[SPARK 2] Bug Report

Subject: Re: [SPARK 2] Bug Report Wed Dec 14, 2011 4:12 pm

@ shd :

I have corrected the >> template stuff. As i m using visual to program I tend to be a little lazy with the norm (even if i think all modern compilers can well recognize it as the closing of a template now)

For the code not compiling on other compilers than visual, I tested it long ago, so I guess I ll have to make some corrections quick.

Subject: Re: [SPARK 2] Bug Report Wed Dec 21, 2011 8:34 pm

Just to let you know I modified the code so that SPARK 2 compiles under gcc.

I still have some warnings (I think its a compiler bug) and a friend relationship between 2 nested classes I havent managed to make so I had to make some stuff public.

About that last problem, if someone has a solution... I dont manage to get Octree::Cell to be friend of Octree::Array<T> even with forward declaration... Either I m tired, or the solution is tricky syntax or it is a compile bug...

Messages : 4 Date d'inscription : 2011-12-06

Juff wrote:: Just to let you know I modified the code so that SPARK 2 compiles under gcc.

Thanks a lot, you probably saved me quite a lot of work by this Smile

I'll try it myself tomorrow.

[EDIT]
Sorry for late response, I've been busy.
How did you actually compiled it on G++?
On 336 revision, i just tried to make it. Unfortunately part of old errors appear, like:
* missing SPK:IO:Attribute at SPK:IO:Descriptor (so i have to declare 'class Attribute').
* at SPK_Zone.h:70:113 there is type conflict, so i have to change parameter to const (hell, i don't even know if it can be const)
and more things like that.
I'm using gcc 4.6.2. For me, this errors look reasonable but of course i may be wrong since as i said, i don't feel c++ guru at all.

Messages : 4 Date d'inscription : 2012-02-20

I tried to use spark with irrlicht, but it doesn't compile because there is a spelling error.
I've posted a patch in SourceForge to fix the error.

Also, I don't recommend placing binaries in the same repository.
It can be built from source code, and it requires extra space to save entire history of binaries.

Subject: Re: [SPARK 2] Bug Report Mon Feb 20, 2012 6:53 pm

Hi, thanks for your report, i ll check that

Messages : 4 Date d'inscription : 2012-02-20

Hmm, Corrected version only works with Irrlicht 1.8 but 1.7.
pack_texureBlendFunc was renamed to pack_textureBlendFunc in 1.8.

Messages : 7 Date d'inscription : 2012-03-25

When I only add #include "SPARK.h" in my project, I get the following error

Code:: /usr/local/include/spark2/Core/SPK_MemoryTracer.h: In member function ‘void SPK::SPKMemoryTracer::dumpMemory()’:
/usr/local/include/spark2/Core/SPK_MemoryTracer.h:104: error: ‘compareAllocTime’ was not declared in this scope
In file included from /usr/local/include/spark2/SPARK_Core.h:37,
from /usr/local/include/spark2/SPARK.h:26,

EDIT:
Fixed the above error by disabling #define SPK_TRACE_MEMORY in SPK_MemoryTracer.h

But then I got the following when trying to compile the SPKTestIrrlicht.cpp

Code:: SPK_IRR_QuadRenderer.cpp:(.text+0x3d): undefined reference to `SPK::QuadRenderBehavior::QuadRenderBehavior(float, float)'
SPK_IRR_QuadRenderer.cpp:(.text+0x4d): undefined reference to `SPK::Oriented3DRenderBehavior::Oriented3DRenderBehavior()'
/usr/local/lib//libSPARK_IRR.a(SPK_IRR_QuadRenderer.cpp.o): In function `SPK::IRR::IRRQuadRenderer::IRRQuadRenderer(irr::IrrlichtDevice*, float, float)':
SPK_IRR_QuadRenderer.cpp:(.text+0xfb): undefined reference to `SPK::QuadRenderBehavior::QuadRenderBehavior(float, float)'
SPK_IRR_QuadRenderer.cpp:(.text+0x10b): undefined reference to `SPK::Oriented3DRenderBehavior::Oriented3DRenderBehavior()'

Messages : 1 Date d'inscription : 2012-10-25

\sparksdk\src\1.5.5\src\core\spk_model.cpp:200

Code:: if (nbInterpolatedParams > 0)
{
interpolatedParams = new int[nbMutableParams]; //<<=== here
for (size_t i = 0; i < nbInterpolatedParams; ++i)
interpolatedParams[i] = model.interpolatedParams[i];
}

must be:

Code:: interpolatedParams = new int[nbInterpolatedParams];

Subject: Re: [SPARK 2] Bug Report Fri Oct 26, 2012 2:34 pm

Thanks. This is fixed on the svn

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Hi is there a way to reset the particles (lifetime , age , etc) in a group ? The Group::reallocate only works if the newCapacity != currentCapacity

so this doesn't work :

Code:: group->reallocate(group->getCapacity());

There should be an option to somehow "restart" the groups.I have "stop" "play" "pause" buttons in my editor , and i need to reset the systems when i press the "stop" button.

Any ideas ?

PS:
I'm not sure yet if this a bug , but it seems that calling continguously the reallocate() method sometimes (randomly) crashes...

Subject: Re: [SPARK 2] Bug Report Sat Mar 02, 2013 4:35 pm

I don't have the source code of my editor on the computer I'm using now, but I think I implemented it by emptying groups and reseting emitters.

