lib+blender: Fix up coordinate system mismatch

My brain has a hard time keeping up with all the different things going
on, but I got it to the point that the camera seems to work as expected.
I ended up going the easy route, and just adding a toggle to the API to
switch a specified camera to use Blender's native coordinate system, and
still defaults to c-ray coords (y up, +z forward) so existing json
scenes work normally.

bindings/blender_init.py

@@ -22,6 +22,8 @@
 
 import ctypes as ct
 from array import array
+import math
+import mathutils
 
 from bpy.props import IntProperty
 from bpy.props import PointerProperty
@@ -128,8 +130,8 @@ def sync_scene(self, renderer, depsgraph, b_scene):
 			bl_cam = bl_cam_eval.data
 
 			cr_cam = cr_scene.camera_new()
-			mtx = ob_main.matrix_world
-			euler = mtx.to_euler('ZXY')
+			mtx = fix_coord(ob_main.matrix_world)
+			euler = mtx.to_euler('XYZ')
 			loc = mtx.to_translation()
 			cr_cam.set_param(c_ray.cam_param.fov, 80) #todo
 			cr_cam.set_param(c_ray.cam_param.pose_x, loc[0])
@@ -140,6 +142,7 @@ def sync_scene(self, renderer, depsgraph, b_scene):
 			cr_cam.set_param(c_ray.cam_param.pose_yaw, euler[2])
 			cr_cam.set_param(c_ray.cam_param.res_x, self.size_x)
 			cr_cam.set_param(c_ray.cam_param.res_y, self.size_y)
+			cr_cam.set_param(c_ray.cam_param.blender_coord, 1)
 			# TODO: We don't tell blender that we support this, so it's not available yet
 			if bl_cam.dof.use_dof:
 				cr_cam.set_param(c_ray.cam_param.fstops, bl_cam.dof.aperture_fstop)

bindings/c_ray.py

@@ -213,6 +213,7 @@ class cam_param(IntEnum):
 	time = 9
 	res_x = 10
 	res_y = 11
+	blender_coord = 12
 
 class camera:
 	def __init__(self, scene_ptr):

include/c-ray/c-ray.h

@@ -219,6 +219,7 @@ enum cr_camera_param {
 
 	cr_camera_res_x,
 	cr_camera_res_y,
+	cr_camera_blender_coord,
 };
 
 typedef cr_object cr_camera;

src/common/vector.h

@@ -35,8 +35,6 @@ struct intCoord {
 	int x, y;
 };
 
-static const struct vector g_world_up = { 0.0f, 1.0f, 0.0f };
-
 //For defaults
 static inline struct vector vec_zero() {
 	return (struct vector){ 0.0f, 0.0f, 0.0f };

src/lib/api/c-ray.c

@@ -380,6 +380,10 @@ struct camera default_camera = {
 	.fstops = 0.0f,
 	.width = 800,
 	.height = 600,
+	.look_at = { 0.0f, 0.0f, 1.0f },
+	.forward = { 0.0f, 0.0f, 1.0f },
+	.right = { 1.0f, 0.0f, 0.0f },
+	.up = { 0.0f, 1.0f, 0.0f }
 };
 
 cr_camera cr_camera_new(struct cr_scene *ext) {
@@ -442,6 +446,13 @@ bool cr_camera_set_num_pref(struct cr_scene *ext, cr_camera c, enum cr_camera_pa
 			cam->height = num;
 			return true;
 		}
+		case cr_camera_blender_coord: {
+			cam->look_at = (struct vector){0.0f, 0.0f, -1.0f};
+			cam->forward = vec_normalize(cam->look_at);
+			cam->right = (struct vector){1.0f, 0.0f, 0.0f};
+			cam->up = (struct vector){0.0f, -1.0f, 0.0f};
+			return true;
+		}
 	}
 
 	cam_update_pose(cam, &cam->orientation, &cam->position);

src/lib/datatypes/camera.c

@@ -18,16 +18,12 @@ void cam_recompute_optics(struct camera *cam) {
 	cam->aspect_ratio = (float)cam->width / (float)cam->height;
 	cam->sensor_size.x = 2.0f * tanf(deg_to_rad(cam->FOV) / 2.0f);
 	cam->sensor_size.y = cam->sensor_size.x / cam->aspect_ratio;
-	cam->look_at = (struct vector){0.0f, 0.0f, 1.0f};
 	//FIXME: This still assumes a 35mm sensor, instead of using the computed
 	//sensor width value from above. Just preserving this so existing configs
 	//work, but do look into a better way to do this here!
 	const float sensor_width_35mm = 0.036f;
 	cam->focal_length = 0.5f * sensor_width_35mm / deg_to_rad(0.5f * cam->FOV);
 	if (cam->fstops != 0.0f) cam->aperture = 0.5f * (cam->focal_length / cam->fstops);
-	cam->forward = vec_normalize(cam->look_at);
-	cam->right = vec_cross(g_world_up, cam->forward);
-	cam->up = vec_cross(cam->forward, cam->right);
 }
 
 void recomputeComposite(struct camera *cam) {
@@ -36,7 +32,7 @@ void recomputeComposite(struct camera *cam) {
 		struct vector positionAtT = spline_at(cam->path, cam->time);
 		transforms[0] = tform_new_translate(positionAtT.x, positionAtT.y, positionAtT.z);
 	} else {
-		transforms[0] = tform_new_translate(-cam->position.x, cam->position.y, cam->position.z);
+		transforms[0] = tform_new_translate(cam->position.x, cam->position.y, cam->position.z);
 	}
 	transforms[1] = tform_new_rot(cam->orientation.roll, cam->orientation.pitch, cam->orientation.yaw);
 
@@ -77,7 +73,7 @@ struct lightRay cam_get_ray(const struct camera *cam, int x, int y, struct sampl
 	const float jitter_x = triangleDistribution(getDimension(sampler));
 	const float jitter_y = triangleDistribution(getDimension(sampler));
 
-	const struct vector pix_x = vec_scale(vec_negate(cam->right), (cam->sensor_size.x / cam->width));
+	const struct vector pix_x = vec_scale(cam->right, (cam->sensor_size.x / cam->width));
 	const struct vector pix_y = vec_scale(cam->up, (cam->sensor_size.y / cam->height));
 	const struct vector pix_v = vec_add(
 							cam->forward,