You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
your plugin included in QGIS 3.10 for linux seems to be plagued by race conditions when it becomes executed in multiple threads.
the issue doesn't appear in case of the DirectoryWriter, but the MBTilesWriter will hardly work at all and produce corrupt files with lots of missing tiles and other defects...
as a symptom of this particular issue, you should often see error messages like this:
ERROR 5: /tmp/test7.mbtiles: Access window out of range in RasterIO(). Requested (-569088,-368384) of size 256x256 on raster of 512x512.
the strange negative numbers are a side effect of unexpected changed values of the self._first_tile variable resp. colliding zoom factors which are running in parallel and overwrite the expected state one another...
i did try to fix the issue by replacing self._first_tile and self._zoom_ds with dictionaries indexed by the zoom factor to reduce the observable collisions. this had some positive effect, but it doesn't completely fix the issues. :(
unfortunately i do not know, how to fix this kind of thread lock requirements in case of QGIS plugins in a more adequate manner.
the versions of my affected debian linux installation:
your plugin included in QGIS 3.10 for linux seems to be plagued by race conditions when it becomes executed in multiple threads.
the issue doesn't appear in case of the DirectoryWriter, but the MBTilesWriter will hardly work at all and produce corrupt files with lots of missing tiles and other defects...
as a symptom of this particular issue, you should often see error messages like this:
the strange negative numbers are a side effect of unexpected changed values of the
self._first_tile
variable resp. colliding zoom factors which are running in parallel and overwrite the expected state one another...i did try to fix the issue by replacing
self._first_tile
andself._zoom_ds
with dictionaries indexed by the zoom factor to reduce the observable collisions. this had some positive effect, but it doesn't completely fix the issues. :(unfortunately i do not know, how to fix this kind of thread lock requirements in case of QGIS plugins in a more adequate manner.
the versions of my affected debian linux installation:
The text was updated successfully, but these errors were encountered: