Open palmerj opened 5 years ago
Same issue occurs on Ubuntu with GDAL 2.4.1 installed from source.
@palmerj You just lack patience ;-) Lossless WEBP is super slow. Adding -co NUM_THREADS=ALL_CPUS will make things faster. Eventually it completed in a few minutes. Actually what happens is that the raster is small enough to fit in GDAL block cache in the translation phase. The real compression work is deferred in the FlushCache() step, which unfortunately has no proper progress report mechanism. In e3511844a11959bd81f5956a1e7e2de96900e7ac , I've added a rather poor-man progress report for that step when CPL_DEBUG=ON and for console applications. A proper fix would be much more involved and require revising the virtual FlushCache() method in the API and drivers to add a progress report, and also have GDALTranslate() takes that into account.
If you want more accurate progress report, you can also decrease GDAL_CACHEMAX, like --config GDAL_CACHEMAX 1
... (but might affect performance in some cases of course)
Wow ok thanks WebP encoding is very slow! The progress reporting certainly got me thinking there was something wrong.
I'm re-opening the issue since FlushCache() inherent lack of progress mechanism should ideally be fixed in a cleaner way that I did, and thatcould benefit to non-console applications
Expected behavior and actual behavior.
gdal_translate should finish but just hangs. Putting on debug produces:
I have to control+c to exit program. The output file is corrupted.
Steps to reproduce the problem.
Data is here: https://www.dropbox.com/s/7w1hwm73l676hae/BA31_2623.tif.zip?dl=1
Operating system
MacOSX 10.14.4
GDAL version and provenance
2.4.1 installed from MacOS osgeo-homebrew