Defects Fixed
This is a history of defects that have been fixed in different releases of Stella Architect, iThink Version 1.7, Stella Professional, and Stella Designer. There is also a list of Known Issues for the current release. The list of features by version is available in the Release Notes for Stella 1.7.1
The following defects were fixed in Version 1.7.1
- The model compare feature was not displaying properly on Macs.
- The allocation input device on the interface would sometimes incorrectly label 0% slices with 100% and place the label in the wrong location.
- Changing a variable to not be a graphical could cause a crash if the variable was used in a graphical input on the interface.
- Module names on multiple lines would cause the assign module input options to display only part of the name during cross level assignment.
- Navigation actions combined with sounds would sometimes fail to play the sounds.
- Font settings for text on arrayed inputs on the interface would also change other parts of appearance.
- Deleting a large number of variables on the Mac would sometimes cause the software to crash.
- Units were not consistently showing up in graph legends on the interface.
- Sometimes the size and position panel settings would show extra text overlapping with the settings themselves.
- Label justification in tables with time running down was incorrect on the model.
- Autofill was sometimes failing to find all appropriate names in the panel.
- On the interface stocks used in input objects were getting the ~ indicating they were not constants even when they were.
- Encapsulation was not connecting modules to some outputs.
- Spell check was slowing down the software.
- Encapsulation was not creating the correct cross level ghost for variables only used in stock initialization.
- Encapsulation would sometimes place variables at negative coordinates making them invisible.
The following defects were fixed in Version 1.7
- Array names were not being properly validated for conflict when opening models.
-
.csv files that only had quoted variables names were not importing correctly.
-
When importing to apply to all with non apply to all data, the values were not correctly stored.
-
Global styles panels for graphs and tables were missing some settings.
- ENDVAL was not consistently retaining the correct value after restoring outputs.
- Input devices with variables that were also used for triggering simulation events, were resetting one time step late.
- Selecting stocks in large models could be slow.
- When a variable was added using the Alt key to repeat the last operation the variable was not highlighted for editing.
- Colors in scatter graphs were not selecting every graph color but instead only the even ones.
- Conveyor leakages with variable leak times were usually not leaking as much in total as the leak fraction would indicate.
- Integer leakages when DT was not 1 would not always result in a true integer amount moving out of the conveyor.
- Values displayed in the equation viewer were incomplete.
- Changing externally saved module filenames could cause spurious error messages.
- The arrowheads on large loop labels were positioned incorrectly.
- Random number seeds that are expressions were not consistently generating random values.
- The find dialog was not displaying all instances of names when A-Z was selected as the order with orer by module selected.
In Earlier Releases
The following defects were fixed in Version 1.6.2
- Triggered events that paused a simulation could case the recorded values on Stocks to miss the DT on which the pause occurred.
- The last page visited navigation button ignored previous and next page button changes as older versions of the software did.
- The dropdown for graph selection on the interface was not working in full screen mode.
- Closing the simulation log during a simulation could cause the software to close.
- There was a problem resolving module names in models saved in iThink or STELLA version 10.0.3 and before.
- Reordering runs in the data manager dialog was not working.
- The export interval on tables was sometimes being incorrectly reset.
- Large intervals in the PULSE function could cause spurious behavior in published models and incorrect error messages on the desktop.
- Some automatically created scales were not rounding correctly.
- Autoscaling was sometimes adjusting scales when it should not have been.
- Clicking on S-Run after adjusting knobs in Stella Live would sometimes run only a single run.
- Double clicking on name only variables would not always correctly open the panel.
-
Modules save as a separate file could give a spurious message about file name conflicts when clicking on the module.
-
Mismatched dimension definitions in modules saved as separate files are now reported.
-
Using Partial Run with "All variables not in sectors" selected did not work with incomplete models.
- Scrubbing on graphs now displays the time value with the same precision as DT.
- Lettered polarity settings made in the model root were not being applied in modules.
- Canceling a run during optimization would not always stop the optimization and might leave remnant controlled variables that could not be restored.
- Importing values into empty equations (using the set parameters option) was not working.
- Outflows from stocks with unit converted inflows were not always correctly constrained.
- Importing from csv files with only two dimension array entries was not always working correctly.
