Main Content

Run Polyspace Analysis on Custom Code in C Function Block

You can run a Polyspace® analysis on the custom C code in a C Function (Simulink) block from Simulink®. Polyspace checks the custom C code for errors and bugs while keeping the model specific information such as design range specification, nature and number of inputs that are specified in the Simulink model.

Prerequisites

Before you run Polyspace from Simulink, you must link your Polyspace and MATLAB® installations. See Integrate Polyspace with MATLAB and Simulink.

To open the model used in this example, in the MATLAB Command Window, run:

openExample('polyspace_code_prover/CScriptDemoExample')
open_system('psdemo_model_link_sl_cscript');

Open Model for Running Polyspace Analysis on Custom Code in C Function Block

This example uses:

Open Script

The model contains a C Function block called Command Strategy inside the controller subsystem.

The Command Strategy block implements a look-up table using custom C code and outputs a value result based on two inputs x and y.

Run Polyspace Analysis

Run Polyspace Analysis from Simulink Editor

Click the Apps tab and select Polyspace Code Verifier to open the Polyspace tab.

  1. Select Bug Finder or Code Prover in the Mode section of the Polyspace tab. A Code Prover analysis detects run-time errors while a Bug Finder analysis detects coding defects and coding rule violations.

  2. To run a Polyspace analysis on the custom C code in the C Function block, select Custom Code Used in Model from the drop-down list in the Analyze section.

    A section of the Polyspace tab on the Simulink toolstrip. In this section, you can specify which code to analyze, and then run analysis.

  3. To start the Polyspace analysis, click the Run Analysis button. The MATLAB Command Window displays the progress of the analysis.

  4. After the analysis, the Polyspace user interface opens with the results. You can choose to not open the results automatically after the analysis by unselecting Open results automatically after verification in Settings. To open the results after the analysis is finished, click the Analysis Results button.

  5. To see the unfiltered results of the Polyspace analysis, click Clear active filters from the Showing drop-down list in the Results List pane. If you run a Code Prover analysis, the unfiltered results for the controller subsystem contain two red checks and an orange check.

  6. To organize the results by family, click square drop down menu and select Family.

    Results list organized by family of results. The top level shows result types such as Run-time Check, Global Variable, and so on. Within run-time checks, the results are organized by check color.

To switch between a Bug Finder and Code Prover analysis, return to the Simulink Editor from the Polyspace user interface. Switch between Bug Finder and Code Prover in the Mode section and run the analysis again.

Run Polyspace Analysis from MATLAB

You can run a Polyspace Code Prover™ analysis on the custom code for this model from MATLAB Editor or the Command Window using this code:

% Load the model 'psdemo_model_link_sl_cscript'
load_system('psdemo_model_link_sl_cscript');
% Create a 'pslinkoptions' object
mlopts = pslinkoptions('psdemo_model_link_sl_cscript'); 
% Specify whether to run 'CodeProver' or 'BugFinder' Analysis
mlopts.VerificationMode = 'CodeProver';
% Specify custom code as analysis target and run the analysis
pslinkrun('-slcc','psdemo_model_link_sl_cscript',mlopts);

Identify Issues in C Code

To identify issues in the custom C code, use the information in the Result Details pane and the Source pane of the Polyspace user interface. If you do not see these panes, go to Window > Show/Hide View and select the missing pane. For details on the panes, see Result Details in Polyspace Desktop User Interface and Source Code in Polyspace Desktop User Interface.

Identify C Function Block Inputs and Outputs in Source Pane

Polyspace wraps the code in the C Function block in a custom code wrapper. The inputs and outputs of the C Function block are declared as global variables. The custom C code is called as a function.

/* Variables corresponding to inputs ..*/
// global In... 
/* Variables corresponding to outputs*/
// global Out...
/* Wrapper functions for code in block  */
// void ...(void){
    //...
}

  • The global variables corresponding to inputs start with In, such as In1_psdemo_model_link_sl_cscript_98_Command_strategy.

  • The global variables corresponding to outputs start with Out, such as Out1_psdemo_model_link_sl_cscript_98_Command_strategy.

  • The void-void function contains the custom C code with the input and output variables replaced by the global variables. If you have multiple C Function blocks, then the code in each block is wrapped in separate functions.

The global variables reflect the properties of the input and output of the C Function block, including their data range, data type, and size. If you have multiple inputs, then the order of the global variables is the same as the order of the input defined in the C Function block. This table shows the input and output variables of the block in this example and their corresponding global variables in the Source pane.

