Question pour le créateur de Purebasic (Fred)

[Lna]

Un coup de pompe cérébral ça peut arriver à tout le monde, tant que celui-ci n'est pas funèbre, il y a encore l'espoir qui nous fera vivre.

Le passage à la version 4 à été effectué sur PC, le plus dur est fait, sur Linux et Mac c'est de l'adaptation. C'est moins difficile mais ça prend aussi du temps.
Déjà que l'on bataille avec les api, alors en plus devoir trouver les équivalents pour Linux et Mac.

Je pense qu'il sera nécessaire de passer par des étapes intermédiaires dans le sens du ce qui est vraiment indispensable pour commencer.
Les premières lacunes de la V 3.94 à la V 4, c'était déjà les types de variables, .c .q .d , xor not, select/case (multi-paramètres) /endselect
Ce serait un bon départ de les avoir avec les fonctions de math qui vont avec.

Pour les changements d'ordre de paramètres, il reste toujours possible de se fabriquer un convertisseur en attendant.
Jusque là, on a toujours réussi à se débrouiller.

Il ne faut pas non plus reprocher à Fred de prendre un peu de distance avec tout ça.
A force de passer autant d'années dessus, ça use. Il faut admettre qu'avec la V 4, que la production à frappée haut.

Je remercie Fred et ses associés pour toutes les bonnes choses qu'ils nous ont apportés et qui continueront à nous apporter.
Le PureBasic, c'est la seule chose de positive après l'an 2000, qui ne nous ait pas déçu.

Et si Fred préparait simplement les élections, la domination du monde pourrait commencer là ?
Votons pour lui quand même, on verra bien après s'il sera clément ou non avec ses fidèles sujets ?

@ peluche les nounours

[poshu]

Funèbre => Funeste?
PC => Windows?

Fallait pas laisser les femmes s’exprimer (ou juste pour voter fred alors.)

[SPH]

Meme si fred arretais PB, la version 4 est suffisement puissante pour tout realiser, alors......

[Thyphoon]

Meme si fred arretais PB, la version 4 est suffisement puissante pour tout realiser, alors......

Il y a du boulot quand même côté 3D... Et on peut toujours faire mieux

[poshu]

si la version nux de pb integre les threads, alors wii, pure basic sera assez complet pour bien des applications. Comme le dis Thyphoon, manque un moteur 3D qui marche.

[Lna]

Funèbre => Funeste?

J'avais bien choisi funèbre, en référence au mot pompe : pompe funèbre
Comme quoi la situation n'est pas si désespérée que ça.

PC => Windows?

C'est juste, avec l'habitude on ne voit que windows comme os principal du PC.
Et vu la difficulté à installer Linux, ça risque encore de durer.

Fallait pas laisser les femmes s’exprimer (ou juste pour voter fred alors.)

A l'état fetal tu devais moins faire le malin, et que même si tu sortais d'un oeuf comme caliméro sans être vilain, et bien les coqs comme les canards n'en pondent pas, coin.

Fallait pas laisser poshu comprendre tout seul non plus, ba voilà. (C'est moins facile lorsque l'on n'a pas le livret avec les corrections, les maîtres d'école paraissent soudainement moins bons, en son absence )


Pour les versions Linux et Mac, il est important de rattraper les lacunes que j'ai citées, dans l'objectif de remettre ces versions à niveau avec les langages basic existants, voir autres.
Et pour le reste, ce n'est que du bonus.

@ peluche

[Progi1984]

Ogre 1.2.4 est sortie le 12/11/2006 :

Code : Tout sélectionner

Changes since v1.2.3:

