man: Added writing-addons.md

This commit is contained in:
Daniel Marjamäki 2022-06-19 12:23:29 +02:00
parent ee85de5ac3
commit b6a364eda4
1 changed files with 400 additions and 0 deletions

400
man/writing-addons.md Normal file
View File

@ -0,0 +1,400 @@
---
title: Writing addons
subtitle: Version 2.9
author: Cppcheck team
lang: en
documentclass: report
---
# Introduction
This document provides an overview about writing Cppcheck addons.
# Overview of data
## Tokens
See class `Token` in cppcheckdata.py
Cppcheck splits up the code in tokens. Operators, numbers, identifiers, etc.
Example C code:
ab = a + b;
Addon code:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
print(token.str)
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
ab
=
a
+
b
;
There are various properties in the `Token` class and some of those will be discussed below.
## AST - Abstract syntax tree
Cppcheck creates a syntax tree for every expression.
Example C code:
ab = a + b;
Addon code:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
out = token.str
if token.astParent:
out += f' parent:"{token.astParent.str}"'
if token.astOperand1:
out += f' astOperand1:"{token.astOperand1.str}"'
if token.astOperand2:
out += f' astOperand2:"{token.astOperand2.str}"'
print(out)
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
ab parent:"="
= astOperand1:"ab" astOperand2:"+"
a parent:"+"
+ parent:"=" astOperand1:"a" astOperand2:"b"
b parent:"+"
;
## ValueType
Data type of expressions are provided by `class ValueType` in cppcheckdata.py.
Example C code:
short a;
a = a + 10;
Addon code:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
print(f'{token.str} : {token.valueType}')
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
short : None
a : ValueType(type='short', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
; : None
a : ValueType(type='short', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
= : ValueType(type='short', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
a : ValueType(type='short', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
+ : ValueType(type='int', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
10 : ValueType(type='int', sign='signed', bits=0, typeScopeId=None, originalTypeName=None, constness=0, pointer=0)
; : None
The `pointer` property is a simple counter.
int p => pointer=0
int *p => pointer=1
int **p => pointer=2
The `constness` property is a bitmask.
int * * => constness=0
const int * * => constness=1
int * const * => constness=2
int * * const => constness=4
const int * const * => constness=3
const int * const * const => constness=7
## Variable
Information about a variable is provided by `class Variable` in cppcheckdata.py.
Example code:
short a[10];
a[0] = 0;
Addon code:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
print(f'{token.str} : {token.variable}')
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
a : Variable(Id='0x55f432fc1580', nameTokenId='0x55f432fe0850', typeStartTokenId='0x55f432fc0eb0', typeEndTokenId='0x55f432fc0eb0', access='Global', scopeId='0x55f432fe0360', isArgument=False, isArray=True, isClass=False, isConst=False, isGlobal=True, isExtern=False, isLocal=False, isPointer=False, isReference=False, isStatic=False, constness=0)
[ : None
10 : None
] : None
; : None
a : Variable(Id='0x55f432fc1580', nameTokenId='0x55f432fe0850', typeStartTokenId='0x55f432fc0eb0', typeEndTokenId='0x55f432fc0eb0', access='Global', scopeId='0x55f432fe0360', isArgument=False, isArray=True, isClass=False, isConst=False, isGlobal=True, isExtern=False, isLocal=False, isPointer=False, isReference=False, isStatic=False, constness=0)
[ : None
0 : None
] : None
= : None
0 : None
; : None
## Scope
See `class Scope` in cppcheckdata.py.
Every token is in some scope.
If you see {} in the code then that is a scope of some kind. There is one scope that is not surrounded by {}; the global scope.
Example c code:
int x;
void foo()
{
if (x)
{
x = 0;
}
}
Example addon #1 (list all scopes):
import cppcheck
@cppcheck.checker
def func(cfg, data):
for scope in cfg.scopes:
print(scope.type)
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
Global
Function
If
Example addon #2 (show scope for each token):
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
print(f'{token.str} : {scope.type}')
Output:
$ cppcheck --dump test.c
$ python3 runaddon.py myaddon.py test.c.dump
Checking 1.c.dump...
Checking 1.c.dump, config ...
int : Global
x : Global
; : Global
void : Global
foo : Global
( : Global
) : Global
{ : Function
if : Function
( : Function
x : Function
) : Function
{ : If
x : If
= : If
0 : If
; : If
} : If
} : Function
### Special tokenlist tweaks and else if
The cfg.tokenlist has some tweaks. In C/C++ code it is optional to use `{` and `}` around the conditional code,
if the conditional code is only a single statement. However Cppcheck adds "{" and "}" tokens if those are missing.
One more tweak is that in cfg.tokenlist there is no "else if" scope.
Example C code:
void foo(int x)
{
if (x > 0)
--x;
else if (x < 0)
++x;
}
The tokens in the cfg.tokenlist will look like this:
void foo(int x)
{
if (x > 0)
{
--x;
}
else
{
if (x < 0)
{
++x;
}
}
}
And there are 2 "If" scopes here. And 1 "Else" scope.
## Function
The class `Function` in cppcheckdata.py represents a function that is declared somewhere.
Example code:
void foo(int x);
void bar()
{
foo(1);
}
Example addon #1:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
print(f'{token.str} : {token.function}')
Output:
void : None
foo : Function(Id='0x55c5f9330f60', tokenId='0', tokenDefId='0x55c5f93510d0', name='foo', type='Function', isVirtual=False, isImplicitlyVirtual=False, isInlineKeyword=False, isStatic=False, argumentId={1: '0x55c5f9351260'})
( : None
int : None
x : None
) : None
; : None
void : None
bar : Function(Id='0x55c5f9331040', tokenId='0x55c5f93314a0', tokenDefId='0x55c5f93314a0', name='bar', type='Function', isVirtual=False, isImplicitlyVirtual=False, isInlineKeyword=False, isStatic=False, argumentId={})
( : None
) : None
{ : None
foo : Function(Id='0x55c5f9330f60', tokenId='0', tokenDefId='0x55c5f93510d0', name='foo', type='Function', isVirtual=False, isImplicitlyVirtual=False, isInlineKeyword=False, isStatic=False, argumentId={1: '0x55c5f9351260'})
( : None
1 : None
) : None
; : None
} : None
## Value flow
For Cppcheck, by default variables and expressions have "unknown" values. If the value is only constrained by the data type that is typically not enough information to write warnings with precision. Unless all the values possible in the data type are bad.
In Cppcheck terminology an expression will have a "possible" value if Cppcheck can determine that there will be a specific value in some control flow paths.
In Cppcheck terminology an expression will have a "known" value if Cppcheck can determine that there will be a specific value in all control flow paths.
An expression can have several "possible" values. But it can't have several "known" values.
Example code:
void foo(int x) // <- values of x is only constrained by data type. there are no "possible" or "known" values here.
{
a = x; // <- assuming that condition below is not redundant, x can have value 2.
if (x == 2)
{
b = x + 2; // <- value of x is always 2 when this code is executed. It's "known".
}
else
{
c = x;
}
d = x + 10; // <- value of x can be 2 when this code is executed. It's "possible".
}
Addon:
import cppcheck
@cppcheck.checker
def func(cfg, data):
for token in cfg.tokenlist:
if token.values:
print(f'line {token.linenr} str="{token.str}"')
for value in token.values:
print(f' {value}')
Output:
$ ../cppcheck --dump 1.c ; python3 runaddon.py myaddon.py 1.c.dump
Checking 1.c ...
Checking 1.c.dump...
Checking 1.c.dump, config ...
line 3 str="="
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='possible', inconclusive=False)
line 3 str="x"
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='possible', inconclusive=False)
line 4 str="2"
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=None, valueKind='known', inconclusive=False)
line 6 str="="
Value(intvalue=4, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='known', inconclusive=False)
line 6 str="x"
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='known', inconclusive=False)
line 6 str="+"
Value(intvalue=4, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='known', inconclusive=False)
line 6 str="2"
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=None, valueKind='known', inconclusive=False)
line 12 str="="
Value(intvalue=12, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='possible', inconclusive=False)
line 12 str="x"
Value(intvalue=2, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='possible', inconclusive=False)
line 12 str="+"
Value(intvalue=12, tokvalue=None, floatvalue=None, containerSize=None, condition=4, valueKind='possible', inconclusive=False)
line 12 str="10"
Value(intvalue=10, tokvalue=None, floatvalue=None, containerSize=None, condition=None, valueKind='known', inconclusive=False)
If the value is determined from a condition then the attribute `condition` points out the line that has the condition.