Global Variable Name in Source PaneScopeVariable Name in C Function Block
In1_psdemo_model_link_sl_cscript_98_Command_strategyInputx
In2_psdemo_model_link_sl_cscript_98_Command_strategyInputy
Out1_psdemo_model_link_sl_cscript_98_Command_strategyOutputresult

Identify issues in the custom code by reviewing the wrapped code in the Source pane. Use the tooltip in the Source pane and the information in the Result Details pane to fix the issues. This workflow applies to Code Prover and Bug Finder analyses.

Illegally dereferenced pointer

The red check Illegally dereferenced pointer highlights the dereferencing operation after the for loop.

tmp = *p + 5;
The Result Details pane states that the pointer *p is outside its bounds. To find the root cause of the check, follow the lifecycle of the pointer leading to the illegal dereferencing.

  1. At the start of its lifecycle, the pointer *p points to the first element of array which has 100 elements.

  2. Then p is incremented 100 times, pointing *p to the nonexistent location array[100].

  3. The dereferencing operation in tmp = *p+5; becomes illegal, causing a red check.

Out of Bounds array index

The red check Out of Bounds array index highlights the array indexing operation in the if condition.

 if (another_array[return_val - i + 9] != 0)
The Result Details pane states that the size of another_array is 2 while the index value return_val-i+9 ranges from 2 to 18. To find the root cause of the check, track the values of the variables return_val and i using the tooltip. When you hover over any instance of the variables in the Source pane, the tooltip is displayed.

  1. The value of i is 100.

  2. The value of return_val ranges from 93 to 109 because of the prevailing condition: if ((return_val > 92) && (return_val < 110)).

  3. The index value (return_val-i+9) evaluates to a range of 2 to 18.

  4. The index values are out of bounds for the array another_array, causing a red check.

Overflow

The orange Overflow check highlights the assignment to return_val. The Result Details pane states that the check is related to bounded input values. To find the root cause of the check, check the data type and corresponding range of the variables by using the tooltip.

  • The input values x and y correspond to these respective global variables

    • In1_psdemo_model_link_sl_cscript_98_Command_strategy

    • In2_psdemo_model_link_sl_cscript_98_Command_strategy

  • The first input x is an unbound unsigned integer. Because x is unbound, it has the full range of an unsigned integer, which is from 0 to 65535.

  • The second input y is a bounded unsigned integer ranging from 0 to 1023.

  • x-y is assigned to the unbound signed integer return_val. Because return_val is unbound, it has full range from -32768 to 32767.

  • The range of x-y is 1023 to 65535, while the range of return_val is -32768 to 32767.

  • Some possible values of x-y cannot fit into return_val, causing the orange check.

For details about interpreting results of a Polyspace Code Prover analysis, see Interpret Code Prover Results in Polyspace Desktop User Interface.

Fix Identified Issues

Modify the custom C code or the model to fix the issues. You can fix a Polyspace check in several ways. The examples here illustrate the general workflow of fixing Polyspace checks.

Illegally dereferenced pointer

You can address this check in several ways. Modify the C code so that a nonexistent memory address is not accessed.

  1. Return to the Simulink Editor and double-click the C Function block to open the custom code.

  2. Use the index operator on array to access a valid array index. You can access indices from 0 to 99 because array has 100 elements. Accessing indices beyond this range results in a run-time error in Simulink.

    // access any index between 0 to 99
tmp = array[50] + 5;
    Alternatively, assign the address of a valid memory location to p before the dereferencing operation. For example, *p can point to the 51st element in array.
    // After the for loop, point p to a valid memory location
p = &(array[50]);
// ...
tmp = *p + 5;

Out of Bounds array index

You can address this check in several ways. Modify the code so that the size of another_array[] remains larger than or equal to the index value return_val-i+9.

  1. Return to the Simulink Editor and double-click the C Function block to open the custom code.

  2. Modify the prevailing condition on return_val so that the index value return_val-i+9 evaluates to 0 or 1 for the possible values of return_val.

    if ((return_val > 91) && (return_val < 93))
//...
    Alternatively, declare another_array with size 19.
    int another_array[19];

Overflow

You can address this check in several ways as well. Modify the C code or the model so that the range of the right side of the assignment operation remains equal to or larger than that of the left side.

  1. Return to the Simulink Editor.

  2. Saturate the input variables x and y in the model so that their difference can fit into a 16-bit integer. The workflow for fixing Overflow by using saturation blocks is described in Run Polyspace Analysis on Custom Code in C Caller Blocks and Stateflow Charts.

Alternatively, increase the size of return_val in the custom C code to accommodate x-y.

  1. Return to the Simulink Editor and double-click the C Function block to open the custom code.

  2. Declare return_val as a 32-bit integer.

    int32_T return_val;

For details about addressing Polyspace results, see Address Results in Polyspace User Interface Through Bug Fixes or Justifications.

See Also

|

Related Topics