The following defects were fixed in Version 1.6.1
- Stella live runs on published models would sometimes give incorrect results. This has been fixed online for 1.6, 1.5.2 and 1.5.1 as well.
- Controls could be applied to discrete outflows (though they had no effect).
- Some models from the online model editor would not open on the desktop.
- Queues with a number of inflows and outflows could end up near 0 but not completely empty.
- Images did not always scale correctly on high DPI monitors.
- Rich text pastes would sometimes show the incorrect font.
- The sensitivity setup panel would not always commit the changes made to it.
- Name only ghosts would not resize properly.
- Time sliders on interface and model were not in sync when moved manually.
- Number preview in time formatting was showing random value if calculation was not possible.
- Font settings for buttons were being ignored.
- Restore input devices buttons were not working.
- During page load in published model some buttons types were always starting as active.
- Page keys were active in Runtime and locked models.
- Default Graph Series Style Settings were not being applied.
- Font sizes in the Online CLD editor were not saving.
- The pulse function when passed a large interval would generate a spurious divide by zero error.
- The LOG, LOG10 and SQRT builtins were not displaying an error when passed an invalid argument.
- Story pages beginning with numbers and containing periods would prevent interface editing.
- Sensitivity run settings were being ignored after using a Stella live knob.
The following defects were fixed in Version 1.6
- Modules using array offset addressing (using var(2) instead of var[1,2]) were not working when published. This fix has been pushed back for modules published with 1.5.1 and 1.5.2 as well.
- The knob would not work correctly when displaying long labels.
- Long variables names could sometimes get cut off when positioned away from normal name position.
- Connection queries when connecting to modules were not including nested variables taking imported values.
- Divide by zero messages were not opening the equation for the variable triggering the message.
- Variable styles were sometimes overwritten by matching global styles.
- Debugging of interface using a chromium based debugging engine was not working properly.
- Resetting variable sizes to model defaults was not working with variables marked name only.
- Models saved with medium security passwords where the password began with a digit, as well as some other passwords, would fail to recognize the password.
The following defects were fixed in Version 1.5.2
- Undocked results panels could jump in front of the interface window.
- Simulation events were not consistently triggering when used on Stocks or when RK2 or RK4 integration was used.
- Interfaces with sliders could generate extra simulations when used in live mode on the interface or in the model window causing unexpected behavior. (Also fixed for model published with 1.5.1.)
- Controls on a published interface would not always animate in response to changes in their controller behavior. (Also fixed for model published with 1.5.1.)
- Titles on group input devices were being lost after saving and reopening models.
- Importing data from a blank file could cause a crash.
- Generation reporting
- Changing dimensions on the variables when the interface was open could cause the software to crash.
- Preserving data on runs that had been saved could result in runs that could not be deleted or used to display results.
- Renaming a payoff would disconnect the payoff from an optimization and could invalidate the optimization for multi-criteria optimizations.
- Non apply to all equation errors were being erroneously repeated on multiple equations.
- Opening models with missing files related to Simulation Events would not trigger an error report.
The following defects were fixed in Version 1.5.1
- Before a model was run (or after starting a new run) interface output objects would always change when changing inputs. The updating of these now follows the same rules as during a pause depending on the settings of individual output objects.
- The RUNCOUNT builtin would display a number 1 too small before the model was run or after a restore all devices.
- Adjusting a point on a sketchable graph would sometimes reset all the points.
- Creating an import template with graphical functions that had been overridden on the interface could cause erroneous results in the files.
- Creating an import template for an arrayed apply to all constant with some elements overridden was giving the original model equation only.
- The PULSE builtin was not repeating values correctly when called with 2 arguments.
- Using All Combinations in Sensitivity with multiple inputs would result in the assignment of values to the wrong variables.
- The payoff and optimization names defined in the panel would sometimes append a spurious number.
- Sliders and knobs were not always getting the right default bound values for negative numbers or when controlling variables that were not constant.
- Resetting runs from the data manager was missing initial values for stocks which were defined as constants, as well as some flows (see Data Manager) for more discussion of flows.
The following defects were fixed in Version 1.5
- Using ad-hoc settings for sensitivity could cause the software to close unexpectedly.
- Import and export was not working with some new Excel installations on Windows.
- Export to Excel on the Mac could fail if the workbook was closed after the first export.
