Xaphier

Yeah. I was thinking about delaying the change till the texture is loaded, but that would force even bigger changes and I first wanted to see if the current stuff works.

Thanks and I am looking at it. Seems like the best fix for the minimap problem is an image with an explicit alpha channel for the file /textures/compas(like this one: http://img248.images...i/compassp.png/ ) The problem with the map images should be fixed now

Thanks, minimap should be fixed in CVS now.

Thanks alberich, hopefully both problems are now fixed in CVS.

Infos about the driver bug: http://ubuntuforums.org/archive/index.php/t-1134387.html To compile the source code, at the moment cmake 2.6.4 is needed The used libs are: SDL, SDL-Image, cal3d, boost-1.37 (filesystem system thread serialization), zlib, guichan & guichan-sdl and of course opengl

I hope now I fixed the problem (I never had it).

Should be fixed now

That's why I added the last row. idx0 +1 is not always used, only when idx1 is greater than 5. That are the 4 values in the last row. If idx1 is 5 or less, idx0 + 1 is not used as index

Hi! I checked the code again (because I wrote it), but i can't see any problem there. Seems like your gcc makes a mistake. i idx0 idx1 idx0 + 1 0 0 0 - 1 0 3 - 2 0 6 1 3 1 1 - 4 1 4 - 5 1 7 2 6 2 2 - 7 2 5 - 8 3 0 - 9 3 3 - 10 3 6 4 11 4 1 - 12 4 4 - 13 4 7 5 14 5 2 - 15 5 5 -

I have code for loading dds images and I have code that can detect the supported FSAA (Full Scene Anti-Aliasing) modes of a graphic card (using glx). I am working at FSAA detection for windows. For mac I can only write code, but I can't test or debug because I have no mac. If anyone want to have it, just tell me.

The DirectDraw Surface (.dds) file format is from Microsoft. It is used for storing textures (1d, 2d, 3d and cube maps) with and without mipmaps. The compression formats that are used are supported by the graphic cards (DXT1-5 etc.), so the texture can be loaded compressed. Also support for float point textures is there. OpenGL has functions (glCompressedTexImage1D, glCompressedTexImage2D & glCompressedTexImage3) for that. Wikipedia: DDS Microsoft DDS infos

I would say, the md5 in the e3d file is wrong after the change. So either the md5 check is disabled in the client or the right md5 is saved in the file If you want to disable the md5 check, than remove these lines: MD5Open(&md5); size = el_get_size(file) - header_offset; mem = &((Uint8*)el_get_pointer(file))[header_offset]; MD5Digest(&md5, mem, size); MD5Close(&md5, md5_digest); if (memcmp(header.md5, md5_digest, sizeof(MD5_DIGEST)) != 0) { LOG_ERROR("Wrong MD5! File '%s' is corrupt! MD5 [%x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x] MD5 [%x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x]", filename, header.md5[0], header.md5[1], header.md5[2], header.md5[3], header.md5[4], header.md5[5], header.md5[6], header.md5[7], header.md5[8], header.md5[9], header.md5[10], header.md5[11], header.md5[12], header.md5[13], header.md5[14], header.md5[15], md5_digest[0], md5_digest[1], md5_digest[2], md5_digest[3], md5_digest[4], md5_digest[5], md5_digest[6], md5_digest[7], md5_digest[8], md5_digest[9], md5_digest[10], md5_digest[11], md5_digest[12], md5_digest[13], md5_digest[14], md5_digest[15]); return -1; } line 87-97 in io/elc_io.c

Sorry for the delay. Yes, boost is needed. I have no mac so I can't test. But at least here is explained how to compile boost for mac os: http://www.boost.org/doc/libs/1_38_0/more/...x-variants.html

Some features of the new engine: Support for OpenGL 2.0 vertex and fragment shaders Support for GL_ARB_vertex_program and GL_ARB_fragment_program Support for VBO's Support for FBO's Support for full screen post processing filters e.g. gaussian blur Support for loading on demand for resources like shaders, textures, mesh etc. Support for xml files for resource description to prevent hard coding these properties into the source code Support for 3d terrain Support for light managment Support for dds image file format Support for loading compressed textures without decompressing them Support for software decompression of DXT1-5, ATI1 & ATI2 compressed dds images Support for loading mipmaps Support for writing dds images Support for parallel spit shadow mapping Support for static instancing (merging multible objects into one to save draw calls) Optional support for SDL_image Optional support for zip files using zlib Dependencies: libxml2 (easy to replace with some other xml parser) iconv (easy to replace with with other char encoding converter) SDL Boost version 1.35 Scons (as build system, should be easy to replace with some other build system) Optional zlib Optional SDL_image Used build systems: Gentoo 2008.0, 64 bit with gcc-4.3.1 and gcc-4.1.2 Windows XP, 32 bit with Microsoft visual studio express 2008 Test systems: Gentoo 2008.0, 64 bit, AMD Athlon™ 64 X2 Dual Core Processor 3800+ and ATI 1950 XP graphic card using latest ATI closed source drivers Windows XP, 32 bit, Intel® Pentium® M processor 1.86 Ghz and Intel 915 GM graphic card using latest Intel drivers Files: Source code Precompiled windows binary with dlls Data (needed for sdl_map_viewer) Images: I would be happy to hear about any comment. I also will do my best to explain

So, the first alpha version of the map editor is done. Here are some pictures of it:

Vertex programs for actor animation New selection Water shader Bug fixes

Can you try with the current cvs? It should be fixed.

"PARAM matrix"

Can you please try current CVS? It should log more infos in case of an error.

@Florian: Can you please post what is in your info.log. Mainly the lines "Max parameters per program: ...", "Max instructions per program: ..." and "Max bones per mesh: ..." are important.

