lp1933001: Preserve phase of decoded FLAC samples

1 files changed, 9 insertions, 3 deletions

diff --git a/src/sources/soundsourceflac.cpp b/src/sources/soundsourceflac.cpp
index a6624a7e82..9c0ba7a5c3 100644
--- a/src/sources/soundsourceflac.cpp
+++ b/src/sources/soundsourceflac.cpp
@@ -403,12 +403,18 @@ namespace {
 // get rid of the garbage in the most significant bits before scaling
 // to the range [-CSAMPLE_PEAK, CSAMPLE_PEAK - epsilon] with
 // epsilon = 1 / 2 ^ bitsPerSample.
-constexpr CSAMPLE kSampleScaleFactor = CSAMPLE_PEAK /
+//
+// We have to negate the nominator to compensate for the negative denominator!
+// Otherwise the phase would be inverted: https://bugs.launchpad.net/mixxx/+bug/1933001
+constexpr CSAMPLE kSampleUnshiftFactor = -CSAMPLE_PEAK /
         (static_cast<FLAC__int32>(1) << std::numeric_limits<FLAC__int32>::digits);
 
 inline CSAMPLE convertDecodedSample(FLAC__int32 decodedSample, int bitsPerSample) {
-    DEBUG_ASSERT(std::numeric_limits<FLAC__int32>::is_signed);
-    return (decodedSample << ((std::numeric_limits<FLAC__int32>::digits + 1) - bitsPerSample)) * kSampleScaleFactor;
+    static_assert(std::numeric_limits<FLAC__int32>::is_signed);
+    static_assert(kSampleUnshiftFactor > CSAMPLE_ZERO, "preserve phase of decoded signal");
+    const auto leftShift = (std::numeric_limits<FLAC__int32>::digits + 1) - bitsPerSample;
+    const auto shiftedSample = decodedSample << leftShift;
+    return shiftedSample * kSampleUnshiftFactor;
 }
 
 } // anonymous namespace