Why do we need this?
* Now the user is not warned about temp memory overflow which may introduce hard to debug bugs with memory management.
Describe at least one specific and practical problem this solves for people developing a game
* The buffer overflown and the pointers have been overwritten - undefined behavior, the pc fly into space. This fix fixes what the warning is meant to be. Warn in overflow situation to avoid potential bugs.
Does this add complexity/friction for people making games? If so, how do you justify that?
* There's no complexity in this PR.
Concurrency:
- Refactored spinlock out of OS api (deprecated old procs)
- Concurrency utilites:
void spinlock_init(Spinlock *l);
void spinlock_acquire_or_wait(Spinlock* l);
bool spinlock_acquire_or_wait_timeout(Spinlock* l, f64 timeout_seconds);
void spinlock_release(Spinlock* l);
void mutex_init(Mutex *m);
void mutex_destroy(Mutex *m);
void mutex_acquire_or_wait(Mutex *m);
void mutex_release(Mutex *m);
void binary_semaphore_init(Binary_Semaphore *sem, bool initial_state);
void binary_semaphore_destroy(Binary_Semaphore *sem);
void binary_semaphore_wait(Binary_Semaphore *sem);
void binary_semaphore_signal(Binary_Semaphore *sem);
Macro MEMORY_BARRIER
- Concurrency tests
Docs:
- custom_logger.c example to show how one can make a custom logger and have logs displayed in-game
Utility:
- draw_text_and_measure() which is just an overload of draw_text but it also does a measure and returns it
Misc:
- Added u64 thread_id to global context. This is set in main() for main thread and in thread startup when you dispatch a Thread
- Fixed a bug where plain rects would be drawn with the incorrect color
- Fixed a bug where quads from earlier frames would be drawn
- Improved assert messages to be more clear about what might be happening if they fail
- Added more checks in debug to detect heap corruption and what not
- Fixed a bug where the program would crash because a heap block was perfectly full
- Fixed Y placement of window when changing the window rect
- Fixed window sizing when setting scaled_width or scaled_height
- Updated readme
- Improved assert messages to be more clear about what might be happening if they fail
- Added more checks in debug to detect heap corruption and what not
- Fixed a bug where the program would crash because a heap block was perfectly full
- Fixed Y placement of window when changing the window rect
- Fixed window sizing when setting scaled_width or scaled_height
- Updated readme
- Added asserts to make sure we detect similar bugs faster
- Made simd shippable & configurable
- Some documentation in oogabooga.c on configuration
- Very optimized release build script
- Cleaned some stuff up
- Refactor drawing to keep blocks persistent in memory
- Make string usage easier & more consisten (macros for string vs const char*)
- rdtsc intrinsic (get cycle count)
- Profiling macros (generates a google trace json)
- Write D3D11 debug messages to stdout
- String_Builder & thorough tests for it
- I didn't get rotation going quite right with the xform, but it's 2am goodnight
- Smashed my head and had bad thoughts about microsoft many times
- stuff
- lotsa example code in entry.c
- We just add to a frame of quads which the renderer deals with in gfx_update
- draw_quad, draw_rect, draw_rect_rotated
- Currently just implemented with legacy opengl, so we can't do custom shading
- Added third party code from lodepng so we can load pngs without any other dependencies (no C includes)
- ALLOCATOR_REALLOCATE
- Basic Vector2, Vector3, Vector4 stuff, and some math utilities
- Added <math.h> dependency for now.
- Graphics renderer frontend layer
- Naive get_random() implementation
- Jai-like print procedures
- %s formats 'string' and to format a char* you do %cs
- It detects if %cs pointer is outside of program memory or stack and asserts
- AND same for if a char* is passed to %s
- Print directly writes to stdout without any allocations
- Basic utlity procedures
- Buncha string tests
- Beef up assert to display file & line as well
- Don't define memcpy procedures if compiler has intrinsics for them
- Init memory arena for allocations in initialization time before heap is ready (stack memory)
- os_compare_and_swap
- Spinlock "primitive" (wrapper around a bool using compare_and_swap)
- Switched to using spinlock instead of os mutex in heap for synchronization.
- is_pointer_valid() which checks if address is in program_memory or thread stack
- Dyamically load minimum crt symbols in os backend so we don't need to #include c headers
- Some more os calls:
- write to stdout (with our string type)
- macros for variadic argument lists
- Dynamic library loading
- Some compile options that make sense for build.bat, and a build_release.bat
- oogabooga:
- Big buffer program_memory that always has same base address (goodbye serialisation problems)
- Big buffer grows as needed, crash when out of memory
- Very sexy platform abstraction & basic os call procedures
- Heap allocator around the program memory.
- Jaiyfication (context, context.allocator, push_allocator, temp allocator)
- Some tests to make sure we ooga the right boogas
