QGIS » QGIS Application

... also, do you have them in multiple favorites?

with "I observe it with both versions." do you mean that this bug is not a regression that was introduced recently on master?

I mean that I can see it in both windows, either the master terminal or the 2.18 terminal.

I just compiled 2.18.12 and it still happens, as said not ALWAYS, but with some fiddling and trying it happens at a certain point...

Slightly other output?

src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [7ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/nl.gpkg' not recognized as a supported file format. 
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [12ms] item favourites:/|home|richard|geodata state changed 1 -> 2
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [2ms] item favourites:/|home|richard|geodata state changed 2 -> 1
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [9ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/nl.gpkg' not recognized as a supported file format. 
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [12ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/prov.gpkg' not recognized as a supported file format. 
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [6ms] item favourites:/|home|richard|geodata state changed 1 -> 2
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [1ms] item favourites:/|home|richard|geodata state changed 2 -> 1
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [5ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/nl.gpkg' not recognized as a supported file format. 
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [11ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/prov.gpkg' not recognized as a supported file format. 
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [11ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/world.gpkg' not recognized as a supported file format. 
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [9ms] item favourites:/|home|richard|geodata state changed 1 -> 2
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [67ms] item favourites:/|home|richard|geodata state changed 2 -> 1
src/providers/gdal/qgsgdaldataitems.cpp: 288: (dataItem) [6ms] [thread:0x55963379e9f0] GDALOpen error # 4 : `/home/richard/geodata/nl.gpkg' not recognized as a supported file format. 
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [8ms] item favourites:/|home|richard|geodata state changed 1 -> 2
src/core/qgsbrowsermodel.cpp: 418: (itemStateChanged) [2ms] item favourites:/|home|richard|geodata state changed 2 -> 1

Richard Duivenvoorde wrote:

I mean that I can see it in both windows, either the master terminal or the 2.18 terminal.

Hi Richard, I think that Alessandro's question is because he needs to understand of this behavior was introduced by the work he is doing on the Broswer and GPKG support in QGIS 3, but it seems that after all this is also happening in QGIS 2.18.* so after all is not a regression, correct?

cheers!

As I said, I've seen this before and I think I know why it happens:

1. when opening a GPKG sqlite creates two temporary files (wal and shm), this triggers a new re-scan of the folder, re-opening the files and re-creating the shm and wal, which triggers a new re-scan of the folder .... infinite loop! (an attempt to fix this was done in bcb48f2bdc8f274cbc68f376489430f45e0b17d1
2. worse: each file is opened twice with GDAL and OGR providers, that could be completely avoided in GDAL 2.x (but this is somewhere down in my TODO list)

There is no perfect solution to this issue but I can try to mitigate the effect.

BTW, I'd be surprised if this was a regression, I believe that the bug was just sitting there and wait for someone to test the gpkg.

No way to add filters, that was my first idea too..

The signal is fired any time something changes in the directory, my fix was to disconnect the signal while the children of the tree are populated (that's when the providers try to open the files - and where GDAL fails) and reconnecting them at the end.

And as an additional issue, if you start creating a file (for example a new shape or gpkg or whatever) the signal is fired and the provider fails because the file is still being written (but this is another issue).

BTW, let me try to reproduce it, because I thought it was fixed and I couldn't reproduce it anymore, maybe having many gpkg in the same folder triggers the error.

I cannot reproduce in master. Can you please confirm that on master it is fixed?

If you run simultaneously 2.8, then you might see the error in master because of the bug in 2.18 that keeps opening the files and creating the temporary shm and wal and trigger the re-scan.

Tested with two master QGIS versions concurrently running, and then it seems fine!

As soon as I start a 2.18, master still starts freaking out.

So: close this one?

No, please leave this open, I need to further investigate if are there other conditions in which it can happen.