- Opening a second instance of Stella would sometimes display the crash recovery dialog for the model open in the existing instance.
- Initial thumbnail navigation would sometimes not work without clicking on more than two thumbnails.
- The INVNORM builtin was not working correctly and could cause the software to crash when called with one argument.
- Copying and pasting on high resolution Windows computers was giving poor quality results.
- Importing graphical function values was not resetting previously set controls.
- Clicking on the Causal Lens with a sector selected could cause the software to crash.
- Graphs with NaNs (which should show gaps) were not displaying correctly on the interface.
The following defects were fixed in Version 1.4.3
- Models using the DELAYN, SMTHN and DERIVN builtins with a variable or expression for order and including array expressions would not work.
- Certain flow sequence dependencies could cause spurious divide by zero messages.
- Table appearance on the interface was not always following the panel settings.
- X axis labeling on graphical inputs on the interface was improperly rounded.
The following defects were fixed in Version 1.4.2
- The cycle time values assigned to the second and later conveyor leakages were not being computed correctly. Similarly there were issues computing cycle times on nonnegative stocks with multiple outflows. The errors related to the apportioning of different time stamps when combining material and were generally small.
- The number format on numeric displays would not always update as scale and range settings were changed.
- Modules using theTREND builtin would sometimes return the wrong results after being published.
- Graphicals with constant inputs that were marked invalid would remain invalid even after unchecking graphical.
- Tables were doubling entries when DT was bigger than one.
- After recovering an unsaved model there was no save query when closing the software.
- Sliders would change their horizontal alignment when the restore button became visible.
- Setting values using a button was not working correctly with inputs in modules.
- Setting loop label visibility in stories and model view widgets was not working correctly.
- Pasting interface elements would sometimes move them to a different location.
- Sending objects to the back and front on the interface was not always persisting when the model was closed and reopened.
- The show restore button option on interface was not always remembering its state.
- Sketchable graphs were not correctly displaying discrete graphicals.
- Some international characters were not correctly displayed in the graph legends of published models (this is fixed for modules published with 1.4.1 as well).
- Stella would sometimes crash when clicking on tables on the interface.
- Arrayed expressions were executing outside of array boundaries which would cause incorrect results for the MEAN and STDDEV builtins when used with expressions.
The following defects were fixed in Version 1.4.1
- Cutting and pasting stock and flow structure could sometimes incorrectly connect an existing flow to a pasted stock.
- Exported data going to Excel would sometimes have values missing at intermediate times.
- Autorecovered files were not available after a recovery and second unexpected program termination.
- Annotation popup text styles were not being properly saved with the model.
- Buttons and graphs were losing borders when resizing interfaces.
- Undo after pasting pages into an interface could cause a crash.
- Copying of story pages could cause Stella to crash.
- Insert before on the interface Pages view was inserting the page in the wrong position.
- Error messages relating to incomplete module definitions have been cleaned up and deleting cross level ghost will now also delete automatically created connectors into modules.
- When the properties panel was undocked on the interface it was being brought to the foreground when editing the model.
- Pasting model structure with nested modules and name conflicts was not always properly updating cross level variable definitions.
- Opening a second instance of Stella would sometimes delete auto recovery files from the first instance.
- Cross level ghosts from descendent modules would not be retained in graphs or tables after saving and reopening a model.
- Stacked graphs would sometimes not display properly for long runs.
- Names on the right hand side would sometimes be placed at the wrong location in the model diagram.
The following defects were fixed in Version 1.4
- When published, larger models would sometimes only partially initialize. This could be quite noticeable in numeric displays and spacial map, and is made obvious by the new animation objects. (The fix for was also applied to sims published with 1.3.1).
- The last value in a discrete graphical was not being used (the second last was).
- The line thickness on delay marks was not copying to the clipboard correctly (Windows only).
- Using a graphical input control on a graphical function variables with both active and initial equations resulted in the output from the original graphical function being used in the modified graphical function and giving bad results. (Note also the behavior change in graphical using I (initial) equations mentioned in features section.)
- Equations with initial (I) equations that had an active equation using INIT were not getting the correct value from INIT after the first simulation.
- Starting a model from the command line with export to or import from an Excel file was not correctly connecting to the spreadsheet.
