da53685587
Font class now can change its point-size on the fly - this obviously changes the 'state' of the Font, so subsequent draws will reflect the changes. This will likely not be a problem, as Text input will likely handle this by specifying size and color changes whenever a render operation is needed. Font also now holds the SDL_RWops object so that the font can be closed and re-opened at different sizes. Also added some minor comments in GUI.
2015-03-09 16:40:10 -07:00
b8d6462723
Added basic code for creating and handling GUI events - at the moment just (SDL_USEREVENT + 10) using the SDL_USEREVENT union struct. Maybe define a custom struct that is equal in size to SDL_Event (and has type in the same position) and cast when needed? It'd make it a bit easier/nicer than using data1/data2. Cleaned up GuiElement and added margin functionality.
2015-03-07 22:56:27 -08:00
549b391c4e
A magical commit occurs, wherein a much better GUI layout engine is implemented. Much shorter, much more logical, and much more expandable. Still needs some work as well as some tidying up.
2015-03-07 04:07:29 -08:00
8c2dfb042a
Added Program HashTable to core. Some delete/remove functionality added to various Render* classes. Gui/GuiList somewhat works. GUI is now specified in terms of (0,0) being at the top-left of the screen. Need to recode so GuiElement contains majority of the needed functionality for updating (and that even GuiList(s) contain proper positioning and sizing. Embedding elements in elements will be much easier and more functional if it's abstracted to the base class rather than a specific subclass.
2015-03-06 21:39:07 -08:00
1f0a4cdc53
Updated OSX, iOS, and Android for SDL2_ttf.
2015-03-06 14:13:11 -08:00
1318be4fc0
SDL_ttf now required. Added basic Font class and default TTF font (Amaranth).
2015-03-06 05:53:09 -08:00
482db6b6f1
Updated VS project, works ok.
2015-03-05 20:04:44 -08:00
b3f9d3d0c8
Updated linux build - moved getpwuid/etc. to SysInfo.
2015-03-05 18:45:22 -08:00
d4ea6ce4b0
updated TODO. Gui system is as abominable as ever. Need to rework it so offset_x/offset_y contain the absolute x/y(maybe just rename to that) as created during doUpdate(). Added GuiButton/GuiList elements. The former is a basic button that (at the moment) relies on a textured image - will later have a text texture generated with SDL2_ttf. There now exists a SysInfo class, of which the global instance, sysinfo, is provided. This provides the app, user, and external data directories (still needs to be fully implemented). It also provides screen density DPI, xdpi, and ydpi, at least for Android. RenderObjects may now be shown/hidden.
2015-03-05 17:11:11 -08:00
27624848bd
Added much dodgy code. Fixed bug in Android wherein if AAsset directory did not exist, a close directory call was still made (thereby exploding everything). Added bad mojo for getting user paths for OS X and Linux (USER/Library/Application Support/RtB on OS X and HOME/.RtB/ on Linux). GLSL shaders now reside in their own directories for different versions (e.g., 'data/shaders/100/default_vs.glsl'). Core now supplies getTexture and getMesh functions - these do not yet account for additions to the Asset Caches or similar, as if they successfully load once, they'll forever reference that same object. Added lots of bad stuff to the GUI, including indecision. I want GUI elements to render to a special context so that you can easily have scrolled content (or, perhaps in the future, an embedded browser for embedding HTML content). At the moment it is unused, but a RenderView is created for each GuiElement created. Ideally this RenderView FBO would only be updated whenever a change occurs to the GuiElement, so as to not consume too much. Gui rendering does exist now, although it's prone to change. RenderViews can now be created with a predefined FBO - this sets the STATIC_FBO flag which indicates not to create nor destroy any framebuffers. Additionally, Core now provides a RenderView for the default window fbo, v_fbo. Something is also distinctly screwed up with the RenderCamera/matrix/etc. - I believe an incorrect transformation is happening at some point, as the face culling seems to be reversed.
2015-03-05 03:13:25 -08:00
12b2eed290
Updated VS project - works fine, but probably does not copy over all the needed DLLs during build (zlib, etc.).
2015-03-04 01:12:49 -08:00
a3e703f8fc
Updated Linux build.
2015-03-04 01:05:55 -08:00
3c534a9b98
SDL2_image is now used. OMesh::loadObj now ensures that UVs and normals are created equal to the # of vertices - this is not exactly right, as it simply zeros values that don't exist, but it's better than nothing. Vertex indexing should be implemented using GL_DRAW_ELEMENTS, but for now we'll deal. Default cube now generates proper normals and uvs attached to triangular faces. GLSL vertex shaders now flip the outputted fragment UV (1.0 - uv.t). The Texture class has been created which is responsible for turning some image format data into an OpenGL texture. It handles things similarly to Mesh, in that it first loads the data then has to be 'built'.
