Re-work fade operation to be closer to Mixbus; things

below the top region's fades are implicitly faded in
the opposite sense; restore short crossfades option.


git-svn-id: svn://localhost/ardour2/branches/3.0@12022 d708f5d6-7413-0410-9779-e7cbd77b26cf

1 files changed, 22 insertions, 12 deletions

diff --git a/libs/ardour/test/playlist_read_test.cc b/libs/ardour/test/playlist_read_test.cc
index 55c6abe4ad..e016b3adc6 100644
--- a/libs/ardour/test/playlist_read_test.cc
+++ b/libs/ardour/test/playlist_read_test.cc
@@ -74,8 +74,8 @@ PlaylistReadTest::overlappingReadTest ()
 {
 	/* Overlapping read; ar0 and ar1 are both 1024 frames long, ar0 starts at 0,
 	   ar1 starts at 128.  We test a read from 0 to 256, which should consist
-	   of the start of ar0, with its fade in, followed by ar1's fade in (mixed with ar0)
-	   and some more of ar1.
+	   of the start of ar0, with its fade in, followed by ar1's fade in (mixed with ar0
+	   faded out with the inverse gain), and some more of ar1.
 	*/
 
 	boost::shared_ptr<AudioRegion> ar0 = boost::dynamic_pointer_cast<AudioRegion> (_region[0]);
@@ -102,9 +102,11 @@ PlaylistReadTest::overlappingReadTest ()
 	/* ar0's fade in */
 	for (int i = 0; i < 64; ++i) {
 		/* Note: this specific float casting is necessary so that the rounding
-		   is done here the same as it is done in AudioPlaylist.
+		   is done here the same as it is done in AudioPlaylist; the gain factor
+		   must be computed using double precision, with the result then cast
+		   to float.
 		*/
-		CPPUNIT_ASSERT_DOUBLES_EQUAL (float (i * float (i / 63.0)), _buf[i], 1e-16);
+		CPPUNIT_ASSERT_DOUBLES_EQUAL (float (i * float (i / (double) 63)), _buf[i], 1e-16);
 	}
 
 	/* bit of ar0 */
@@ -112,10 +114,12 @@ PlaylistReadTest::overlappingReadTest ()
 		CPPUNIT_ASSERT_EQUAL (i, int (_buf[i]));
 	}
 
-	/* ar1's fade in */
+	/* ar1's fade in with faded-out ar0 */
 	for (int i = 0; i < 64; ++i) {
 		/* Similar carry-on to above with float rounding */
-		CPPUNIT_ASSERT_DOUBLES_EQUAL (i + 128 + float (i * float (i / 63.0)), _buf[i + 128], 1e-4);
+		float const from_ar0 = (128 + i) * float (1 - float (i / (double) 63));
+		float const from_ar1 = i * float (i / (double) 63);
+		CPPUNIT_ASSERT_DOUBLES_EQUAL (from_ar0 + from_ar1, _buf[i + 128], 1e-16);
 	}
 }
 
@@ -143,10 +147,14 @@ PlaylistReadTest::transparentReadTest ()
 
 	_apl->read (_buf, _mbuf, _gbuf, 0, 1024, 0);
 
-	/* ar0 and ar1 fade-ins, mixed */
+	/* ar0 and ar1 fade-ins; ar1 is on top, so its fade in will implicitly
+	   fade in ar0
+	*/
 	for (int i = 0; i < 64; ++i) {
-		float const fade = i / 63.0;
-		CPPUNIT_ASSERT_DOUBLES_EQUAL (float (i * fade) * 2, _buf[i], 1e-16);
+		float const fade = i / (double) 63;
+		float const ar0 = i * fade * (1 - fade);
+		float const ar1 = i * fade;
+		CPPUNIT_ASSERT_DOUBLES_EQUAL (ar0 + ar1, _buf[i], 1e-16);
 	}
 
 	/* ar0 and ar1 bodies, mixed */
@@ -154,10 +162,12 @@ PlaylistReadTest::transparentReadTest ()
 		CPPUNIT_ASSERT_DOUBLES_EQUAL (float (i * 2), _buf[i], 1e-16);
 	}
 
-	/* ar0 and ar1 fade-outs, mixed */
+	/* ar0 and ar1 fade-outs, mixed (with implicit fade-in of ar0) */
 	for (int i = (1024 - 64); i < 1024; ++i) {
-		float const fade = (1023 - i) / 63.0;
-		CPPUNIT_ASSERT_DOUBLES_EQUAL (float (i * fade) * 2, _buf[i], 1e-16);
+		float const fade = (1023 - i) / (double) 63;
+		float const ar0 = i * fade * (1 - fade);
+		float const ar1 = i * fade;
+		CPPUNIT_ASSERT_DOUBLES_EQUAL (ar0 + ar1, _buf[i], 1e-16);
 	}
 }