Mantis - ISSUES

0000301: CMake Makefiles don't link against stdc++ (OSX)

When generating Unix Makefiles with CMake on OSX (confirmed on 10.5 and 10.6),
the link.txt does not contain "-lstdc++" and therefore, the build fails due to unmatched references.

How to fix?
Add stdc++ to the various LIBs etc. vars; maybe some other way?

0000304: setPolygonMode doesn't work on the iPhone

OpenGL ES doesn't support setPolygonMode(). Objects can only be rendered solid.

0000302: No coherence in CMAKE vars (regarding addition features like ZZLIB)

The cmake scripts set
OGRE_CONFIG_DISABLE_ZIP
OGRE_CONFIG_DISABLE_DDS
...

However, OgreMain/CMakeLIsts.txt checks against OGRE_CONFIG_ENABLE_ZIP

Therefore, if these flags aren't manually set, Ogre is built without ZIP (etc.) support which makes it unable to run even the samples.

0000303: VS SDKs include a reference to CMake in pre-build step

Apparently the pre-build step is included and refers to C:\Program Files\Cmake. The intention was not to have any reference to CMake in the prebuilt SDK unless someone chose to use it, but this probably got missed because this was the default install location.

This needs to be removed / resolved.

0000300: CMake picks up debug dependencies for release build if release versions are not available

http://www.ogre3d.org/forums/viewtopic.php?f=4&t=56168 [^]

Apparently on Windows if the user has only built the debug dependencies, CMake will pick those up for the release build too which causes errors at runtime. The dependencies affected are:

Dependencies\lib\debug\OIS_d.lib
Dependencies\lib\debug\freetype2311_D.lib
Dependencies\lib\debug\FreeImaged.lib
Dependencies\lib\debug\zziplibd.lib
Dependencies\lib\debug\zlibd.lib

0000264: There is probably a bug in the function Camera::setDirection()

I am confused by the code in Camera::setDirection(). I have posted it in the forum, but I got no response. And I still think there is issue in the code. So I file a bug here. Please check if I am wrong.
Below is the post link in the forum.
http://www.ogre3d.org/forums/viewtopic.php?f=2&t=54686&p=371819&hilit=Confused#p371819 [^]

thanks

0000299: Sequence of light changes can cause crash because of light hash==0

http://www.ogre3d.org/forums/viewtopic.php?f=4&t=56018 [^]

void SceneManager::useLightsGpuProgram(const Pass* pass, const LightList* lights)
{
// only call the rendersystem if light list has changed
if (lights->getHash() != mLastLightHashGpuProgram)
{
// Update any automatic gpu params for lights
// Other bits of information will have to be looked up
mAutoParamDataSource->setCurrentLightList(lights);
mGpuParamsDirty |= GPV_LIGHTS;

mLastLightHashGpuProgram = lights->getHash();

}
}
//---------------------------------------------------------------------
void SceneManager::bindGpuProgram(GpuProgram* prog)
{
// need to reset the light hash, and paarams that need resetting, since program params will have been invalidated
mLastLightHashGpuProgram = 0;
mGpuParamsDirty = (uint16)GPV_ALL;
mDestRenderSystem->bindGpuProgram(prog);
}

- lights change,
- useLightsGpuProgram is called with new light list, currently mLastLightHashGpuProgram = 0;
- AutoParamDataSource refreshes its mCurrentLightList
- mLastLightHashGpuProgram is modified to reflect the HASH of Lights
- somewhere later in the code bindGpuProgram(GpuProgram* prog) is called, RESETS mLastLightHashGpuProgram = 0;
- HERE I DELETE THE ONLY LIGHT IN THE LIST in frameRenderingStarted
- useLightsGpuProgram is called with new light list, mLastLightHashGpuProgram = 0;
- AutoParamDataSource DOES NOT REFRESH its mCurrentLightList because, the new light list is empty and has a hash of 0, where mLastLightHashGpuProgram = 0; due to bindGpuProgram() code above, so AutoParamDataSource.mCurrentLightList has invalid contents (reference to deleted light)
- AutoParamDataSource::getLight(0)->getXYZW()

BOOM...CRASH!!!

