Merge r334996 to M44 branch 2403

https://crrev.com/9c5cf06617232b00842383ff001c49fe058fc6b4 r334996

lut_inverse_interp16: interpolate degenerate zeros in TRC curves

Degenerate TRC have runs of 0's / 1's from the left / right edge of
the TRC curve (a curve we want to invert, to be able to use a color
profile as a monitor profile). The inverse TRC of degenerate curves
is not well-defined in the run of 0's area (infinite slope).

Some monitor calibration software in the field produce 0 runs, thus
degenerate TRC, and QCMS setting the inverse of the run to 0's area
to 0 causes a color cast (burn) effect on web images corrected with
the associated color profile:  issue 492379 .

Fix that by interpolating the (imaginary) inverse in the 0 run area
instead of clamping the inverse to 0. Add more assert conditions to
the curve inversion algorithm and also check for an inverted binary
search range (these two changes from the related Mozilla bug).

TBR=kbr@chromium.org
BUG=458024,492379

Review URL: https://codereview.chromium.org/1209863003.

Cr-Commit-Position: refs/branch-heads/2403@{#400}
Cr-Branched-From: f54b8097a9c45ed4ad308133d49f05325d6c5070-refs/heads/master@{#330231}

2 files changed, 16 insertions, 3 deletions

diff --git a/third_party/qcms/README.chromium b/third_party/qcms/README.chromium
index 3548a0f..01f362b 100644
--- a/third_party/qcms/README.chromium
+++ b/third_party/qcms/README.chromium
@@ -66,6 +66,8 @@ The following changes have been made since qcms was imported:
    - https://code.google.com/p/chromium/issues/detail?id=471749
  - Limit vcgt table to a maximum of 1024 entries
    - https://code.google.com/p/chromium/issues/detail?id=471749
+ - lut_inverse_interp16: interpolate degenerate zeros in TRC curves
+   - https://code.google.com/p/chromium/issues/detail?id=458024
 
 For the Chromium changes, since the import, in a patch format run:
   git diff b8456f38 src
diff --git a/third_party/qcms/src/transform_util.c b/third_party/qcms/src/transform_util.c
index 5eeafa2..aa08df0 100644
--- a/third_party/qcms/src/transform_util.c
+++ b/third_party/qcms/src/transform_util.c
@@ -334,8 +334,6 @@ uint16_fract_t lut_inverse_interp16(uint16_t Value, uint16_t LutTable[], int len
                 if (Value == 0) return 0;
                 // if (Value == 0xFFFF) return 0xFFFF;
                 sample = (length-1) * ((double) Value * (1./65535.));
-                if (LutTable[sample] == 0)
-                    return 0;
                 if (LutTable[sample] == 0xffff)
                     return 0xffff;
 
@@ -353,6 +351,13 @@ uint16_fract_t lut_inverse_interp16(uint16_t Value, uint16_t LutTable[], int len
                     l = 1;
                 if (r > 0x10000)
                     r = 0x10000;
+
+                // If the search range is inverted due to degeneracy,
+                // deem LutTable non-invertible in this search range.
+                // Refer to https://bugzil.la/1132467
+
+                if (r <= l)
+                    return 0;
         }
 
         // Seems not a degenerated case... apply binary search
@@ -376,14 +381,20 @@ uint16_fract_t lut_inverse_interp16(uint16_t Value, uint16_t LutTable[], int len
 
         // Not found, should we interpolate?
 
-
         // Get surrounding nodes
 
+        assert(x >= 1);
+
         val2 = (length-1) * ((double) (x - 1) / 65535.0);
 
         cell0 = (int) floor(val2);
         cell1 = (int) ceil(val2);
 
+        assert(cell0 >= 0);
+        assert(cell1 >= 0);
+        assert(cell0 < length);
+        assert(cell1 < length);
+
         if (cell0 == cell1) return (uint16_fract_t) x;
 
         y0 = LutTable[cell0] ;