The following defects were fixed in Version 1.3.1
- Using I equations on graphical variables would give wrong results in published models.
- Moving flow positions with the keyboard would not undo properly.
- The space bar was not always being recognized in stories in presentation mode.
- Importing model with conflicting module names could sometimes leave module names duplicated.
- Deleting stories could cause the software to crash.
- Changing discrete stock types could cause incorrect model results or other errors until the model was saved an reopened.
- After closing a model locked to only show the interface menus were left grayed.
- Turning on cycle time could leave to incorrectly constrained outflows.
- Deleting a dimension from a variable could cause a crash.
- Graphical frames were sometimes drawn outside the display area in published models.
- Cross level ghosts were not always appearing correctly in the causal lens.
- When using storytelling in presentation or experiment mode simulations would sometimes pause arbitrarily.
- Graphs on published models were not always displayed correctly in IE and Edge (this fix was applied to models published with 1.3 as well).
- Using a biflow on a non negative stock could incorrectly constrain the outflows of the stock.
- Flow ordering could sometimes be different in published models leading to different results.
- Text boxes with only spaces (sometimes used for borders) were reopening without the spaces and the prompt to type in text was therefore visible.
- Model view animations were not always responding to changes in the state.
- Deleting variables after deleting the active run in the data manager could cause the software to crash.
The following defects were fixed in Version 1.3
- Using both the INIT builtin and intial equations to break circularities could sometimes lead to spurious behavior.
- The ENDVAL builtin, when used with an explicit initial, would often return the second from last value in the previous simulation.
- Circularities involving Queue outflows were not always being correctly identified.
- Undo was not always maintaining discrete stock and flow attributes, such as capacities and leak zones, when variables were deleted or had their types changed.
- The graphical panel could sometimes get in a state where it would flicker between having, and not having, a vertical scroll bar.
- Parenthesis matching was not working for long equations.
- Labels in the equation panel were sometimes being truncated.
- The x-axis value displayed when scrubbing graphs could be misleading if the runs had different start or end points.
- Fixed scale graphs on the model would sometimes draw graph lines outside of the graph proper.
- Editing units could sometimes generate spurious unit errors.
- Changing a conveyor to a queue and back could sometimes cause spurious model results.
The following defects were fixed in Version 1.2.2
- The highlighting of name only converters for storytelling was not working correctly.
- Changing connections in a model that had implicit circularities requiring marking of stocks would sometimes cause spurious divide by 0 messages.
- Importing a mixture of overwriting and non-overwriting sources could cause the software to freeze.
- Connecting would sometimes fail because of incorrectly identified circularities.
- Closing models with separate initial equations would sometimes cause an error.
- Simulations were being stopped when shifting between model and interface.
- Data manager changes were not always correctly reflected on comparative graphs and tables.
- Exported variable markings were being incorrectly applied when connecting to a module.
- Graph labels were misleading when graph ranges were very small.
- There was excessive memory use on export.
The following defects were fixed in Version 1.2.1
- The DELAY and DERIVN builtins were not reporting correct values when using Runga Kutta integration (RK2 or RK4).
- Importing of multiple CSV files could cause extra values to be incorrectly added to graphicals and display spurious error messages.
- Undoing and redoing structure additions more than once could cause the software to crash.
- Very small graph scales could cause the graph to display incorrectly when values were out of bounds.
- Setting user parameters from the data manager or during import could cause leakage fractions to be set incorrectly.
- Sector positioning made it difficult to have the sectors directly touch so that all variables could be included in a sector.
- Exported CSV files were not correctly escaping quotes on variable names with embedded quotes.
- Negative polarity symbols where not being displayed correctly on the interface.
- Single clicking on knobs in explore mode when save all variables was not set would net reset the constant value.
- Changing the passthru nature of a stock (by creating or removing an implicit circularity) could cause the software to crash.
- Setting the report interval on a table to 0 could cause a crash.
- Extraneous highlighting was occurring in the equation panel with variable names requiring quotes.
- Selection of connector handles was difficult when many connectors were coming into a flow or converter.
The following defects were fixed in Version 1.2
- Reordering array elements was not always correctly reflected in Non Apply to All equations.
- Name positions set using the shift key (so that all 360 degrees of location can be used) were not all available.
