Cppcheck: Reduction of False Positives with a MSVC Project File

[Stephan Bergmann]

On 07/09/18 06:31, Luke Benes wrote:
> By using our MSVC Project file to teach cppcheck about includes and 
> preprocessor configurations, I was able to reduce the number of warnings 
> from 9,141 to 25.  You can see the Report here:
> 
> https://drive.google.com/file/d/1NCV4Zl8vcWl30f6cDfTbGwxvW7lhMxw-/view?usp=sharing
> 
> While the results are thin, are there any valid issues here?

Commenting on the referenced cppcheck_reports2/index.html:

 > C:\Program Files (x86)\Windows 
Kits\10\Include\10.0.17134.0\shared\WTypesbase.h
 > 357	syntaxError		error	syntax error
 > C:\Program Files (x86)\Windows Kits\10\Include\10.0.17134.0\ucrt\malloc.h
 > 150	variableScope	398	style	The scope of the variable '_Marker' can 
be reduced.

The above two are in system code; nothing we can do about.

 > 
D:\cygwin\home\Hearthstone\lode\dev\core\dbaccess\win32\source\odbcconfig\odbcconfig.cxx
 > 119	resourceLeak	775	error	Resource leak: hModule
 > 123	resourceLeak	775	error	Resource leak: hModule

The above two should indeed be fixed

 > 
D:\cygwin\home\Hearthstone\lode\dev\core\framework\inc\helper\shareablemutex.hxx
 > 73	noExplicitConstructor	398	style	Class 'ShareGuard' has a 
constructor with 1 argument that is not explicit.

For all the nine noExplicitConstructor occurrences:  For one, what 
puzzles me is Cppcheck's fixation on "1 argument" ctors, as "explicit" 
has long surpassed being only relevant for ctors that can be called with 
one argument; if it warns about these, why doesn't it warn about ctors 
taking more arguments, too?  For another, for each individual ctor, they 
may be arguments for and against it being "explicit".  Would need deeper 
inspection, but generally feels on the level of rather unhelpful noise.

 > 
D:\cygwin\home\Hearthstone\lode\dev\core\framework\source\fwi\helper\shareablemutex.cxx
 > 33	copyCtorPointerCopying	398	style	Value of pointer 'm_pMutexRef', 
which points to allocated memory, is copied in copy constructor instead 
of allocating new memory.

The above is a false positive.

 > D:\cygwin\home\Hearthstone\lode\dev\core\include\osl\endian.h
 > 77	preprocessorErrorDirective		error	#error undetermined endianness

The two "preprocessorErrorDirective" are apparently fallout from 
Cppcheck (not being called in a correct way and thus?) not being able to 
determine which #if branch to take.

 > D:\cygwin\home\Hearthstone\lode\dev\core\include\osl\mutex.hxx
 > 120	noExplicitConstructor	398	style	Class 'Guard' has a constructor 
with 1 argument that is not explicit.
 > 127	noExplicitConstructor	398	style	Class 'Guard' has a constructor 
with 1 argument that is not explicit.
 > 154	noExplicitConstructor	398	style	Class 'ClearableGuard' has a 
constructor with 1 argument that is not explicit.
 > 161	noExplicitConstructor	398	style	Class 'ClearableGuard' has a 
constructor with 1 argument that is not explicit.
 > 200	noExplicitConstructor	398	style	Class 'ResettableGuard' has a 
constructor with 1 argument that is not explicit.
 > 207	noExplicitConstructor	398	style	Class 'ResettableGuard' has a 
constructor with 1 argument that is not explicit.
 > D:\cygwin\home\Hearthstone\lode\dev\core\include\rtl\byteseq.h
 > 199	noExplicitConstructor	398	style	Class 'ByteSequence' has a 
constructor with 1 argument that is not explicit.
 > 210	noExplicitConstructor	398	style	Class 'ByteSequence' has a 
constructor with 1 argument that is not explicit.

(see above)

 > D:\cygwin\home\Hearthstone\lode\dev\core\include\sal\alloca.h
 > 60	preprocessorErrorDirective		error	#error "unknown platform: please 
check for alloca"

(see above)

 > D:\cygwin\home\Hearthstone\lode\dev\core\sal\osl\w32\dllentry.cxx
 > 224	ignoredReturnValue	252	warning	Return value of function 
CreateThread() is not used.
 > 224	leakReturnValNotUsed	771	error	Return value of allocation 
function 'CreateThread' is not stored.

 From the comments around the above two, they appear to be intentional.

 > D:\cygwin\home\Hearthstone\lode\dev\core\sal\rtl\byteseq.cxx
 > 43	variableScope	398	style	The scope of the variable 'nElements' can 
be reduced.

The above could indeed be fixed.

 > 82	incorrectStringBooleanError	571	warning	Conversion of string 
literal "### null ptr!" to bool always evaluates to true.
 > D:\cygwin\home\Hearthstone\lode\dev\core\sot\source\sdstor\stgavl.cxx
 > 137	incorrectStringBooleanError	571	warning	Conversion of string 
literal "The pointer is not allowed to be NULL!" to bool always 
evaluates to true.
 > 149	incorrectStringBooleanError	571	warning	Conversion of string 
literal "The pointer is not allowed to be NULL!" to bool always 
evaluates to true.
 > 180	incorrectStringBooleanError	571	warning	Conversion of string 
literal "The pointer is not allowed to be NULL!" to bool always 
evaluates to true.
 > 191	incorrectStringBooleanError	571	warning	Conversion of string 
literal "The pointer is not allowed to be NULL!" to bool always 
evaluates to true.
 > 302	incorrectStringBooleanError	571	warning	Conversion of string 
literal "The pointers may not be NULL!" to bool always evaluates to true.

The above six are all occurrencs of the assert(X && "explanation") 
idiom, and as such false positives.  (Odd that there are only warnings 
for these six occurrences; I'm pretty sure there's more across the code 
base.  Maybe that's because "the 'LibreOffice.sln' project is incomplete 
and cannot yet generate full builds.")