use float instead of double
This commit is contained in:
parent
f4f91325cc
commit
e81666e657
10 changed files with 114 additions and 115 deletions
|
|
@ -39,3 +39,24 @@ void GBGLDestroy(GBGL gl) {
|
|||
|
||||
free(gl);
|
||||
}
|
||||
|
||||
void GBGLClear(GBGL gl) {
|
||||
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
|
||||
}
|
||||
|
||||
void GBGLIgnoreDepth(GBGL gl) {
|
||||
glDisable(GL_DEPTH_TEST);
|
||||
}
|
||||
|
||||
void GBGLCareDepth(GBGL gl) {
|
||||
glEnable(GL_DEPTH_TEST);
|
||||
}
|
||||
|
||||
void GBGLSetLight(GBGL gl) {
|
||||
GLfloat f[4];
|
||||
int i;
|
||||
|
||||
for(i = 0; i < 4; i++) f[i] = gl->engine->client->light0[i];
|
||||
|
||||
glLightfv(GL_LIGHT0, GL_POSITION, f);
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue