Digitizing: deleting vertexes with backspace can cause spikes when layer and project CSR don't match

[JuhoErvasti]

What is the bug or the crash?

When you're digitizing a geometry on a layer whose CRS is different from the project CRS and use the backspace to cancel added vertexes up to the second to last vertex it causes a spike. Probably better explained by this gif:

It's most apparent with polygon layers but also happens on line layers where there's going to be a duplicate line segment in the geometry.

Happens regardless of the digitizing tool.

Steps to reproduce the issue

1. Download and extract the geopackage here
2. Open the project inside the geopackage
3. Start adding a feature to the polygon layer and create 3 vertexes
4. Press backspace to delete the latest vertex from the rubberband
5. Observe the spike that appears

Versions

QGIS version	3.39.0-Master
QGIS code revision	d85f3c2a28
Libraries
Qt version	5.15.3
Python version	3.10.12
GDAL/OGR version	3.6.4
PROJ version	9.1.1
EPSG Registry database version	v10.076 (2022-08-31)
GEOS version	3.11.1-CAPI-1.17.1
SQLite version	3.37.2
PDAL version	2.5.0
PostgreSQL client version	16.4 (Ubuntu 16.4-1.pgdg22.04+1)
SpatiaLite version	5.0.1
QWT version	6.1.4
QScintilla2 version	2.11.6
OS version	Ubuntu 22.04.4 LTS
This copy of QGIS writes debugging output.
Active Python plugins
processing	2.12.99
MetaSearch	0.3.6
db_manager	0.1.20
grassprovider	2.12.99

Supported QGIS version

• I'm running a supported QGIS version according to the roadmap.

New profile

• I tried with a new QGIS profile

Additional context

[JuhoErvasti]

On more investigation this seems to happen only when the layer and project CRSs are different.

[agiudiceandrea]

The issue didn't occur using QGIS 3.28.15.

[JuhoErvasti]

Okay, interesting. Now that I discovered it's linked to the CRSs I tested again in 3.38.2 and there it does reproduce for me.

Also another issue with the rubberband happens when you backspace-delete all vertexes. The rubberband behaves oddly, although the actual digitized geometry is correct.

This is in 3.38.2 as well and only when the CRSs don't match.:

I've created a test project here, inside a GeoPackage in case someone else wants to see if it replicates:
digitizing_bug_58813.zip

[agiudiceandrea]

Yes, the reported issue occurs using both the current versions QGIS 3.38 and QGIS 3.34 and also the previous versions 3.36 and 3.32. It didn't occur using QGIS 3.28.

[JuhoErvasti]

Oh, my bad then. I mixed up the version numbers.

[nyalldawson]

@JuhoErvasti i can't reproduce using current master and your demo project. is this still an issue?

[agiudiceandrea]

@nyalldawson , both the incorrect reported behaviours occurs even using the latest QGIS 3.39.0-Master b144272 from OSGeo4W on Windows 10.

[nyalldawson]

I can reproduce the weird behaviour when deleting the second last vertex, but not a CPU spike

[JuhoErvasti]

@nyalldawson I was referring to a spike in the polygon geometry here, not a CPU spike. Probably could've made that clearer 😅. I don't have the latest master compiled yet, but it does reproduce on 3.39.0-Master d85f3c2.

[agiudiceandrea]

Me neither observe a CPU spike. Maybe the reporter didn't intend "CPU spike".