[alonfaraj]

What does this PR do?

On a multi band GPKG raster, every band reads band 1 metadata. So after gdal raster edit --stats, all bands show band 1 stats. Opening in update mode saves wrong values to the file.

Repro

gdal raster create --of GPKG --co TILE_FORMAT=PNG --size 1,1 --bbox 0,0,1,1 --band-count 3 --burn 40 --burn 120 --burn 200 rgb.gpkg
gdal raster edit --stats rgb.gpkg
gdal raster info rgb.gpkg | grep -E "^Band|Minimum="

Band 1 Block=256x256 Type=Byte, ColorInterp=Red
  Minimum=40.000, Maximum=40.000, Mean=40.000, StdDev=0.000
Band 2 Block=256x256 Type=Byte, ColorInterp=Green
  Minimum=40.000, Maximum=40.000, Mean=40.000, StdDev=0.000    <--120
Band 3 Block=256x256 Type=Byte, ColorInterp=Blue
  Minimum=40.000, Maximum=40.000, Mean=40.000, StdDev=0.000    <- 200
Band 4 Block=256x256 Type=Byte, ColorInterp=Alpha
  Minimum=40.000, Maximum=40.000, Mean=40.000, StdDev=0.000    <- 255

Reproduced on 3.13.3 and master.

What are related issues/pull requests?

AI tool usage

• AI (Y-a-t-il-un-Copilot-dans-l'avion, Chat-j'ai-pété, Jean-Claude Dusse or something similar) supported my development of this PR. See our policy about AI tool use. Use of AI tools must be indicated.

Tasklist

• Make sure code is correctly formatted (cf pre-commit configuration)
• Add test case(s)
• Add documentation
• Updated Python API documentation (swig/include/python/docs/)
• Review
• Adjust for comments
• All CI builds and checks have passed
• ADD YOUR TASKS HERE

Environment

Provide environment details, if relevant:

• OS: Ubuntu 26.04
• Compiler: gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0

[jratike80]

It would be kind to add a bit more information about what happens, even you do have a good test case.
So, before creating statistics correct statistics are reported because GDAL computes them from the source data

gdalinfo -stats rgb.gpkg
...
Band 1 Block=256x256 Type=Byte, ColorInterp=Red
  Minimum=40.000, Maximum=40.000, Mean=40.000, StdDev=0.000
...
Band 2 Block=256x256 Type=Byte, ColorInterp=Green
  Minimum=120.000, Maximum=120.000, Mean=120.000, StdDev=0.000
...
Band 3 Block=256x256 Type=Byte, ColorInterp=Blue
  Minimum=200.000, Maximum=200.000, Mean=200.000, StdDev=0.000
 ...
Band 4 Block=256x256 Type=Byte, ColorInterp=Alpha
  Minimum=255.000, Maximum=255.000, Mean=255.000, StdDev=0.000

After running gdal raster edit --stats rgb.gpkg the gpkg_metadata table has correct statistics

1	1	dataset	http://gdal.org	text/xml	<GDALMultiDomainMetadata>
  <Metadata domain="BAND_1">
    <MDI key="STATISTICS_MAXIMUM">40</MDI>
    <MDI key="STATISTICS_MEAN">40</MDI>
    <MDI key="STATISTICS_MINIMUM">40</MDI>
    <MDI key="STATISTICS_STDDEV">0</MDI>
    <MDI key="STATISTICS_VALID_PERCENT">100</MDI>
  </Metadata>
  <Metadata domain="BAND_2">
    <MDI key="STATISTICS_MAXIMUM">120</MDI>
    <MDI key="STATISTICS_MEAN">120</MDI>
    <MDI key="STATISTICS_MINIMUM">120</MDI>
    <MDI key="STATISTICS_STDDEV">0</MDI>
    <MDI key="STATISTICS_VALID_PERCENT">100</MDI>
  </Metadata>
  <Metadata domain="BAND_3">
    <MDI key="STATISTICS_MAXIMUM">200</MDI>
    <MDI key="STATISTICS_MEAN">200</MDI>
    <MDI key="STATISTICS_MINIMUM">200</MDI>
    <MDI key="STATISTICS_STDDEV">0</MDI>
    <MDI key="STATISTICS_VALID_PERCENT">100</MDI>
  </Metadata>
  <Metadata domain="BAND_4">
    <MDI key="STATISTICS_MAXIMUM">255</MDI>
    <MDI key="STATISTICS_MEAN">255</MDI>
    <MDI key="STATISTICS_MINIMUM">255</MDI>
    <MDI key="STATISTICS_STDDEV">0</MDI>
    <MDI key="STATISTICS_VALID_PERCENT">100</MDI>
  </Metadata>
</GDALMultiDomainMetadata>

However, now gdalinfo -stats rgb.gpkg´ and gdalinfo rgb.gpkg` tries to use existing statistics without recomputing them, and the statistics are wrongly taken from the band 1 statistics for all layers.

gdalinfo -stats rgb.gpkg
...
Band 1 Block=256x256 Type=Byte, ColorInterp=Red
  Min=40.000 Max=40.000
...
Band 2 Block=256x256 Type=Byte, ColorInterp=Green
  Min=40.000 Max=40.000
 ...
Band 3 Block=256x256 Type=Byte, ColorInterp=Blue
  Min=40.000 Max=40.000
  ...
Band 4 Block=256x256 Type=Byte, ColorInterp=Alpha
  Min=40.000 Max=40.000