2015-03-04 00:59:46 -08:00
12020b1da4
Updated VS
2015-03-03 03:29:09 -08:00
d1121245af
Updated linux build
2015-03-03 03:26:21 -08:00
f05725438d
Implemented some of the GUI subsystem. Events are first passed through the GUI (should limit this to only particular events) before being sent to the top-most state. Later on, the GUI will be capable of locking the sending of particular events, such as motion or mousemove, if an element is currently focused/held. Basic GuiElement added. Each GuiElement may have children, each of which may have children in turn. The Gui will send getHit requests to the top-most/parent GuiElement, after which that GuiElement will send getHit requests down to its children and their children if need be. If a match is found, a pointer to that element is returned, otherwise NULL. More to be added, but hopefully the GUI system is robust enough - it tries to somewhat decouple from the state, in that elements do not have to have their lower level hit detection processed. There may be some oddities in the design that have yet to be discovered due to there being no visual front end (and the creation of RenderObjects for said visuals).
2015-03-03 03:24:07 -08:00
268cbb5a1f
Update VS project - works fine as well.
2015-03-02 20:13:43 -08:00
8b93c18bd5
Updated ios, osx, and android builds - all work fine.
2015-03-02 20:08:00 -08:00
a567769fcf
Core now uses a state stack. States may be pushed on the stack with Core::pushState(State*) and popped off with Core::popState(). Pushing a state will trigger the old state's onCede(), push the state onto the stack, then call the new state's onInit(if DID_INIT is not set) then onRise(). Popping a state will trigger the top state's onCede and the previous state(if it exists)'s onRise. Core::doProcess now has a delta tickrate accumulator that ensures the world is processed at Core::tick_rate. Rendering is not capped to this tickrate, as doRender is called outside of the Event handling and state processing. Finally, functionality that was in Core for testing is now in TestState - it pushed onto the stack during Core::initSystem
2015-03-02 19:58:35 -08:00
685c28891e
GUI Notes that were forgotten.
2015-03-02 11:04:49 -08:00
5b5a1276cf
Added more notes to GUI, with some specifics for icon requirements and how the basic view will be. Very terse at the moment. Added default asset cache locations for iOS: the app's resources/data folder and the app's Documents folder.
2015-03-02 03:25:39 -08:00
ba6d8e21c8
Updated VS project for States and set header includes for src/ and src/states.
2015-03-02 01:54:08 -08:00
6f30b0a662
Updated Linux makefile to include MenuState
2015-03-02 01:45:45 -08:00
0142296949
Added State class and MenuState subclass. Notes added for GUI and STATES in root. iOS, OS X, android, and Linux build environments now use include directories for src/ and src/states. This is to prevent excessive clutter in the src/ directory - although some might say that point was passed long ago.
2015-03-02 01:18:01 -08:00
b1efc3ef74
AssetCache works fine on Windows now. Need to reorganize AssetCache's _WIN32 preprocessor conditionals - it's pretty ugly.
2015-03-01 17:25:19 -08:00
a531be218c
Moved one line of code for no real reason (stemming from foolishness involved with not cleaning the build)
2015-03-01 16:54:21 -08:00
6e0a0b3226
ApkAssetCache now fully works. Unlike AssetCache, it does not build a cache by traversing the directory hierarchy - this is due to limits of how the Android assets archive works -, but rather adds to the cache on get/load requests. This is more inefficient, but it shouldn't be much of a consideration. Fixed preprocessor conditionals around loops to stop vi from exploding. Some minor code cleanup to remove warnings.
2015-03-01 16:46:13 -08:00
977af72999
AssetCaches no longer keep all .CACHE entries. If an entry does not exist on the file system, it is not added to the cache. As a result of this, dud entries are cleared whenever the cache file is written to disk.
2015-03-01 06:07:50 -08:00
abf9b299b5
Added Android Java JNI functions that set 'app' and 'ext' directories that correspond to the application data directory and the external storage app directory respectively. Shaders now use Assets. Assets no longer exist in a 'data/' directory within an AssetCache directory, but rather as their location relative to the root of the AssetCache directory. fileToMem now exists as a separate function to asset_fileToMem, as the old method does not quite fit the new AssetCache scheme. Windows version needs to be fixed to match this update. Android version needs proper Asset management via the ApkAssetCache class - this involves iterating through files in a JNI acceptable manner.
2015-03-01 05:45:36 -08:00
c5f013d693
Updated Linux makefile. added <errno.h>, <sys/stat.h>, and <dirent.h> to AssetCache.cpp. Asset system works on Linux now. I also did A Bad Thing and replaced MAX_PATH with 1024. Also, VS allows for some majorly bad template code to be accepted.
2015-03-01 02:44:28 -08:00
bc831f05b0
AssetManager/AssetCache system now mostly in place. AssetCache(s) create/read from a .CACHE file in the target directory. This file is human-readable and is the index for the cache. Each line contains the file's path, its checksum, and the filesize. This file is updated when the filesize of the file changes. Currently new files are found and added on start - to be added is a refresh function. Also to be added is an update via modification time - this will require a new entry in the cache index file. A single-line version heading should also be added.
...
Data files are now loaded from the Cache - Mesh and Program build from passed buffers now. These are generally read from a loaded Asset. Next up: updating it to work on Linux, Android, iOS, and OS X.
