ok, I believe I've now a fix in https://github.com/OSGeo/gdal/pull/9700
. Probably a subtle multi-threading issue related to different memory
ordering between M1 and Intel CPUs (or just that it was easier to
trigger on M1)
Le 16/04/2024 à 21:01, Even Rouault via gdal-dev a écrit :
Hi,
I've investigated that today, and I can quite reliably trigger a
similar error with our existing tests on CI, but this is impossible to
diagnose further without direct access to a machine where the error
triggers (when simulating taking the the error code path on Linux, the
fallback code works, whereas it doesn't on OSX Arm64, so there's
something quite subtle happening here). So if someone has a OSX Arm64
and is willing to give me ssh access to it, that would be welcome.
In the meantime, I've submitted
https://github.com/OSGeo/gdal/pull/9685 which is just a workaround to
disable (by default) the threaded RTree creation on that platform. It
can be explicitly enabled by setting OGR_GPKG_ALLOW_THREADED_RTREE to YES
Even
Le 16/04/2024 à 08:13, Rahkonen Jukka via gdal-dev a écrit :
Hi,
It seems that at least two MacOS ARM64 users have faced a problem
with the new, faster rtree creation method so maybe it is worth
having a look. Obviously a big dataset and right hardware is needed
for testing
https://gis.stackexchange.com/questions/479958/how-to-fix-failed-to-prepare-sql-error-when-creating-gpkg-file-from-osm-extrac/479964#479964
-Jukka Rahkonen-
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/gdal-dev
--
http://www.spatialys.com
My software is free, but my time generally not.
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/gdal-dev
--
http://www.spatialys.com
My software is free, but my time generally not.
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/gdal-dev
