Address sanitizer detected a memory leak in CmdFeatureMenuBuilder::MenuItem
objects caused by RiuContextMenuLauncher objects that were never deleted.
Problem: The RiuContextMenuLauncher class has two constructors:
1. RiuContextMenuLauncher(widget, CmdFeatureMenuBuilder&) - sets parent ✓
2. RiuContextMenuLauncher(widget, QStringList&) - missing parent ✗
The second constructor was missing the parent initialization, causing all
RiuContextMenuLauncher objects created with QStringList (used in 10+ locations)
to never be deleted, leading to memory leaks of their internal MenuItem vectors.
Root cause: Missing parent-child relationship meant Qt couldn't automatically
clean up these objects when their associated widgets were destroyed.
Fix: Added `: QObject( widget )` to the second constructor to match the first
constructor's behavior. This ensures Qt's automatic parent-child cleanup
handles the RiuContextMenuLauncher lifecycle properly.
This fixes the systemic issue affecting multiple plot types including:
- Correlation plots, Analysis plots, VFP plots, Well allocation plots
- All locations using `new RiuContextMenuLauncher(widget, {commands})`
Address sanitizer detected a memory leak in CmdFeatureMenuBuilder::MenuItem
objects allocated during context menu creation.
Problem: The RiuContextMenuLauncher constructor was creating a temporary
CmdFeatureMenuBuilder object, populating it with menu items, then copying
the entire object to the member variable m_menuBuilder. This caused
unnecessary memory allocation and copying of std::vector<MenuItem> data.
Root cause: Inefficient constructor implementation:
caf::CmdFeatureMenuBuilder menuBuilder; // temp object
for (cmd : commandIds) menuBuilder << cmd; // populate temp
m_menuBuilder = menuBuilder; // copy entire vector (leak source)
Fix: Build the menu directly into the member variable instead of creating
a temporary object and copying it:
for (cmd : commandIds) m_menuBuilder << cmd; // direct construction
This eliminates the unnecessary temporary object creation and copy operation
that was causing the MenuItem vector allocation detected by the sanitizer.
The fix maintains identical functionality while being more efficient.
Address sanitizer detected a memory leak in QProxyStyle objects
created during GUI theme updates.
Problem: The updateGuiTheme() function was called multiple times (e.g., during
initialization) and repeatedly created new PdmUiTreeViewStyle objects for
every PdmUiTreeViewWidget without checking if the widget already had the
correct style set.
Root cause: Each call to updateGuiTheme() would:
1. Iterate through ALL widgets in the application
2. Create a new PdmUiTreeViewStyle for every tree view widget
3. Call setStyle() which should take ownership, but repeated calls
were causing memory leaks
Fix: Added a check to only create and set a new PdmUiTreeViewStyle if the
widget doesn't already have one. This prevents repeatedly creating new
styles for the same widgets during multiple theme update calls.
The fix uses dynamic_cast to check if the current style is already a
PdmUiTreeViewStyle before creating a new one.
* Add drag'n'drop support for ensemble summary vectors and ensembles into histogram plots
* Make sure histogram plot is shown and all update signals are connected when loading a project.
* Support ensemble parameter drag'n'drop from data sources into histogram plots.
* Show ensemble parameters in data sources tree.
* Add copy/paste of histogram curves
* Re-use all existing parameters in a plot when dragging a new ensemble into it
* Adds a bounds check on the node index to prevent accessing invalid memory locations.
This prevents potential crashes caused by out-of-bounds access when retrieving summary variable values.
* Updates warning message for incomplete deltas
* Plot based on a single realization from a file set ensemble is not restored correctly on project load
* Populate result addresses if no addresses are present
* Adds CNV_ (convergence) result name
* Avoid include of RiaResultNames.h in header file
* Removes obsolete code
* Return first non-empty return value from reader
* Simplifies water cut curve visibility check
* Defers the page update until after all plots have been added.
* Add settings to control number of threads
* Refactors ensemble curve handling and visibility
Improves the efficiency of ensemble curve set management by separating realization and statistics curves into distinct collections. This allows for more targeted operations, such as selectively showing/hiding realization curves without affecting statistics.
Also defers data loading of curves until they are actually displayed, optimizing performance.
The visibility updates are modified to optionally skip updating the parent plot, providing more control over replotting behavior and preventing unnecessary updates.
- Introducing a configuration system to automatically detect file patterns for restart and parameter files across ensemble realizations.
- Adding UI elements to control improved summary import settings.
- Modifying summary file reader to use the determined file patterns.
* Add support for reload of ensemble file set
* Fix Replace for ensemble file sets
* Workaround for update of connected plots
* Use summary icon for ensemble data sources
Removes the dedicated `SUMMARY_ENSEMBLE_STATISTICS` category and instead uses a `statisticsType` field within `RifEclipseSummaryAddress` to represent ensemble statistics.
This change simplifies the code and avoids potential confusion by representing statistics as attributes of existing summary types, rather than a separate category. It also improves code robustness.
Create a QVariant using caf::PdmPointer, and send this variant to the feature. The feature converts the QVariant to an ensemble curve set.
Make sure the right-click menu in the summary plot shows the Create from filtered cases independent to the selection in project tree.
Refactors zoom handling in multi-plot scenarios to ensure consistent
axis ranges and proper synchronization between subplots, especially
when linking subplot axes.
Removes direct zoom event handling from subplots and centralizes zoom
logic in the multi-plot container.
This ensures that the 'Zoom All' operation correctly adjusts the
Y-axis ranges based on the configured aggregation method and linked
axes settings.
* Show collapsed range as a single line
* Improves plot annotation label placement
* Adds option to show/hide axis properties
* Adjusts plot info label alignment
* Adds filter by summary value functionality
* Enables ensemble creation from filtered cases
* Use vector name as axis object text, avoid identifier text
* Select the plot clicked at in the Project Tree
* Make sure RFT plots are updated when summary case and ensemble is changed
* Make sure all plots are updated for multiselect and toggle statistics
* Allow paste of plot on top level plot collection
* Allow multiple top level plots as destination for Append operations
* Use same naming for ensemble grid cases as ensemble summary cases
* Create Polygons: Ask for polygon area cutoff when polygon count is high
* Show Create Polygons from Contour Map in polygon collection menu
* Update titles in crossplots
* Clear message panel when project is closed
* Show text message for start and complete of replace ensemble
Interpolate distance from mouse cursor to the line between index above and below closest index. This will make sure selection works when zoomed, and clear highlighting if mouse click is far from curves.
Avoid pixel size for easier config across multiple environments
Set default font size for plots to 10 points
Use Small instead of XSmall for default font in plot
Remove pt postfix to indicate point size
Kristian Bendiksen
JJ
Magne Sjaastad
Santiago Ospina De Los Ríos
jonjenssen
magnesj