    * DirectX9:
          o Eliminated setting of texture transform to D3D device if texture transforms are disabled, since D3D texture coordinate processing will ignore it 
          o If "Full Screen" or "Video Mode" configuration option missing when creating an auto D3D window, should throw exception correctly
          o D3D9RenderSystem::setVertexBufferBinding now dealing with gaps in vertex buffer bindings correctly, could manifest iteself as problems if entities created before shadow type picked (generally not recommended anyway)
          o Correctly report support for non-POW2 textures in D3D9, and also detect 'limited' support option which many cards have (no mips etc). Flag was not used in Ogre since POW2 issues are dealt with another way but external apps might be checking it.
          o Fix divide by zero when locking DXT compressed textures
    * OpenGL:
          o Buffers which had temporary blend copies were not forcibly destroying those buffer copies when the original was destroyed, leading to wasted memory that was only recovered when using more general shutdown methods
          o Full screen antialiasing (FSAA) now supported for Win32/GL even if in 16bpp display mode
          o GLSL now correctly populating GL_SAMPLER_2D_RECT and GL_SAMPLER_2D_RECT_SHADOW type uniforms
          o Detect FBO depth/stencil formats a little more robustly
          o Ensure GLSL programs are properly unbound when the next program bound is a different type
          o Check for existence of wglSwapIntervalEXT before trying to use it, fixes crash on very old drivers
          o Hardware pixel buffer locking fix - was only downloading the buffer when locked read-only, now downloads provided the discard lock option isn't used and the texture was not created write-only
          o Hardware pixel buffer unlocking optimisation - don't upload the buffer if the original lock mode was read only
    * Shadows:
          o Should disable fogging explicitly for shadow receiver pass if using additive texture shadows, since fogging state might otherwise leak from a previous modulative mode
          o Fix memory growth issue with stencil shadows when destroying and recreating entities - EntityShadowRenderable was creating vertex buffer binding and vertex declarations itself when they were already being created by VertexData
    * Materials:
          o Material::getNumLodLevels should fall back on first scheme if scheme is not defined for this material
          o Fixed incorrect auto gpu program parameters: inverse_transpose_view_matrix, inverse_transpose_world_matrix and inverse_transpose_worldview_matrix
    * Node: Fix missing calls to Node::Listener if user causes transformation derivation to happen earlier than regular Node::update()
    * Fix reinitialisation of auto windows after Root::shutdown
    * BillboardChain::clearChain should notify index dirty and update bounds.
    * RibbonTrail:
          o setMaxChainElements, setNumberOfChains should reset all trails to initial state if there have tracked node. 
          o setNumberOfChains should not allow changing the number of chains to less than number of tracking nodes. 
          o Overridden clearChain to reset trail to initial state if it has a tracked node.
    * Mesh:
          o Fixed an assert that would be raised when trying to lock an already locked buffer when buildTangentVectors is called on a skeleton animated mesh with no edge list built.
    * Overlay: 
          o Fix copying of nested overlay container elements from template
          o Fix incorrect static_cast<unsigned> instances in BorderPanelOverlayElement which caused problems with some metrics
          o On destruction of a root overlay container, remove from parent overlay to avoid dangling bad reference
    * Quaternion: Fixed shortest path slerp bug, also use lerp rather than just return first quaternion in case two quaternions are very close.
    * OctreeSceneManager: detect mOctree deletion during shutdown if callbacks from nodes being destroyed hit OctreeSceneManager methods, fixes some rare shutdown issues
    * TerrainSceneManager:  Fix terrain reloading problems (setWorldGeometry called more than once)
    * Pass: dirty pass hashes immediately on change rather than on load. If pass is loaded on demand as part of queueing an object for rendering, and the pass had previously been rendered, the pass hash would be dirtied too late for it to be corrected (since the hash is recalculated when the queue is cleared, which has already happened), leading to a hash discrepancy
    * Fix small memory growth issue when destroying and recreating StaticGeometry and ManualObject
    * PixelFormat: PF_UNKNOWN should be marked as inaccessible from PixelUtil::isAccessible
    * Removed MemoryManager::instance() from MemoryManager methods being used inside non-static MemoryManager methods
    * Profiler: Init hierarchicalLvl earlier so that it is valid when used by intermediate code
    * BillboardSet::mBoundingRadius now initialised correctly
    * Normalise result of Vector3::getRotationTo to protect against variations in extreme cases
    * MaterialSerializer: Serialise number of mips and alpha-only option on texture_unit 'texture' attributes
    * AnimationTrack: Ensure the result for zero-weight and trackless animations is the same, caused incorrect blending if mixed
    * HDR compositor scene HDR rendering pass should use "input none", and suppressed warning that attempt to render queue 5 before 96.
    * Entity:
          o shareSkeletonInstanceWith should not delete sharing prepared temp blend buffers
          o Stop sharing skeleton instance during entity destroy and there is only one entity sharing the skeleton
          o Don't re-prepare temp blend buffers when stopping sharing skeleton instance, it's unnecessary in this case
    * Fix Camera::getCameraToViewportRay so that it deals with custom view / projection matrices properly
    * XSI Exporter: Fix a pose animation bug where the start of a pose animation was truncated to fit to a later animation and incorrectly indexed
    * Code::Blocks:
          o allow MinGW to use mem pipe instead of temp files to speedup build process
          o added missing Demo_Ocean project file for CB + MinGW
    * Documentation updates

[GG]

Ogre 1.4.0RC1 est sorti...

[Progi1984]

Ogre 1.4.0 Final est sorti

Code : Tout sélectionner

Changes since RC2:

