big val in y axis while displaying nav mesh

[GoogleCodeExporter]

What steps will reproduce the problem?
1. while some of the config is set to certain val for example ,like max 
edge len or max sample error. however ,it is not easy do deside what val 
to use...
2.
3.

What is the expected output? What do you see instead?
the detour fun is ok ,the path is correct ,however, the nav mesh display 
is awesome, it seems that some y axis val of the detail mesh is not 
properly set. set the attachment

What version of the product are you using? On what operating system?
the latest in svn

Please provide any additional information below.
i hope it can be fixed soon, it is very annoying...

Original issue reported on code.google.com by zhibin.l...@gmail.com on 27 May 2010 at 8:38

Attachments:

• problem.bmp

[GoogleCodeExporter]

Is it possible to have that mesh (.obj) for testing, together with the values 
which
result bad coordinates? If you don't wish to share the mesh here, you can a 
link to
the file to my personal email address (you can find it in the readme.txt).

Original comment by memono...@gmail.com on 27 May 2010 at 8:47

[GoogleCodeExporter]

the obj is as attached, with the Max edge error set to 3.0, it always happens 
in my 
project, may be it is because we use very big size model, which lead to quite 
big 
val of cs...i don't know, i hope you can fix it ,thx. 

Original comment by zhibin.l...@gmail.com on 27 May 2010 at 9:58

Attachments:

• lava.obj

[GoogleCodeExporter]

I will take a look at it.

A simple solution on your side is to use better value for max edge error. Value 
of
3.0 way off and will make a lot of shortcuts, which may lead to other problems. 
Good
value for max edge error is somewhere between 1.1 and 1.5.

See this blog post for explanation of the different values:
http://digestingduck.blogspot.com/2009/08/recast-settings-uncovered.html

Original comment by memono...@gmail.com on 27 May 2010 at 10:05

[GoogleCodeExporter]

the same obj,with the config of cs = 3.0 , max_climb = 3.0, min region size = 
0, 
merge region size = 0....... crash!!!
it seems there are some bug in the function: simplifyContour, it require more 
and 
more memory here,and finally crash. Crash on the code below:

// If the max deviation is larger than accepted error,
            // add new point, else continue to next segment.
            if (maxi != -1)
            {
                // Add space for the new point.
                simplified.resize(simplified.size()+4);
......

i wish you can fix the problem too. thx again

Original comment by zhibin.l...@gmail.com on 27 May 2010 at 10:52

[GoogleCodeExporter]

Can you post your other generation values too?

Original comment by memono...@gmail.com on 27 May 2010 at 11:02

[GoogleCodeExporter]

as attachment, all the config, except, cs = 3, max_climb = 3, min_region_size = 
0, 
merge_region_size = 0 ,other param is set to default

Original comment by zhibin.l...@gmail.com on 27 May 2010 at 12:01

Attachments:

• config.JPG

[GoogleCodeExporter]

Please check the blog post how to select the values. If your cell size is 3.0 
then it
is odd that your agent radius is 0.6, also max climb should be less than agent 
height.

Original comment by memono...@gmail.com on 27 May 2010 at 12:05

[GoogleCodeExporter]

i am sorry, but can you provide your blog address? thx

Original comment by zhibin.l...@gmail.com on 27 May 2010 at 12:42

[GoogleCodeExporter]

See comment #3 above.

Original comment by memono...@gmail.com on 27 May 2010 at 12:45

[GoogleCodeExporter]

more example of bug:
1.it seems that there are sth wrong with the find path straight now, see the 
attachment wrong_path. while the find path iter is ok, the straight path is not 
right.

2. with certain config, as attachment wrong_detail, the detail mesh is not 
right.

totally speaking, the config of building is always boring me, i have to try 
different config again and again. i hope that the algorithm would be more 
robust. thx

Original comment by zhibin.l...@gmail.com on 28 May 2010 at 3:10

Attachments:

• detour.obj
• wrong_path.JPG
• wrong_detail.JPG

[GoogleCodeExporter]

In the first picture (wrong_path.jpg) the generation settings are way off. 
Maybe you are using old version. Please 
try the latest one from perforce.

I fixed the case detail mesh case which you describe in the second picture in 
change list R167.

Sorry to hear that you have had hard time trying to come up with good input 
parameters. In practice the only 
parameters you need to touch are the agent size and cell size. The others are 
good defaults. It all starts from the 
agent size, as the blog post describes.

Original comment by memono...@gmail.com on 28 May 2010 at 6:36

[GoogleCodeExporter]

it is so nice of you to fix the problem so quickly.
however, even when i use the latest version of demo.exe from the svn ,the 
problem 
still exists. see the attachment. 
we are applying recast to a new project, maybe more bug would be met in future.
we hope that we can make the recast better and better together.

thank you again

Original comment by zhibin.l...@gmail.com on 28 May 2010 at 8:23

Attachments:

• wrong_path_latest.JPG

[GoogleCodeExporter]

Sorry, I did the fix on my Mac, I did not update the demo exe file.

Original comment by memono...@gmail.com on 28 May 2010 at 8:39

[GoogleCodeExporter]

another problem.
since my scn is so large, i wonder that if i can use the tile_mesh, but 
applying 
diffent config to different tile. for example , while the character in the 
wild, i 
would use the bigger cs. if the character is in the town, i would like to use 
small 
cs to handle the more complex enviroment.
would it cause any problem?

Original comment by zhibin.l...@gmail.com on 28 May 2010 at 9:20

[GoogleCodeExporter]

In theory it should be possible, Detour can handle slight inaccuracy at tile 
borders. I have not tried it in practice.

Original comment by memono...@gmail.com on 28 May 2010 at 9:41

[GoogleCodeExporter]

Excuse me
i am confuse about the usage of m_saltBits.
especially in the code :
    if (m_saltBits < 10)
        return false;
why it should be wider than 10 bit?
can you also explain the meaning of m_tileBits and m_polyBits, thx

Original comment by zhibin.l...@gmail.com on 29 May 2010 at 8:28

[GoogleCodeExporter]

This blogpost explains how the handles work:
http://digestingduck.blogspot.com/2010/03/save-games-and-all-that.html

Original comment by memono...@gmail.com on 29 May 2010 at 10:30

[GoogleCodeExporter]

Closing this issue for now. If you have if you have further questions, please 
use the discussion group at http://groups.google.com/group/recastnavigation

Original comment by memono...@gmail.com on 8 Jun 2010 at 5:22

• Changed state: Done