Properties are not hidden when using `Property::setHidden`

[JustusBraun]

Generated by Generative AI

No

Operating System:

Linux user-hostname 6.14.0-34-generic #34~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Sep 23 15:35:20 UTC 2 x86_64 x86_64 x86_64 GNU/Linux

ROS version or commit hash:

jazzy, rolling

RMW implementation (if applicable):

No response

RMW Configuration (if applicable):

No response

Client library (if applicable):

No response

'ros2 doctor --report' output

ros2 doctor --report

<COPY OUTPUT HERE>

Steps to reproduce issue

1. Open RViz2
2. Add a PointCloud2 Display
3. The there are lots of properties shown that used to be hidden based on the selected rendering style

Expected behavior

Only the properties related to the currently selected options are shown.

Actual behavior

The properties of all selectable options are shown simultaneously.

Additional information

This bug is a regression and worked fine in commit 4f940dd.
I have tracked the Issue down to the changes made in commit 9fc8054 (PR #1587) . If I remove the nullptr check added in line 122 of property_tree_widget.cpp the expected behavior returns.

rviz/rviz_common/src/rviz_common/properties/property_tree_widget.cpp

Lines 119 to 132 in ffe1ab8

	void PropertyTreeWidget::propertyHiddenChanged(const Property * property)
	{
	if (model_) {
	if (property->parent() != nullptr) {
	const auto & parent_index = model_->parentIndex(property);
	if (parent_index.isValid()) {
	setRowHidden(property->rowNumberInParent(), parent_index, property->getHidden());
	} else {
	printf("Trying to hide property '%s' that is not part of the model.\n",
	qPrintable(property->getName()));
	}
	}
	}
	}

Apparently the parent pointers of the properties are unset.

[ros-discourse]

This issue has been mentioned on Open Robotics Discourse. There might be relevant details there:

https://discourse.openrobotics.org/t/ros-news-for-the-week-of-november-10th-2025/50990/1

[sloretz]

I have tracked the Issue down to the changes made in commit 9fc8054 (PR #1587) . If I remove the nullptr check added in line 122 of property_tree_widget.cpp the expected behavior returns.

Wow thank you for the issue, and for tracking down the cause! 🧇

The nullptr check is interesting. It checks that property->parent() is not null, but I don't see anything using property->parent() in that scope. Can we avoid the crash and get the correct behavior by moving that nullptr check to the scope that actually tries to use property->parent()?

[JustusBraun]

My understanding is that the property->parent() pointers are from QT and are used to link QTObjects for automatic memory management. But the Properties have their on object hierarchy with automatic deletion of children etc:

rviz/rviz_common/src/rviz_common/properties/property.cpp

Lines 67 to 108 in 74c9bc8

	Property::Property(
	const QString & name,
	const QVariant default_value,
	const QString & description,
	Property * parent,
	const char * changed_slot,
	QObject * receiver)
	: value_(default_value),
	model_(0),
	child_indexes_valid_(false),
	parent_(0),
	description_(description),
	hidden_(false),
	is_read_only_(false),
	is_expanded_(false),
	save_(true)
	{
	setName(name);
	if (parent) {
	parent->addChild(this);
	}
	if (receiver == 0) {
	receiver = parent;
	}
	if (receiver && changed_slot) {
	connect(this, SIGNAL(changed()), receiver, changed_slot);
	}
	}
	Property::~Property()
	{
	// Disconnect myself from my parent.
	if (getParent() ) {
	getParent()->takeChild(this);
	}
	// Destroy my children.
	for (int i = children_.size() - 1; i >= 0; i--) {
	Property * child = children_.takeAt(i);
	child->setParent(nullptr);
	delete child;
	}
	}

So the parent() pointers are never set to something. I was also unable to reproduce the crash described in PR #1587 and I cannot find any uses of the parent() method except in add_display_dialog.cpp.