0.5 bug report

[bappy]

first , it would be cool if we can have a bug report section in the forum....

second , here is some bug i found in the 0.5 release... say me if you have the same or if it s my card (geforce ti 4200)

...direct x 8 and 9 renderer:

problem with shadow:
-NO SHADOW in Fulscreen mode...
- there is a little bug that we can clearly see in the techdemo when there is the shadow of the running sydney and sphere map sydney. if you look at each shadow volume, we can see a second tiny shadow objet in black that is at the opposite of the light for every shadow caster... it s strange, it s like having the objet that project shadow in tiny and black...

... opengl

i have a problem with the 3dsmax famous dragoon... when the light is enable, in irrlicht , the head vertex color are purple , just the head, all other are good... and if i disable light, all is ok and vertex color are ok
very strange, this problem just appear with this object...

...directx AND opengl

when light is enable and shadow....when we look a shadow caster object, there is some little black point that appear in the texture of the object when the camera is looking the face that is in shadow...

if you want niko, i can send you a program that will show you all this problem.. maybe it will be clearly

[niko]

I know the first problem, but the two others are really strange. Yes, send the program, would be interesting.

[Zaelsius]

More bugs:

DirectX 8.1 and 9.0b

- No shadows in fullscreen mode.
- Horrible performance(but shadows ok) on windowed mode.
- Mipmapping is wrong. It happens in both tech-demo and tutorials. The mipmaps are used too near of the camera. Here is a screen shot:




OpenGL 1.2

- Ok. Haven't tested it too much, just Tech-demo on fullscreen.

[level1]

Hi here's another thing

the quaternion multiplication code is not correct, there are some minuses too much ( sorry for my wrong english)

Code: Select all

//! multiplication operator
inline quaternion quaternion::operator*(const quaternion& other) const
{
quaternion tmp;

tmp.W = (other.W * W) - (other.X * X) - (other.Y * Y) - (other.Z * Z);
// changed the following lines
tmp.X = (other.W * X) + (other.X * W) + (other.Y * Z) - (other.Z * Y);
tmp.Y = (other.W * Y) + (other.Y * W) + (other.Z * X) - (other.X * Z);
tmp.Z = (other.W * Z) + (other.Z * W) + (other.X * Y) - (other.Y * X);

return tmp;
}

The idea with the IOSOperator is quiet cool, congratulation

cheers
level1

[Guest]

There is still a visible corner of the sky box (at least in DX8.1 in fullscreen mode)

In fullscreen DirectX8.1 works good in techdemo but OpenGL almost stops...
I know that GeForce2 is not the best video card but still... 0.4.2 worked somehow

Does anyone have the same problem?

[thesmileman]

I do not have the sky problem and I have a GF2 as well

[Guest]

strange...
look carefully in techdemo. look on the sky...
there are tree small rays in the "corner" going from the same point. I can see it . In another corner I can see just one ray.
Actually, rays are segments.

I can send you screen shots via email...

[Zaelsius]

Yes, I can see the skybox corners too. I have GeForce 4 MX

[Guest]

BTW, there is one more obvious... with CGUIFileOpenDialog, actually with its ScrollBar...
It doesn't calculate the size of the scroll thumb correctly until you don't move it. Initially it's always very thin. Try Userinterface demo.

And one more: it's not a bug but a useful feature. It would be very useful if to implement keyboard events handling in listbox. Like KEY_UP, KEY_DOWN, etc...

[niko]

Sounds that there is a lot to do again.

[buhatkj]

i have checked it all carefully against other implementations and the formulas, and im pretty sure it's all correct. i think the trouble is probably elsewhere.

i have noticed before in irrlicht when doing movement code based on rotating around the y axis that i have to invert the z coordinate for the vector to be correect. is that normal or could it be related to this?

i'm not too strong with math, but i try...
-Ted

[Guest]

From http://www.euclideanspace.com/maths/alg ... /index.htm

The multiplication rules for quantities are:

i*i = j*j = k*k = -1
i*j = k,
j*i = -k
j*k = i,
k*j = -i
k*i = j,
i*k = -j

So for the general case just expand out the terms and group as follows:


(a + i b + j c + k d)*(e + i f + j g + k h)

=

a(e + i f + j g + k h)
+ i b(e + i f + j g + k h)
+ j c(e + i f + j g + k h)
+ k d(e + i f + j g + k h)

=

a*e + i a*f + j a*g + k a*h
+ i b*e + i*i * b*f + i*j * b*g + i*k * b*h
+ j c*e + j*i * c*f + j*j * c*g + j*k * c*h
+ k d*e + k*i * d*f + k*j * d*g + k*k * d*h

=

a*e + i a*f + j a*g + k a*h
+ i b*e - b*f + k * b*g -j * b*h
+ j c*e - k * c*f - c*g + i * c*h
+ k d*e + j * d*f -i * d*g - d*h

=

a*e - b*f - c*g - d*h
+ i (b*e + a*f + c*h- d*g)
+ j (a*g - b*h+ c*e + d*f)
+ k (a*h + b*g - c*f + d*e)

note the signs!

my conclusion: the irrlicht implementation is not correct.

cheers

level1

[VeneX]

I am sorry but the mipmapping is bad, especially in huge levels.
See http://irrlicht.sourceforge.net/phpBB2/ ... php?t=1244


Other bug at scaling, I have scaled a level with 75 75 75, he did nice,
When I scale it with 75 55 75 (flatter) there come holes into existance. Which disappear when you move the camera but others will appear again...

Shadow (I think not bug but it is strange):
http://irrlicht.sourceforge.net/phpBB2/ ... php?t=1439

Shadows aren't working in fullscreen.

This is bad rendering:

What happend here?

[stoffe]

About bug reporting - do you want to have them here in the forum, or in the tracker, or both?

I submitted this (ISceneManager->getMesh crasches on big .obj files on Dev-CPP) the other day, maybe it would be better to keep it in the forum so others could discuss/confirm/deny etc?