From: Matthew Fernandez <matthew.fernandez@gmail.com>
Date: Wed, 21 Apr 2021 01:26:01 +0000 (-0700)
Subject: do pov_begin_graph computation entirely on doubles
X-Git-Tag: 2.47.2~40^2~6
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=f2e44e914edb4dd951883a15fa33be3aa4a56922;p=graphviz

do pov_begin_graph computation entirely on doubles

This code was triggering a number of -Wfloat-conversion compiler warnings, and
there is really no need to truncate values during computation. We can do the
entire thing on doubles with better accuracy, though the final result is
truncated to three decimal places.
---

diff --git a/plugin/core/gvrender_core_pov.c b/plugin/core/gvrender_core_pov.c
index 2f30f3cc3..29b9dcd84 100644
--- a/plugin/core/gvrender_core_pov.c
+++ b/plugin/core/gvrender_core_pov.c
@@ -392,8 +392,6 @@ static void pov_begin_job(GVJ_t * job)
 
 static void pov_begin_graph(GVJ_t * job)
 {
-	float x, y, d, px, py;
-
 	gvprintf(job, "//*** begin_graph %s\n", agnameof(job->obj->u.g));
 #ifdef DEBUG
 	gvprintf(job, "// graph_index = %d, pages = %d, layer = %d/%d\n",
@@ -438,11 +436,11 @@ static void pov_begin_graph(GVJ_t * job)
 #endif
 
 	//setup scene
-	x = job->view.x / 2.0 * job->scale.x;
-	y = job->view.y / 2.0 * job->scale.y;
-	d = 500;
-	px = atanf(x / d) * 180 / M_PI * 2;
-	py = atanf(y / d) * 180 / M_PI * 2;
+	double x = job->view.x / 2.0 * job->scale.x;
+	double y = job->view.y / 2.0 * job->scale.y;
+	double d = 500;
+	double px = atan(x / d) * 180.0 / M_PI * 2.0;
+	double py = atan(y / d) * 180.0 / M_PI * 2.0;
 	gvprintf(job, POV_CAMERA, x, y, -500.0f, x, y, 0.0,
 		 (px > py ? px : py) * 1.2);
 	gvputs(job, POV_SKY_AND_GND);