0000210: glGetString is failing in GLSupport::initialiseExtensions

When starting freeorion or ember (both are using ogre), they are failing with:

main() caught exception(std::exception): basic_string::_S_construct NULL not valid

When debugging it, I came across GLSupport::initialiseExtensions.
I've change the code to the following:

void GLSupport::initialiseExtensions(void)
{
// Set version string
const GLubyte* pcVer = glGetString(GL_VERSION);
fprintf(stderr,"100\n");
printf("GL_RENDERER = %s\n", (char *) glGetString(GL_RENDERER));
printf("GL_VERSION = %s\n", (char *) glGetString(GL_VERSION));
printf("GL_VENDOR = %s\n", (char *) glGetString(GL_VENDOR));
printf("GL_EXTENSIONS = %s\n", (char *) glGetString(GL_EXTENSIONS));
assert(pcVer && "Problems getting GL version string using glGetString");
fprintf(stderr,"101\n");
String tmpStr = (const char*)pcVer;
fprintf(stderr,"102\n");
LogManager::getSingleton().logMessage("GL_VERSION = " + tmpStr);
fprintf(stderr,"103\n");
mVersion = tmpStr.substr(0, tmpStr.find(" "));
fprintf(stderr,"104\n");

The output is now:

olivleh1@kartoffel olivleh1> freeorion
unknown chip id 0x9515, can't guess.
100
GL_RENDERER = (null)
GL_VERSION = (null)
GL_VENDOR = (null)
GL_EXTENSIONS = (null)
101
main() caught exception(std::exception): basic_string::_S_construct NULL not valid
olivleh1@kartoffel olivleh1>

Why isn't glGetString returning anything. It should return "0" on error.
I have no 3D support for my ATI HD3850 card, but It worked with at least ogre 1.2 or 1.4 with Software Rasterizer support. So glGetString not workomg on my system (FreeBSD 7.2, Xorg 7.4, libdrm 2.4.9) is a Bug to me.

With glxgears, glGetString is outputting what it should:

olivleh1@kartoffel olivleh1> glxgears -info
unknown chip id 0x9515, can't guess.
GL_RENDERER = Software Rasterizer
GL_VERSION = 2.1 Mesa 7.4
GL_VENDOR = Mesa Project

0000280: -DFREEIMAGE_LIB prevents dynamic linking to FreeImage

I'm cross-compiling OGRE for win32 with gcc 4.4.2.

The attached patch removes the -DFREEIMAGE_LIB pre-processor flag, which prevents dynamically linking to FreeImage.

I don't know if you intend to support dynamically linking to FreeImage, hence the low severity of the bug.

I'm afraid I'm not yet familiar enough with CMake to provide a patch that makes the definition depend on whether the user wants to link statically or dynamically. If FREEIMAGE_LIB is not required for static linking then it might be best to remove it altogether.

0000289: nmake chokes on precompiled header settings

The precompiled header for MSVC is set via add_definitions; unfortunately CMake adds these parameters also to calls to the linker and resource compiler. This results in nmake based builds to fail when compiling resources. The handling of precompiled headers needs to be specialised for nmake.

0000252: Build error on linux with thread support == 1

GLX is lacking an implementation of GLEW-based GL context sharing.

Attached patch allows compile, but correctness is unclear.

0000240: TextAreaOverlayElement incorrectly loads Font in background thread (OGRE_THREAD_SUPPORT == 2)

If you have overlay scripts with a TextAreaOverlayElement in it, then calling initialiseAllResourceGroups in the background will cause loading of a Font resouce in the background thread. Triggered in function TextAreaOverlayElement::setFontName. Loading a font in the background causes a segmentation fault for the GLX rendersystem if OGRE_THREAD_SUPPORT is set to 2.

This particular code segment in setFontName needs to be moved someplace else.

0000214: StringConverter localisation issues

Currently StringConverter uses the default locale. This means that if a user has set their locale so that decimal points are commas, material parsing can be screwed up.

Material parsing should use a fixed locale of en_GB.

http://www.ogre3d.org/forums/viewtopic.php?f=4&t=50063&p=341435#p341435 [^]

0000172: Background loaded resources - failure & infinite retry

Resources should probably stop trying at some point. What to do?

http://www.ogre3d.org/phpBB2/viewtopic.php?p=228035#228035 [^]