    * OpenGL:
          o Significant speed improvements for dynamic VBOs - glMapBuffer/glUnmapBuffer is now only used for total buffer replacement or large sub-buffer changes; smaller partial updates use a CPU scratch buffer area and glBufferSubData which can be much faster. Also eliminated some redundant copying when using shadow buffers
          o Fix a problem with bleeding of vertex colour tracking state in GL rendersystem.
          o Use a smaller buffer size when testing FBO formats to reduce overhead
          o Correct GeForce 6200 profile, doesn't support floating-point texture filtering, as stated in NV documents
          o Fixed PBuffer check for whether main context is usable - copy/paste error compared width with height
          o Fixed line ending detection in GLSL uniform parsing, was causing problems with *nix line-endings
          o GLDefaultHardwareVertexBuffer now allocates aligned memory
          o Made GLTextureBuffer destructor safe even if failed to construct render-targets for some reason
          o Don't use FBO method to upload texture from memory if either source or target pixel format is luminance, seems never supported by hardware.
          o Upload texture from memory with FBO should take extra care with source image rowPitch/slicePitch (e.g. nonconsecutive image).
    * D3D9:
          o Fall back to TAM_CLAMP and eliminate D3D debug run-time error if TAM_BORDER not supported by hardware
          o Fixed potential double-release in D3D9HardwarePixelBuffer::blitToMemory if fast GetRenderTargetData fails
    * OSX:
          o Updated the Xcode project by changing any header listed as project, or private, to public.
          o Fixed the header referencing to be compatible with Xcode's framework system. This means that Ogre should be fully ready to be released as a stand alone binary, and it will use the headers in the framework instead of needing an extra copy of the source around
          o Added Crowd and Instancing demos
          o The Tools Xcode projects have been updated against the latest Ogre.framework
          o ExampleApplication relative path fix
    * Linux:
          o Fix automake detection beyond version 1.9
          o Fix GLX crash when building the config dialog image in debug mode
          o Generally improved the GLX driver functionality
          o Allow gcc 4.x builds to take advantage of visibility attribute to reduce plugin size a bit. Works for MinGW and linux gcc >= 4.0.
          o Fix detection of X11 paths for 64bit systems
    * Fix instancing demo crash (should have reduced the mesh count when removing cube.mesh)
    * Fixed and unusual situation with auto-organised buffers where there was unexpected blend weights/indices in vertex data
    * Hardware skinning examples:
          o have GLSL and Cg vertex two weight skinners share the same default parameters.
          o added GLSL vertex skinner two weight shadow caster
          o added unified vertex program for skinner two weight shadow casters
          o used matrix multiply instead of individual dot products in GLSL to improve performance on ATI cards
          o
    * Fix use of UTFString class when using STLport
    * Make sure Rectangle2D::setCorners constructs a valid bounding box
    * Fixed ATI GLSL lighting problem for instancing demo
    * Update deferred shading demo to fix tangent access and explicitly set texture coord set. Also disable skies for MRT since skies don't render to MRT and nVidia D3D driver refuses to write to MRTs for other objects after that's done. Still a few issues with this demo though.
    * Code::Blocks:
          o Project updates including adding missing instancing demo
          o Linux project for Code::Blocks added
    * StringConverter::parseBool accepts '1' as true now
    * Enclose max/min with parentheses in headers to suppress nasty max/min macro substitution if user has a Windows header on their path
    * Make static pass dirty list and graveyard threadsafe when threading enabled
    * Fix bug in GpuProgramParameters::_getFloatConstantPhysicalIndex/_getIntConstantPhysicalIndex when expanding in middle of the buffer for a logical entry
    * Fix CEGui renderer in GL being a bit random with the texture filter modes
    * Font bugfix - don't restart texture coordinates to (0,0) for each code-point range
    * Fixed the fact that Entity::_deinitialise() should have been clearing the internal lists
    * Fixed locale problems with StringConverter::parse* methods - now use std::istringstream instead of atof / atoi so they behave consistently with StringConverter::toString which uses ostringstream.
    * Fixed bug which caused BSP loading not to parse .shader files in subfolders
    * Fix const-correctness on PixelUtil::packColour and Bitwise::intWrite
    * TextAreaOverlayElement - needed to reduce the vertex count twice for CR/LF
    * Fixed Pass max_lights in per-light or per-n-lights iteration, max should apply to the sum of all iterations and not per single iteration
    * 'shadow_scene_depth_range' auto-binding needed same behavior as SceneManager when dealing with light list
    * Consistency issues with writeContentsToFile fixed:
          o D3D now writes only the client area to the screenshot, no title / border (matches GL)
          o Fixed 1-pixel border in Win32 GL screenshots caused by off-by-1 params to glReadPixels
          o Used glReadBuffer(GL_FRONT) in fullscreen mode to ensure that the latest results are read in cases where single-stepping is done
    * Minor memory leaks fixed in OgreXmlConverter
    * Ensure PixelUtil::getBNFExpressionOfPixelFormats is called after all global statics are properly initialised
    * Consistency fixes (export and const correctness) for << operators in various classes
    * Make #ifdef more specific on OgrePrerequisites.h to avoid name clashes
    * TextureUnitState animated textures - allow resetting of animation position
    * Default Quaternion::getYaw / getPitch / getRoll to more intuitive versions ie the atan2 of the projection of the local axes. Can return to the previous version by passing 'false' to the methods.
    * Fix Ogre debug calloc macro has incorrect parameter count
    * Added access to ManualObject sections after they are created
    * Added "hasEnded" method to AnimationState
    * Tidied and removed header entries for some previously removed code
    * LEXIExporter updated to 1.0.3
    * Blender Exporter:
          o added check for empty ipos
          o fixed translation key frame values
          o documentation updates
    * XSI Exporter:
          o Fix pose animations for objects not at the origin