GHA last coverage is failing on raster GDAL polygonize cUnit test

[robe]

Given this has been working and we haven't made any changes since this failure, I'm suspecting something upstream changed, perhaps GDAL polygonize again

​https://github.com/postgis/postgis/actions/runs/4677114956/jobs/8284248542

Test: test_gdal_polygonize ...FAILED
    1. cu_gdal.c:170  - CU_ASSERT_DOUBLE_EQUAL(gv[0].val,1.8,FLT_EPSILON)
    2. cu_gdal.c:174  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    3. cu_gdal.c:176  - CU_ASSERT_DOUBLE_EQUAL(gv[1].val,0.0,FLT_EPSILON)
    4. cu_gdal.c:179  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    5. cu_gdal.c:215  - CU_ASSERT_DOUBLE_EQUAL(gv[1].val,0.0,FLT_EPSILON)
    6. cu_gdal.c:218  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    7. cu_gdal.c:270  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    8. cu_gdal.c:272  - CU_ASSERT_DOUBLE_EQUAL(gv[1].val,0.0,FLT_EPSILON)
    9. cu_gdal.c:275  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    10. cu_gdal.c:328  - CU_ASSERT_DOUBLE_EQUAL(gv[0].val,1.8,FLT_EPSILON)
    11. cu_gdal.c:332  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
    12. cu_gdal.c:334  - CU_ASSERT_DOUBLE_EQUAL(gv[1].val,0.0,FLT_EPSILON)
    13. cu_gdal.c:337  - CU_ASSERT_DOUBLE_EQUAL(lwgeom_area(gobserved),lwgeom_area(gexpected),FLT_EPSILON)
Warning 6: PNG driver doesn't support data type Float64. Only eight bit (Byte) and sixteen bit (UInt16) bands supported. Defaulting to Byte

  Test: test_raster_to_gdal ...passed
  Test: test_gdal_to_raster ...passed
  Test: test_gdal_warp ...passed
Suite: raster_geometry
  Test: test_raster_envelope ...passed
  Test: test_raster_envelope_geom ...passed
  Test: test_raster_convex_hull ...passed
  Test: test_raster_surface ...passed
  Test: test_raster_perimeter ...passed
  Test: test_raster_pixel_as_polygon ...passed
  Test: test_raster_get_pixel_bilinear ...xw = 1.5, yw = 0.5, xr = 1.5, yr = 0.5
xw = 1.5, yw = 0.5, value = 20, nodata = 0

Given debbie is not failing, I don't think it's PG16 at fault

[robe]

Docker build has been failing since march (some debian apt issue, which I guess my patch updates and time updates fixed on April 8th). So the issue upstream could have been introduced any time between March 1st and April 11th. At a glance not seeing anything in GDAL commit logs between such a period to cause this issue.

Could also be a change in GEOS I suppose, but debbie, winnie, or berries should have caught those.

[robe]

dbaston suggested this one which seems to be the likely culprit committed March 4th

​https://github.com/OSGeo/gdal/pull/7344

[robe]

The answers are the same it's just the order of polygons and values in the array is different. I'm revising the tests to just sum the values and the areas so to not have to worry about the order in the array.

[Regina Obe <lr@…>]

In 8228878/git:

Change gdal polygonize tests to be sum of areas and values
instead of checking individual elements.
This is to account for GDAL 3.7 results array sorted in different order
from prior versions.
References #5366 for PostGIS 3.4.0

[Regina Obe <lr@…>]

In e58affa/git:

Change gdal polygonize tests to be sum of areas and values
instead of checking individual elements.
This is to account for GDAL 3.7 results array sorted in different order
from prior versions.
References #5366 for PostGIS 3.3.3

[Regina Obe <lr@…>]

In fa7823d/git:

Change gdal polygonize tests to be sum of areas and values
instead of checking individual elements.
This is to account for GDAL 3.7 results array sorted in different order
from prior versions.
References #5366 for PostGIS 3.2.5

[Regina Obe <lr@…>]

In c7b679c3/git:

Change gdal polygonize tests to be sum of areas and values
instead of checking individual elements.
This is to account for GDAL 3.7 results array sorted in different order
from prior versions.
References #5366 for PostGIS 3.1.10

[Regina Obe <lr@…>]

In f7d95150/git:

Make gdal polygonize behave consistently
across all GDAL versions (behavior changed in GDAL 3.7)
Closes #5366 for PostGIS 3.0.10