ComputeRasterMinMaxLocation(): fix on all inf/-inf rasters

Author: rouault

autotest/gcore/basic_test.py

@@ -1001,12 +1001,31 @@ def test_ComputeMinMaxLocation():
         and ret.maxY == 18
     )
 
-    ds = gdal.GetDriverByName("MEM").Create("", 1, 1, 1, gdal.GDT_Float64)
+
+@pytest.mark.parametrize(
+    "datatype", [gdal.GDT_Float16, gdal.GDT_Float32, gdal.GDT_Float64]
+)
+def test_ComputeMinMaxLocation_nan(datatype):
+
+    ds = gdal.GetDriverByName("MEM").Create("", 1, 1, 1, datatype)
     ds.GetRasterBand(1).Fill(float("nan"))
     ret = ds.GetRasterBand(1).ComputeMinMaxLocation()
     assert ret is None
 
 
+@pytest.mark.parametrize("value", [float("inf"), float("-inf")])
+@pytest.mark.parametrize(
+    "datatype", [gdal.GDT_Float16, gdal.GDT_Float32, gdal.GDT_Float64]
+)
+def test_ComputeMinMaxLocation_inf(value, datatype):
+
+    ds = gdal.GetDriverByName("MEM").Create("", 1, 1, 1, datatype)
+    ds.GetRasterBand(1).Fill(value)
+    ret = ds.GetRasterBand(1).ComputeMinMaxLocation()
+    assert ret.min == value and ret.max == value
+    assert ret.minX == 0 and ret.minY == 0 and ret.maxX == 0 and ret.maxY == 0
+
+
 def test_create_numpy_types():
     np = pytest.importorskip("numpy")
     gdaltest.importorskip_gdal_array()

gcore/gdalrasterband.cpp

@@ -8745,7 +8745,7 @@ CPLErr GDALRasterBand::ComputeRasterMinMaxLocation(double *pdfMin,
                 const double dfMinValueBlock =
                     GetPixelValue(eDataType, bSignedByte, pData, pos_min,
                                   sNoDataValues, bValid);
-                if (bValid && dfMinValueBlock < dfMin)
+                if (bValid && (dfMinValueBlock < dfMin || nMinX < 0))
                 {
                     dfMin = dfMinValueBlock;
                     nMinX = iXBlock * nBlockXSize + nMinXBlock;
@@ -8761,7 +8761,7 @@ CPLErr GDALRasterBand::ComputeRasterMinMaxLocation(double *pdfMin,
                 const double dfMaxValueBlock =
                     GetPixelValue(eDataType, bSignedByte, pData, pos_max,
                                   sNoDataValues, bValid);
-                if (bValid && dfMaxValueBlock > dfMax)
+                if (bValid && (dfMaxValueBlock > dfMax || nMaxX < 0))
                 {
                     dfMax = dfMaxValueBlock;
                     nMaxX = iXBlock * nBlockXSize + nMaxXBlock;