[gdal-dev] Possible interlaced 16-bit grayscale PNG read error

[Ray at Daylon]

In GDAL 1.8.0 on little-endian platforms like x64, the PNG driver will 
not byteswap pixels from network order (big endian) to little-endian if 
the PNG is interlaced (or at least if it's 16-bit grayscale interlaced).

Does anyone know if this was corrected in a future version?

In pngdataset.cpp, PNGDataset::LoadScanline() has this code:

     if( bInterlaced )
         return LoadInterlacedChunk( nLine );

The early return causes this downstream byteswapping section:

     if( nBitDepth == 16 )
         GDALSwapWords( row, 2, GetRasterXSize() * GetRasterCount(), 2 );

to be skipped.

LoadInterlacedChunk() as far as I can tell even in 3.9.2. doesn't
do any byteswapping.


Ray Gardener