- Some initial circularities involving conveyor leakages were not being detected.
- Thumbnails for interface pages would not always update when changes were made to pages.
- Architect and Designer would slow down after navigating through a large number of pages.
- Variable names in the simulation log did not always have module prefixes.
- Fixing import links sometimes required saving and reopening a model.
- Dimension mismatch corrections could sometimes leave variables marked invalid till a save and reopen.
- When both the interface and model window were open the model window would appear on top after exiting presentation mode.
- On some Windows computers connecting to our licensing server would unnecessarily require the web solution.
- Adding points to a graphical by typing in a new number of points would sometimes cause a y value to repeat, replacing many other values.
- On the interface, sliders for variables that are not constant could display incorrect values on hover.
- Importing modules that replaced an existing module could leave remnant cross level connections from the old module content.
- The equation viewer was always displaying the entire model instead of the module it was opened in.
- Array definitions added to a model on paste were not always enough to fully support the added equations.
- Units synonyms were not being added on paste.
- Publishing of models that use the INIT function was not always working correctly.
- Divide by 0 errors were not being reported on published or presentation mode models.
The following defects were fixed in Version 1.1.2
- After publishing a model macros could get stuck at their initial values causing erroneous results during simulations made from within Stella Architect or Stella Designer.
- Models using INIT builtin nested inside of other functions could produce incorrect results when published.
- Published models would sometimes preserve the incorrect number of runs for comparative graphs.
- Models using equations to compute array offsets were producing incorrect results when published.
- Simulation events could trigger incorrectly when entering Stella Live in presentation mode or on a published model.
- Stella Live was being disabled for models taking 1.2 seconds instead of the documented 10 seconds to simulate.
- Switching to presentation mode would sometimes clear graph results inappropriately.
- Models generating a divide by 0 could keep running beyond the normal end time when in presentation mode.
- Lab pack activations on the Mac could fail for some users.
The following defects were fixed in Version 1.1.1
- Comparative graphs with multiple pages would sometimes lose or reorder runs when moving between different pages.
- Typing in a value for a numeric or group input would show a rounded value instead of the value entered.
- The number of gridlines used by the Graphical Input on the interface was not always matching the number of points in the graphical function.
- Group input devices on the interface were not adjusting the row height in response to font changes.
- The CTMEAN, CTMIN and CTMAX functions were returning incorrect cycle time statistics.
- Thumbnail scrolling was uneven.
- Some version 10 model graphs would lose their multiscale attribute.
- Exiting Presentation Mode on the interface did not always work correctly.
- Better support for older graphics cards for Designer and Architect.
- Transit times, capacities and other attributes of discrete stocks that used equations were not being invalidated when variables were deleted from the model.
- When running models with the Simulation Log Panel open a large number of messaged coming into the panel would cause the software to response very slowly.
- The preference setting to not rescale graphs until the end of a run was being ignored in presentation mode on the interface and for published interfaces.
- Simulation events that cause a model to pause in a published interface would prevent results from the current time being correctly displayed in tables.
- Simulation events on models which did not have Pause before computing flows checked would delay result reporting.
- Locked tables and graphs from Version 10 models would end up being locked objects (and therefore selectable only by double clicking) when opened.
- Tables and group inputs from version 10 models would sometimes lose titles.
The following defects were fixed in Version 1.1
- The TRHOUGHPUT function was returning the wrong volume when DT was not 1. Similarly, the CTFLOW function wasn't initializing correctly when DT was not 1.
- CSV files with a byte order mark (BOM) for UTF8 weren't being imported correctly.
- Parameters weren't being validated for min < max in sensitivity testing.
- The RUNCOUNT builtin was incrementing during Stella Live.
- The decimal separator wasn't always correctly recognized on the panel with European number formatting active.
- Duplicate run names were being allowed after saving runs.
- Variables marked for import could have connections made to them from other variables.
- Flow order on non-negative stocks would sometimes be incorrect when implicit initialization circularities involving transit time were present.
- The Array Picker would sometimes appear when not needed, and not appear when needed, when editing equations.
- Deactivating an Import Control sheet wouldn't always reset the values to their model equations.
- The AltGR key was being interpreted as requesting a repeated action, making it difficult to enter some special characters in variable names.
