Fix founding difference in pdfium_test on AdobeCMYK_to_sRGB

An optimization to speed up float-to-int rounding caused a different
result for one input value. This tweaks the conversion constant so
that the results are identical across the entire valid range, and
adds a test that checks the part of the range that is most sensitive
to errors.

BUG=pdfium:624

Review-Url: https://codereview.chromium.org/2466203002

2 files changed, 34 insertions, 11 deletions

diff --git a/core/fxcodec/codec/fx_codec_icc.cpp b/core/fxcodec/codec/fx_codec_icc.cpp
index 8e48bfbfea..f0ea6bb43c 100644
--- a/core/fxcodec/codec/fx_codec_icc.cpp
+++ b/core/fxcodec/codec/fx_codec_icc.cpp
@@ -1663,17 +1663,17 @@ void AdobeCMYK_to_sRGB(FX_FLOAT c,
   // Convert to uint8_t with round-to-nearest. Avoid using FXSYS_round because
   // it is incredibly expensive with VC++ (tested on VC++ 2015) because round()
   // is very expensive.
-  // Adding 0.5f and truncating can round the wrong direction in some edge
-  // cases but these do not matter in this context. For instance, the float that
-  // is one ULP (unit in the last place) before 0.5 should round to zero but
-  // this will round it to one. These edge cases are never hit in this function
-  // due to the very limited precision of the input integers.
-  // This method also doesn't handle negative or extremely large numbers, but
-  // those are not needed here.
-  uint8_t c1 = int(c * 255.f + 0.5f);
-  uint8_t m1 = int(m * 255.f + 0.5f);
-  uint8_t y1 = int(y * 255.f + 0.5f);
-  uint8_t k1 = int(k * 255.f + 0.5f);
+  // The 'magic' value of 0.49999997f, the float that precedes 0.5f, was chosen
+  // because it gives identical results to FXSYS_round(). Using the constant
+  // 0.5f gives different results (1 instead of 0) for one value, 0.0019607842.
+  // That value is close to the cusp but zero is the correct answer, and
+  // getting the same answer as before is desirable.
+  // All floats from 0.0 to 1.0 were tested and now give the same results.
+  const float rounding_offset = 0.49999997f;
+  uint8_t c1 = int(c * 255.f + rounding_offset);
+  uint8_t m1 = int(m * 255.f + rounding_offset);
+  uint8_t y1 = int(y * 255.f + rounding_offset);
+  uint8_t k1 = int(k * 255.f + rounding_offset);
 
   ASSERT(c1 == FXSYS_round(c * 255));
   ASSERT(m1 == FXSYS_round(m * 255));
diff --git a/core/fxcodec/codec/fx_codec_jpx_unittest.cpp b/core/fxcodec/codec/fx_codec_jpx_unittest.cpp
index 820a618f5b..3ef14e62c3 100644
--- a/core/fxcodec/codec/fx_codec_jpx_unittest.cpp
+++ b/core/fxcodec/codec/fx_codec_jpx_unittest.cpp
@@ -19,6 +19,29 @@ static unsigned char stream_data[] = {
     0x84, 0x85, 0x86, 0x87,  // Include some hi-bytes, too.
 };
 
+union Float_t {
+  Float_t(float num = 0.0f) : f(num) {}
+
+  int32_t i;
+  FX_FLOAT f;
+};
+
+TEST(fxcodec, CMYK_Rounding) {
+  // Testing all floats from 0.0 to 1.0 takes about 35 seconds in release
+  // builds and much longer in debug builds, so just test the known-dangerous
+  // range.
+  const FX_FLOAT startValue = 0.001f;
+  const FX_FLOAT endValue = 0.003f;
+  FX_FLOAT R = 0.0f, G = 0.0f, B = 0.0f;
+  // Iterate through floats by incrementing the representation, as discussed in
+  // https://randomascii.wordpress.com/2012/01/23/stupid-float-tricks-2/
+  for (Float_t f = startValue; f.f < endValue; f.i++) {
+    AdobeCMYK_to_sRGB(f.f, f.f, f.f, f.f, R, G, B);
+  }
+  // Check various other 'special' numbers.
+  AdobeCMYK_to_sRGB(0.0f, 0.25f, 0.5f, 1.0f, R, G, B);
+}
+
 TEST(fxcodec, DecodeDataNullDecodeData) {
   unsigned char buffer[16];
   DecodeData* ptr = nullptr;