? with new collision dll?

Umm... What?
New collision dll? Where? I haven't seen new SVN commits about vmap management...

This patch fixes the vmap assembler. It has nothing to do with landwalker; I never used that, vmaps are just good as they are, as they come out of the fixed assembler.

[santopazzo]

landwalker is an application that simulate the vmaps & collision dll use and fix the fall bugs...
but it stop work at 5xx maps

sorry for my english

its a tool after the vmap assembler to do extra fix and test

[Hoffa]

vlack, on Jan 21 2009, 09:20 PM, said:

You seem to be right! (or at least this makes sense to me)
It's true, that mobs don't have a working path-finding algorithm, they just march back to their start position in a straight line... Maybe with this behavior something gets borked if they cross walls or other normally non-passable objects on their way back

Should probably implement something like this:


I think wayland is working on it.

Walkable areas are automatically generated (most likely) with tracers that determits the angle of the surface using normals.

[santopazzo]

this is the result after the use of landwalker:

Quote

This program tries to increase Vmap acuracy by borrowing info from collision map
s
This program does not solve multi level issues.
Map list is predifined in source. You need to edit it to change list
========================================================
Processing map 30
Map finetuning report for map : 30
fall underground fixes: 263
avg height dif change : 4.998439
========================================================
Processing map 33
========================================================
Processing map 36
Map finetuning report for map : 36
fall underground fixes: 3440
avg height dif change : 7.934053
========================================================
Processing map 37
Map finetuning report for map : 37
fall underground fixes: 61
avg height dif change : 9.636978
========================================================
Processing map 47
Map finetuning report for map : 47
fall underground fixes: 10209
avg height dif change : 11.473544
========================================================
Processing map 129
========================================================
Processing map 169
Map finetuning report for map : 169
fall underground fixes: 6776
avg height dif change : 10.516128
========================================================
Processing map 189
========================================================
Processing map 209
========================================================
Processing map 269
Map finetuning report for map : 269
fall underground fixes: 7754
avg height dif change : 9.156030
========================================================
Processing map 289
Map finetuning report for map : 289
fall underground fixes: 8675
avg height dif change : 10.231026
========================================================
Processing map 309
========================================================
Processing map 329
Map finetuning report for map : 329
fall underground fixes: 12369
avg height dif change : 9.898681
Could not load header for map at ./maps/ with id 451
========================================================
Processing map 469
========================================================
Processing map 489
Map finetuning report for map : 489
fall underground fixes: 10935
avg height dif change : 10.244144
========================================================
Processing map 509
========================================================
Processing map 529
Map finetuning report for map : 529
fall underground fixes: 3649
avg height dif change : 6.884388
========================================================
Processing map 531
Map finetuning report for map : 531
fall underground fixes: 14862
avg height dif change : 9.607992
========================================================
Processing map 532
========================================================
Processing map 533
========================================================
Processing map 534
========================================================
Processing map 543
========================================================
Processing map 559
Map finetuning report for map : 559
fall underground fixes: 961
avg height dif change : 8.316976
========================================================
Processing map 560
========================================================
Processing map 562
Map finetuning report for map : 562
fall underground fixes: 11257
avg height dif change : 7.199901
========================================================
Processing map 564
Map finetuning report for map : 564
fall underground fixes: 80
avg height dif change : 14.439635
========================================================
Processing map 566
Map finetuning report for map : 566
fall underground fixes: 2346
avg height dif change : 6.552955
========================================================
Processing map 568
========================================================
Processing map 572
Map finetuning report for map : 572
fall underground fixes: 12470
avg height dif change : 6.670658
========================================================
Processing map 580
Map finetuning report for map : 580
fall underground fixes: 28022
avg height dif change : 10.843597
========================================================
Processing map 585
Map finetuning report for map : 585
fall underground fixes: 22724
avg height dif change : 10.450397
Finished all maps

now i have to try this vmaps in the server ^^

[vlack]

santopazzo, on Jan 27 2009, 10:04 PM, said:

this is the result after the use of landwalker:
...
now i have to try this vmaps in the server ^^

Well, now that you told me about this, I'll try it too when I have time
Just for curiosity, how long did it take for landwalker to do the fixes on the vmap data? (An estimated time would be great, and please tell me, on what CPU you achieved that time?)

[Muffler]

hmm
Inn not work i mean char not have Zzz

[vlack]

Muffler, on Jan 29 2009, 07:04 PM, said:

hmm
Inn not work i mean char not have Zzz

This is the same issue as you won't get removed from your horse.

You get Zzz in cities though (that's another type of check I think, that's why it's working).

Current vmaps data is only good for:
-Terrain height check
-LOS check

The other checks (IsIndoor, etc.) currently does not function on it. I think this is a 2.4.3->3.0.3 format change issue, so it needs additional work (maybe even on the extractor, we'll see). There's a change in how the in-door status is checked on a tile (TileAssembler.cpp), so maybe I should look into that first.

[santopazzo]

a lot of time i dont remember but 12h for sure btw i have a old PentiumD 3ghz