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

OMG !! i just searched a bug all day and after 17 hours of hard bug hunting and coffe drinking and frustrating i found this :

Code:: void Group::removeEmitter(const Ref<Emitter>& emitter)
{
   std::vector<Ref<Emitter> >::iterator it = emitters.begin();
   if (it != emitters.end())
      emitters.erase(it);
   else
      SPK_LOG_WARNING("Group::removeEmitter(Emitter*) - The emitter was not found in the group and cannot be removed");

}

Which is apparently wrong ! No matter which emitter you remove , the very first one will be removed every time...

Subject: Re: [SPARK 2] Bug Report Mon Mar 04, 2013 3:17 pm

Oops... That is a stupid bug. Will fix it as soon as I retrieve my dev machine, next week. Thank you for your report.

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Hi i found another bug : The texture stage states are wrong in the dx9 renderers.The texture alpha mode is wrong in each renderer. To have correct BLEND_MODE_ALPHA mode

These flags : (which are correct for alpha blending)

Code:: DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_COLORARG1, D3DTA_DIFFUSE);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_COLORARG2, D3DTA_TEXTURE);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_ALPHAOP, D3DTOP_SELECTARG1);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_ALPHAARG1, D3DTA_DIFFUSE);

Need to be set like this :

Code:: DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_COLORARG1, D3DTA_TEXTURE);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_COLORARG2, D3DTA_DIFFUSE);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_ALPHAOP, D3DTOP_MODULATE);
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_ALPHAARG1, D3DTA_TEXTURE);
// + 1 flag to set :
DX9Info::getDevice()->SetTextureStageState(0, D3DTSS_ALPHAARG2, D3DTA_DIFFUSE);

Cheers

Subject: Re: [SPARK 2] Bug Report Tue Mar 05, 2013 6:40 am

Hi, thanks for reporting.
If you want not to rely on us for bug fix, i can give you the rights to commit code on svn.

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

OK , i have a little spare time so why not. I'm in !

SourceForge : charlie82

Subject: Re: [SPARK 2] Bug Report Tue Mar 05, 2013 9:27 am

Ok done !

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Ok someone should help , it seems i made a mistake while i commit on the SVN... Sorry
Maybe i just wait for you to commit the changes... i'm not that familiar with SVN... Smile

Subject: Re: [SPARK 2] Bug Report Tue Mar 05, 2013 12:44 pm

I have removed what you ve commited

Subject: Re: [SPARK 2] Bug Report Tue Mar 12, 2013 5:04 pm

The Group::removeEmitter bug is now fixed on svn.

If I have time, one day I will change some of the Group data structures from multiple std::vector to intrusive list (notably emitters).

@Juff: why is activeEmitters mutable ?

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Hi i found a bug in the DX9LineRenderer.In DX9LineRenderer::render() seems that the buffer needs to be locked first (buffer.lock(VERTEX_AND_COLOR_LOCK)) otherwise the engine will crash at :

Code:: buffer.setNextVertex(particle.position());

Regards

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Another bug :
Gravity modifier does not take PARAM_MASS into account in Gravity::modify

OFF :

I've added a new way to scale particle systems and groups adding a protected virtual method to SPKObject :

Code:: vitual void updateScale(float scale);

this method recalculates all the parameters (forces , zone sizes , quads point sprites) in Group::updateParticles

example :

Code:: void Emitter::updateScale(float scale)
{
forceMin = forceMin_const * scale;
forceMax = forceMax_const * scale;
zone->updateScale(scale);
}

it works perfectly , except the line trail renderer... does anyone have an idea how to recalculate the duration and the samples by a given scale factor ?
Thanks

Subject: Re: [SPARK 2] Bug Report Sat Apr 06, 2013 6:29 am

For the gravity, it is not a bug. This is physically correct, the mass doesnt change anything :
Newton's second law :
F = m * a
F = C * (m1 * m2) / d² (gravitation law)
F = g * m (simplified with a constant)
=> a = g

For the scale, yes this is a good feature, tricky however because their may be quite a lot a parameter to change by spk object. The philosophy of the updateScale should be similar to the philosophy of the updateTransform. I wont cache the initial values however. To me the updateScale should change the value and is used at systems loading to adapt the scale of a system to the scale of the given world.

Messages : 81 Date d'inscription : 2011-01-28 Age : 42

Yes you are right about the Gravity! just after i made that post i realized that i was wrong , but forget to edit the post... must be sleepy that day.

However there is a small issue in Graph Interpolators import : The y1 values are not loaded :

Code:: if (validGraph && (attrib = descriptor.getAttributeWithValue("graph values")))
   {
      std::vector<T> values0 = attrib->getValues<T>();
      if (values0.size() == nbEntries)
         for (size_t i = 0; i < nbEntries; ++i) entries[i].y0 = entries[i].y1 = values0[i];
      else
         validGraph = false;
   }

   if (validGraph && (attrib = descriptor.getAttributeWithValue("graph values")))
   {
      std::vector<T> values1 = attrib->getValues<T>();
      if (values1.size() == nbEntries)
         for (size_t i = 0; i < nbEntries; ++i) entries[i].y1 = values1[i];
      else
         validGraph = false;
   }

the line :

Code:: if (validGraph && (attrib = descriptor.getAttributeWithValue("graph values")))

should be :

Code:: if (validGraph && (attrib = descriptor.getAttributeWithValue("graph values 2")))

Subject: Re: [SPARK 2] Bug Report Tue Apr 09, 2013 10:19 am

Thank you for the report. With the controls & attributes system I'm finalizing, the de/serialization will be automatic, so it will be fixed with it.

» Bug Report and CMake
» SPARK and C++
» SPARK & Linux ???
» [SPARK 2] Journal de dev
» Marmelade SDK and Spark