Also added a somewhat poor HashTable template implementation. Works okay, but might not be the best.
2015-03-01 02:17:12 -08:00
05a43ddcdb
Updated iOS, OSX, and Android projects
2015-02-28 14:43:09 -08:00
7cc5ad8df4
Added the AssetCache class skeleton.
2015-02-28 03:08:47 -08:00
aeaeae02b8
Actual update of Asset.cpp
2015-02-27 19:57:06 -08:00
b8ac063da1
Updated Linux makefile. Added stdlib.h to Asset.cpp for free()
2015-02-27 19:56:32 -08:00
e1d908ae3e
Added basic AssetManager class skeleton. This will be the governing class for the loading of files and ensuring data remains the same between clients via checksum comparisons.
2015-02-27 19:49:24 -08:00
0ceb4b828d
Added/updated test code for mouse rotation, zooming, and panning.
2015-02-25 15:56:25 -08:00
8c37edcfe7
Added some test code in Core that allow for pinch zooming and two-finger rotation.
2015-02-25 07:51:10 -08:00
00d0ec1c00
Added cube model for testing. Shader textures now start at index 0 rather than 1. Things work on iOS now. closeSystem deletes gl context and destroys window - needs to free textures, etc. still. Added testing finger and mouse motion control. Meshes now generate uv buffer object and normals buffer object. RenderCamera now supports rendering in orthogonal as well as perspective. Orthogonal rendering relies on 'size' float, which indicates the vertical 3D units shown on screen. This is, for RtB, equal to 1 meter. Perspective rendering uses fov_angle, which is the vertical field of view angle. This is, default, pi / 4. Screen resizing now works fine on Android - recreated buffer used non gles type. RenderView now uses nearest mag/min filters rather than linear. Added handleAppEvents which is intended to deal with state changes on mobile devices, such as app minimizing. At the moment, core.closeSystem is called when RtB enters the background.
2015-02-25 04:05:59 -08:00
561c33f092
Updated VS project. Added a _WIN32 ifdef and typedefs for uint32_t and uint8_t due to Windows not including inttypes.h.
2015-02-24 17:18:40 -08:00
b19f0bb727
Updated Linux build, forgot to set v_flags to opengl/resizable. Core now calls updateProjection for each camera and setView for each camera's RenderView on window resize. This functionality will be moved elsewhere, as we do not necessarily want to resize the view or the camera to the entire window.
2015-02-24 17:11:52 -08:00
a0eeec21b2
Much added. Basic asset access functions using fopen() techniques. Currently the system attempts to load from the CWD on most platforms, but on Android it attempts to load from the assets archive. Once updated, it will likely attempt to load from at least 3 locations before falling back to the built-in application assets (external device, user directory, application data directory). Creation/Loading of RtB is now managed by a Core class. This class is, to some extent, a service locator combined with larger engine functionality. A global instance of the class is provided as 'core'. It is responsible for updating the yet-to-be-added States, the base rendering, and the management of assets and much more.
2015-02-24 16:55:11 -08:00
adcf6cd1b5
Added test chest model and basic model rendering shader. Mesh now builds normals buffer along with uv and vp. Added Quaternion class, currently used for RenderObject's rotation matrix generation - needs to be added to RenderCamera. RenderScene now accomplishes what its name implies - the rendering of a scene. Some logic needs to be moved into Program from RenderScene's doRender function.
2015-02-23 02:49:52 -08:00
07cecebad0
Added a single comment. That is all.
2015-02-22 00:14:48 -08:00
b7e867fbf2
Updates VS project. Replaced snprintf with std::ostringstream in Program::attachTexture. Added RenderCamera functionality for orientation and position. Much more to be added, but likely need to implement the State machine so majority of the program logic is complete.
2015-02-21 05:53:40 -08:00
082828465c
Added Program class - it is responsible for loading and compiling GLSL shaders into a GLSL program. Much to be changed in regards to its functionality, especially Variable/Texture-related access. Rendering to FBO and displaying via a framebuffer shader works.
2015-02-20 19:48:24 -08:00
8b11fd78f3
Re-added VAO code to mesh building, hopefully to no error.
2015-02-13 15:12:26 -08:00
f8758111dc
Updated android, ios, and osx projects. Hopefully nothing is broken.
2015-02-12 07:40:40 -08:00
9ab5b42752
Added common checks for GLEW on Windows, need to fix and add other systems. Added LOG_INFO to logging, ERROR logs show SDL message box, and pointing logHook to a function taking 2 const char* params will call said function on log. Added Mesh type that can load an OBJ mesh and build OpenGL vbo/vao. Can also build from passed vertices, uv, and normal arrays. RenderCamera updated. RenderObject updated with Mesh member. RenderScene updated with basic rendering system set (renderTo(RenderView*)). RenderView now creates framebuffers and should be working. RenderView also relies on a Mesh that should generally be a quad, but is not necessarily (it uses a Mesh).
2015-02-12 05:48:10 -08:00
28d93f69d3
Updated linux Makefile.
2015-02-11 02:31:21 -08:00
kts