- Copying and pasting between models would sometimes put incorrect styling on pasted elements.
- Shading on equation panel entries for non a2a equations wasn't displaying correctly in Windows 10.
- Bitmaps rendered to the Windows clipboard as companions for EMF content were of low quality.
- The MEAN function would generate a divide by zero error when applied to an empty queue.
- Sensi runs on models with comparative tables could cause the simulations to progress very slowly.
- Non-negative stocks under RK4 integration could sometimes go negative.
- Run specs wasn't validating start and stop times.
- Reloading runs in the data manager wasn't working correctly with graphicals.
The following defects were fixed in Version 1.0.3
- Graphics frames won't draw a border if none is selected.
- The graphs associated with graphicals weren't drawing on computers configured to use commas as a decimal delimiter (i.e., 1,234 instead of 1.234).
- Better layout of the legend when there were multiple labels.
- Sensitivity simulations are much faster when there are comparative graphs visible.
- The list of undo commands would sometimes include actions that were no longer relevant after multiple changes were reverted.
- Dial positions in Stella Live weren't consistently maintained for variables with a specified input range.
- Posters with navigation would sometimes cause unexpected behavior when they were triggered.
- Scales and ranges on ghosts weren't changing the underlying variable value.
- Pasting graphical points into Excel wasn't organizing the data correctly.
- Flows dragged to the left or top of a diagram could be placed out of range for further editing.
- The Find was organizing variables by subtypes of the basic stock, flow, converter types.
- Initial value circularity errors could persist after being corrected until a model was closed and opened.
- Some equations included in Delay Converters weren't being checked correctly for circularity.
- Drag and drop into the panel would sometimes change list order.
The following defects were fixed in Version 1.0.2
- Output formatted as currency wouldn't always display the $ sign.
-
Text boxes could lose their borders when opened in STELLA/iThink Version 10.
-
The Find dialog wouldn't show all modules in a model when ordered by module.
-
Cross level flows would sometimes save the constrained values of the flow used.
-
Ad-hoc sensitivity values would sometimes drop the first entry.
-
Input devices, numerical displays, and storytelling could be moved or partially lost when moving back and forth between Stella Professional and STELLA/iThink Version 10.
-
The INIT builtin wouldn't always give the same results on each simulation for models that required multiple pass initialization due to cross dependencies of conveyor transit times on initial stock values.
-
The Instantaneous argument to the CTMAX and CTMIN builtins wasn't working correctly. The documentation on CTMEAN, CTSTDDEV, and CYCLETIME has also been corrected.
-
Renaming modules would sometimes leave equations using cross level ghosts in an invalid state.
-
Some units equations were being incorrectly processed, which could cause the software to hang or close.
-
Running models with cycle-time active could sometimes cause outflows to be constrained based on inflow timing.
-
Builtins nested in the initial arguments of other builtins would sometimes cause initial value circularity errors to be missed.
-
Stock outflow priorities could change in models where inflows had a dependency of the outflows of non-negative stocks.
- Not all undone actions were being deleted when making changes after an undo, which could cause unexpected results on later undo attempts.
- Error messages displayed at the bottom weren't always updating correctly after changing equations.
- Showing tooltip values on multidimensional arrays could cause the software to crash.
- Importing data without the overwrite option selected would only be active for a single run (the way this is done has also changed in 1.0.2).
The following defects were fixed in Version 1.0.1
- Numbers exported from tables weren't following the language and region settings of the local computer when being formatted.
- Multiple outflows from queues weren't executing when higher priority downstream stocks had positive, but limited, inflows.
- Multiple message posters marked to be shown only once weren't being shown when the posting condition occurred after the first run.
- When running models with Cycle-Time, biflows weren't being allowed to go negative.
- The revert button on the equation panel didn't always work properly.
- Uniform sensitivity parameters were being saved as Normal.
- MODULO wasn't working when the base was 1.0.
- Flow units inside of modules weren't always being correctly computed.
- Tabbing between array elements in non-apply-to-all equations would sometimes cause crashes.
- Modules didn't support pictures as display icons.
- The Display panel would sometimes show units for variables that hadn't been assigned units.
- Creation of cross level flow ghosts would sometimes cause the software to crash.
- Connecting flows to stocks would fail for certain geometries.