I wrote a patch for EL to be able to dis-/enable OpenGL extentions based on the settings of an "ini" file. The only drawback is that I used boost spirit for the parser. I mainly worte the patch to be able to disable OpenGL extentions for testing.

I think the implementation of the GL_EXT_gpu_program_parameters extention is buggy in the old driver.

I made a demo (gltest.exe) so you can see some effects of the shader system. You can download every thing needed here http://www.eternal-lands.com/shaders_test.zip. Only the normal mapping, parallax mapping and the lava are eye candies. The program shows some materials with shaders. You need GL_ARB_fragment_shader, GL_ARB_shader_objects, GL_ARB_shading_language_100 and GL_ARB_vertex_shader or OpenGL 2.0 instead of all the extensions. Materials: default: Nothing special, only for testing. stones: Nothing special, only for testing. stones_ext: Texture in YCrCb color space. stones_test: Nothing special, only for testing. door_ext: Texture in YCrCb color space. door: Nothing special, only for testing. door_dxt1: Nothing special, only for testing. m0: Nothing special, only for testing. m0_ext: Nothing special, only for testing. m0_dxt1: Nothing special, only for testing. dxt1: DXT1 compressed texture. Mainly for comparison with the YCoCg. dxt3: DXT3 compressed texture. Mainly for comparison with the YCoCg. dxt5: DXT5 compressed texture. Mainly for comparison with the YCoCg. YCoCg: This is an example of storing the texture in the YCoCg color space with DXT5 compression. Looks better than DXT1 compression but needs one byte per pixel. No special extension needed besides texture compression and shaders. rgb8: Uncompressed rgb8 texture. Mainly for comparison with the YCoCg. rgb16: Nothing special, only for testing. rgb16f: Half float point texture. Here you can see if your graphic card supports linear filtering for this type of texture format. rgb32f: Single float point texture. Here you can see if your graphic card supports linear filtering for this type of texture format. rgba8: Nothing special, only for testing. rgba4: Nothing special, only for testing. l8: Nothing special, only for testing. la8: Nothing special, only for testing. no_normal_mapping: Texture without normal mapping. normal_mapping: Texture with uncompressed tangent space normal mapping. normal_mapping_ext: Texture with compressed tangent space normal mapping. Supported are 3Dc, LATC2 and DXT5. parallax_mapping: Texture with tangent space normal mapping and parallax. gray: A shader that "grays" the texture like the texture adaption with night_shift_textures, but without the need to change the texture. lava: A lava shader. Simple, fast and nice looking. with the parameter --start= and the name of the material you want to start with, e.g. --shart=lava begins with the lava material. Default is material default The keys left and right are used to switch to the previous or next material. The program tests runs 93 unit tests. With --show_progess=yes it shows progress. Don't show anything nice. Source: http://www.eternal-lands.com/shader_system.tar.bz2 You need boost to compile it

Adding the following check fixes things, is that the correct solution Xaphier? Index: elconfig.c =================================================================== RCS file: /cvsroot/elc/elc/elconfig.c,v retrieving revision 1.246 diff -a -u -r1.246 elconfig.c --- elconfig.c 3 Jan 2008 09:42:48 -0000 1.246 +++ elconfig.c 3 Jan 2008 14:22:22 -0000 @@ -722,8 +722,11 @@ } shadow_map_size = size; - destroy_shadow_mapping(); - make_shadow_framebuffer(); + if (have_extension(ext_framebuffer_object)) + { + destroy_shadow_mapping(); + make_shadow_framebuffer(); + } } else { I will fix it. The mistake I made is not checking for FBO befor calling make_shadow_framebuffer. destroy_shadow_mapping should work, but must get called allways, because it do different things if we have FBO support or not.

I am searching for people who want to try the shader system for el and give me feed back . It is not well tested, so it could destroy your house or eat your cat. There could be problems with 32 bit architectures and/or big-endian architectures, with single or multiple core cpus, with different graphic cards and divers etc. The reason why I didn't put it in cvs is the size of the images needed for the tests and because it is not fully integrated into el. At the moment, it use only some parts of el, like el_file, load_gl_extensions and some xml stuff. If there is any interest, I could put it in cvs or make an archive. I also can provide binaries for 64 bit linux and 32 bit linux Some features are: Background loading of images dds file format for 1d, 2d, 3d and cube map textures with and without texture compression, custom mipmap levels xml based configuration of textures, shaders and materials. Support for opengl 2.1, npot textures, 3d textures, cube maps, rect textures, texture compression and more Optional support for c++ tr1 shared_ptr and unordered_map. Now boost multi_index with unique hash index is used. Support for custom mipmap levels (only for dds texture files at the moment) Support for maximum mipmap level Light manager with hashed voxel grid is used. Drawbacks and things that are not done till now: Only glsl shaders are supported, no support for arb vertex or fragment programms Resource managment is not complete, no unloading a the moment Light manager is not complete No float point texture support No support for maximum mipmap level Support for render target textures (like shadow maps) is not ready poor documantation You need the boost library to build the programs, because the boost shared_ptr and multi_index templates are used and the boost test library is used for the unit testing. The make file in the tests dir builds the following programs: tools, a xsd builder for the schema used to validate the resource xml files for the shader system tests, an unit testing program to test some functions and objects of the shader system. It should run on any graphic card, because it don't use opengl. gl_tests, a program using the shader system to some texture loading & drawing and shows some shaders in action. This program needs at least opengl 2.0 and texture compression support to run. Possible features for the future: Support for OpenEXR files Support for jpeg2000 through openjpeg Support for 16 bit pngs through libpng Support for calculation in the xml files for shader parameters using boost spirit Editors for the xml resource files