Hi Alessandro
I have found the problem, qgis packaging on buster is totally broken: it mixes
3.6.0 and 3.6.1 dependencies
If you try to install python-qgis you get the following packages installed:
ii libqgis-3d3.6.0 1:3.6.0+15buster amd64
QGIS - shared 3d library
ii libqgis-analysis3.6.0 1:3.6.0+15buster amd64
QGIS - shared analysis library
ii libqgis-core3.6.0 1:3.6.0+15buster amd64
QGIS - shared core library
ii libqgis-core3.6.1 1:3.6.1+15buster amd64
QGIS - shared core library
ii libqgis-customwidgets 1:3.6.1+15buster amd64
QGIS custom widgets for Qt Designer
ii libqgis-gui3.6.0 1:3.6.0+15buster amd64
QGIS - shared gui library
ii libqgis-gui3.6.1 1:3.6.1+15buster amd64
QGIS - shared gui library
ii libqgis-native3.6.0 1:3.6.0+15buster amd64
QGIS - shared native gui library
ii libqgis-native3.6.1 1:3.6.1+15buster amd64
QGIS - shared native gui library
ii libqgis-server3.6.0 1:3.6.0+15buster amd64
QGIS - shared server library
ii libqgispython3.6.0 1:3.6.0+15buster amd64
QGIS - shared Python library
ii python-qgis 1:3.6.0+15buster amd64
Python bindings to QGIS
ii python-qgis-common 1:3.6.0+15buster all
Python bindings to QGIS - architecture-independent files
Packages are installed twice in both version 3.6.0 and 3.6.1 !!!!!
This lead to very unpredictible behavior when using qgis in python (simply not
working or crashing randomly)
David,
> Le 26 mars 2019 à 18:38, David Marteau <dmart...@3liz.com> a écrit :
>
>
> Hi Alessandro
>
> No there is nothing at all in the logs.
>
> I have used the same reduced snippet on qgis 3.4.5 (ltr) on Stretch and qgis
> 3.6.0 (release) on buster, using the same datasource:
>
> ```
> import os
> from qgis.core import QgsApplication, QgsVectorLayer
>
> # prevent display not found error
> os.environ['QT_QPA_PLATFORM'] = 'offscreen'
>
> qgis_application = QgsApplication([], False )
> layer = QgsVectorLayer('path/to/data')
>
> layer.isValid() # Return True on 3.4.5/stretch, False on 3.6.0/buster
> whatever the data
> ```
>
> If the release package has not been updated in the previous 24 h, I suspect
> that something has been broken on the buster distro.
>
> Note that I'm able to load shapefile in python with ogr (gdal-python).
>
>
>> Le 26 mars 2019 à 17:47, Alessandro Pasotti <apaso...@gmail.com
>> <mailto:apaso...@gmail.com>> a écrit :
>>
>>
>> On Tue, Mar 26, 2019 at 5:45 PM David Marteau <dmart...@3liz.com
>> <mailto:dmart...@3liz.com>> wrote:
>>
>> The situation seems to be worst: it seems that no layers can be read from
>> python whatever the format.
>>
>>
>>> Le 26 mars 2019 à 17:38, David Marteau <dmart...@3liz.com
>>> <mailto:dmart...@3liz.com>> a écrit :
>>>
>>> Hi devs,
>>>
>>> We use to build docker images of the latest qgis releases and we have a
>>> regression affecting version 3.6.0 official release on buster
>>>
>>> Qgis: 3.6.0 on Debian
>>>
>>> If we read a layer with python:
>>>
>>>> layer = QgsVectorLayer('./mydata.shp')
>>>
>>> Then the layer is invalid (no warning, no errors)
>>>
>>> Doing the same thing on stretch + Qgis 3.4.5 (ltr) with same data work
>>> flawlessly and lead to a valid layer.
>>>
>>> This was not affecting previous build 24 h ago (release end nightly-release)
>>>
>>> The tests were based with shapefiles from testdata in qgis sources.
>>>
>>> We are not sure it that come from Qgis itself or from a change on the
>>> distribution.
>>>
>>> David,
>>>
>>
>>
>> Hi David,
>>
>> anything in the logs?
>>
>> I would check QGIS_PREFIX_PATH env var, and make sure the provider libraries
>> are found and loaded.
>>
>>
>> --
>> Alessandro Pasotti
>> w3: www.itopen.it <http://www.itopen.it/>
_______________________________________________
QGIS-Developer mailing list
QGIS-Developer@lists.osgeo.org
List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
