Such files are formulated in a way that makes it difficult to read them with the generic GTiff driver:
single strip of uncompressed Float32 data type (1.7 GB on the sample file I worked on)
more than 65,535 GCPs in the GeoTIFF TiePoints tag (> 1 million), which make it impossible to read with libgeotiff. Furthermore those tiepoints form actually a geolocation array, and are thus best represented using the GDAL GEOLOCATION formalism
a unknown TIFF tag of code 65000, which happens to contain XML DIMAP metadata of several megabytes. Part of it contains information on the nodata value, offset and scale, and thus must be parsed to report appropriate band properties
All that combined calls for a specialized driver, instead of cluttering even more the generic GTiff driver.
To help for that, I've developed a simple TIFF reader in https://github.com/libertiff/libertiff and included it here.
coverage: 69.427% (+0.007%) from 69.42%
when pulling 822385e73c2739a9d7f6c370b0e90ff2f16883b9 on rouault:snaptiff
into 6aa2e2e2ab89a35d71d4cb87d1fd3f4a8be5fadd on OSGeo:master.
Such files are formulated in a way that makes it difficult to read them with the generic GTiff driver:
All that combined calls for a specialized driver, instead of cluttering even more the generic GTiff driver. To help for that, I've developed a simple TIFF reader in https://github.com/libertiff/libertiff and included it here.