Randomly discovered during some other bug hunt, only with
AddressSanitizer on. Basically, ASAN detected stack-use-after-return
when playing with spinboxes in the Light Sources preferences when Qt
signal handlers tried to invoke lambda callbacks that referenced
deallocated stack memory.
The main problem is that those lambda functions in the constructor
were captruing by reference. When they were connected to Qt signals
and invoked after constructor completed, they accessed stack vars that
have gone out of scope.
So fixed that by changing lambda captures from `[&]` to explicit captures:
- [this] for lambdas needing only class member access
- [this, updateLight] for lambdas that need both class members and the
`updateLight` lambda.
ArchMaterial's view provider has a setTaskValue helper to set values of widgets inside a task panel. While specific to ArchMaterial, the real place where the helper is called is at BIM_Classification. It appears it's not a general but more like an ad-hoc API for a specific purpose and for a specific object.
In any case, this fix allows widgets that expect a numerical value to have their value set. Before, their value was set as text, unexpectedly.
Before this fix, double-clicking on an ArchMultimaterial object opened its task panel as expected. Unexpectedly, its label was selected and set for edit.
- The tree view's double-click handling expects the view-provider to declare whether it handled the double-click.
- The ViewProvider API's doubleClicked() is documented to return bool (True if handled).
- If a view provider's doubleClicked method returns None/False, the tree falls back to the default Qt behavior (which is beginning inline editing of the item label).
- In ArchMaterial.py the view provider calls self.edit() (or FreeCADGui.ActiveDocument.setEdit(...)) but the method that the tree actually calls (the view provider's doubleClicked) does not return True. Because the method returns None, the tree continues and starts inline label-editing.
In summary: setEdit returning True is fine, but doubleClicked is the method the tree checks; it must return True to suppress the default inline rename.
Python 3 combined the former `urllib`, `urllib2`, `urlparse` Python 2 modules into subpackages of `urllib`. FreeCAD is written in Python 3, thus the `urllib2` import fallback will not work and needs to be removed.
* Sketcher: Fix ctrlA with filter vertex index issue
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
- Enhance _promote_toolbit to handle embedded toolbits that still have a "Custom" shape type.
- If ShapeType is "Custom" but ShapeID is present, attempt to promote to the correct shape class using ShapeID.
- Log promotion for traceability.
- This improves migration of legacy embedded toolbits and ensures proper shape assignment.
* Add Draft workbench to .pre-commit-config
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
As the title says - upon unclone icon is not being removed, but the
cloned item stops mirroring original item which is correct, so this just
aligns the icon removal logic with the correct behavior.
