[Qgis-psc] CI port to 22.04 - complete!
Tim Sutton
tim at kartoza.com
Thu Aug 11 00:21:43 PDT 2022
Wow awesome work Nyall thanks!
Sent from my iPhone
> On 11 Aug 2022, at 07:49, Nyall Dawson <nyall.dawson at gmail.com> wrote:
>
> Hi PSC,
>
> Just reporting back that I've finished porting the CI setup to 22.04
> (for master at least, the backports are still being worked through the
> queue).
>
> In the end this work involved both a lot of fixes to individual tests
> to reflect changed behaviour in newer GEOS/GDAL/Qt versions, plus a
> fair amount of actual bugs fixed in QGIS which were only revealed on
> the newer libraries.
>
> Here's a quick (non-exhaustive) summary of the actual fixes made
> during this work:
>
> - Fixed a crash in the rule based renderer (this one has been plaguing
> us for a looong time -- it had previously been written off as an issue
> in the python tests only, but during this work the Qt 6 based tests
> started failing and led to the actual underlying cause of the crash!)
> - Fixed an issue in the hub lines processing tool which caused failure
> to save features when running on newer GDAL versions (caused by newer
> GDAL rejecting geometries with nan z/m values)
> - Fix virtual layer provider failing to determine correct field types
> on some sqlite versions
> - Fix spatialite provider incorrectly retrieving features with null
> geometries when we explicitly ask for features within a given extent
> on newer spatialite versions
> - Fix 3d layout map exports are completely blank, and setting the 3d
> layout map to copy settings from a scene completely fails
> - Fix GDAL vrt items inside zip files don't show in browser
> - Fix default qml style for raster layers inside zip archives isn't applied
> - Fix loading gdal rasters inside zip archives which also contain a
> .aux.xml file fails if only the zip archive is added as a layer
>
> I also took the opportunity to do a bunch of cleanups to tests, and
> specifically the way rendering tests were previously relying on the
> orfeo toolbox cdash server to allow retrieval and comparison of
> rendered vs expected images. Now it's done via a more modern approach,
> making the rendered image report directly available as an artifact on
> the pull request. (There's lots more we could do to make this further
> developer-friendly). This removes our reliance on the orfeo toolbox
> cdash server which has been broken for a number of months now.
>
> Thanks again for the opportunity to do this work!
>
> Nyall
> _______________________________________________
> Qgis-psc mailing list
> Qgis-psc at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/qgis-psc
More information about the Qgis-psc
mailing list