Misra 17.7 doesn't detect some cases

[Jean-Loic MARTIN]

Hi all,
I am having troubles with Misra rules 17.7 checks.

The Misra add-on detects that memset returns code is not tested and generates an error => Good.
But it doesn't detects that memcpy retunr code isn't tested => Not good.

The code:

void main( CHAR* pacLongitude )
{
    memcpy( acBufferTemp , (const void*) ( pacLongitude ) , 3 );
    u8Degrees = (U8)(atoi(acBufferTemp));
    memset( acBufferTemp , 0 , sizeof(acBufferTemp) );          /* Reset buffer */
    memcpy( acBufferTemp , (const void*) ( pacLongitude + 3U ) , 2 );
    u8Minutes = (U8)(atoi(acBufferTemp));
    memcpy( acBufferTemp , (const void*) ( pacLongitude + 6U ) , 4 );
    return;
}

My line of command with result below:

"c:\Cppcheck\cppcheck.exe"  -DMDF4_TEST_ABG5 --enable=all --inline-suppr --suppress=unusedFunction --suppress=missingInclude --addon=./misra.json ./*.c
Checking essai_misra_17_7.c ...
Checking essai_misra_17_7.c: MDF4_TEST_ABG5=1...
essai_misra_17_7.c:5:11: style: The value returned by a function having non-void return type shall be used [misra-c2012-17.7]
    memset( acBufferTemp , 0 , sizeof(acBufferTemp) );          /* Reset buffer */

memset and memcpy return void pointer.
They both should generate deviation to Misra 17.7 rule.
*
I have checked in dump file and type pointer field is equal to 1. So the python script should generates an error, but doesn't. I don't know why.

[Georgiy Komarov]

Hello,

Thanks for the report.

It seems that problem not in MISRA addon. For some reasons, Cppcheck does not include value type of memcpy function in the dump file.

Here is a minimal example:

void fn()
{
    memcpy(buf, buf2, 3);
    memset(buf, 0, sizeof(buf));
}

This code produces the following tokens in the dump file:

    <token id="0x110c5b0" file="addons/test/memcpy-memset.c" linenr="3" column="5" str="memcpy" scope="0x1112360" type="name" astParent="0x110cb80"/>
    <token id="0x110cb80" file="addons/test/memcpy-memset.c" linenr="3" column="11" str="(" scope="0x1112360" link="0x110ff00" astOperand1="0x110c5b0" astOperand2="0x110fd10"/>

...

    <token id="0x11102e0" file="addons/test/memcpy-memset.c" linenr="4" column="5" str="memset" scope="0x1112360" type="name" astParent="0x110f510"/>
    <token id="0x110f510" file="addons/test/memcpy-memset.c" linenr="4" column="11" str="(" scope="0x1112360" link="0x110df00" astOperand1="0x11102e0" astOperand2="0x110e8b0" valueType-type="void" valueType-pointer="1" valueType-reference="None" valueType-originalTypeName="void *"/>

The ( token after memcpy does not contain the value type, expected by misra.py here: https://github.com/danmar/cppcheck/blob/1daf1ec108f66633f49c51a778988fe337d6cf13/addons/misra.py#L2162

[Daniel Marjamäki]

thanks georgy! Sounds like an easy fix. update memcpy configuration in cfg/std.cfg file. If this line is added: <returnValue type="void*"/> then the return type should be added.

[Georgiy Komarov]

Thanks, here is a fix: https://github.com/danmar/cppcheck/pull/2777
This patch fixes MISRA 17.7 false positives.

[Jean-Loic MARTIN]

Thank you very much for your good job!
I have implemented the fix and it works well... 58 new MISRA errors in my code !

Note: It was false negatives.

[Jean-Loic MARTIN]

Hi all,
i have checked the std.cfg file and it seems that the problem occurs also for the following functions (which have no return type dexcribed):

void * memmove(void *ct, const void *cs, size_t n); 
wchar_t * wmemmove(wchar_t *ct, const wchar_t *cs, size_t n);

Is it possible to add the return type for these two functions please ?

[Daniel Marjamäki]

do you think you could create a github pull request to fix such problems?

[Jean-Loic MARTIN]

I never used github. I have created an account but don't know how to do a accurate pull request.

[orbitcowboy]

Return value types added with https://github.com/danmar/cppcheck/commit/98b62384504f539d06c686fdc5dbaf0cdf14daa7

Could you give it a try?

[Jean-Loic MARTIN]

Hi , with your file, it gives me an error :

nofile:0:0: information: Failed to load std.cfg. Your Cppcheck installation is broken, please re-install. The Cppcheck binary was compiled without FILESDIR set. Either the std.cfg should be available in c:/Cppcheck/cfg or the FILESDIR should be configured. [failedToLoadCfg]

[Daniel Marjamäki]

that is a major problem. how do you compile/install? do you run the CLI or GUI?

[Jean-Loic MARTIN]

Hi Daniel, i have installed CppCheck V2.1 and unzip it in C:\CppCheck. That's all.
Then i call CppCheck directly in pre-build (for naming convention) and post-build rules (for misra rules). I have just downloaded the std.cfg file from https://github.com/danmar/cppcheck/commit/98b62384504f539d06c686fdc5dbaf0cdf14daa7 and copy it in C:\CppCheck/cfg directory. And it gives me this error. I don't know why. I have replaced it with my old std.cfg files and it runs well.