This fixes the FP in cases like this:
```cpp
void f() {
bool b;
bool * x = &b;
if (x != nullptr)
x = 1;
}
```
It tracks the indirection of the uninit value in valueflow.
-Add iterator end patterns
-Add/fix size and access functions
-Remove marking QList and QStringList as std-like strings
-QStringList configuration now inherits from QList like it is actually the case
-Add tests
Use the AST a little bit more to improve the check. In order to do so,
rewrite the check to work from the outer function first and then check
the arguments, instead of the other way around.
It also fixes Trac ticket #9252, no warning is now given for
void* malloc1() {
return(malloc1(1));
}
This FP seems to be common in daca results.
It also makes it possible to improve handling of casts, for example
cppcheck now warns about
void f() {
strcpy(a, (void*) strdup(p));
}
But not for
char* f() {
char* ret = (char*)strcpy(malloc(10), "abc");
return ret;
}
These FP/FN were introduced when the check was switched to use the
simplified token list.
This fixes false positives from daca@home where freopen is used to
reopen a standard stream. There is no longer a warning for
void f() {
assert(freopen("/dev/null", "r", stdin));
}
It is hard to find good references, one that describes it a bit can
be found here:
https://manpages.ubuntu.com/manpages/bionic/man3/xmalloc.3pub.html
xfree() can be used instead of free().
A check, to verify that a memory leak is found if the memory allocated
via xmalloc() is not freed, has also been added.
* Fix#9225 (Crash on valid C++14 code)
This only fixes the crash. Specialization of nested templates is still
broken.
* fix cppcheck warnings
* fixed another cppcheck warning
* Use library for memleak checks
Change memleakOnRealloc and leakReturnValNotUsed to use library
configuration instead of hardcoding "realloc".
In order to do so, some care needs to be taken when matching for a
reallocation function, since it can no longer be assumed that the input
to be allocated is the first argument of the function. This complicates
getReallocationType() and checkReallocUsage() but is necessary in order
to handle for example freopen() properly.
Also, refactor memleakOnRealloc check to reduce duplicated code when
checking "a" and "*a". When doing so, extending the check to look for
arbitrary number of "*" can be done for free (just change an if
statement to a while statement). Most likely, this is an unusual case in
real world code.
* Remove redundant whitespace in Token::Match()
* Run on simplified checks
* Fix cppcheck warning
* Fix FP memory leak with unknown function call in condition
This was introduced in 8513fb81d2 when
fixing memory leaks for global variables allocated in condition. The
refactored code had an inconsistency where c and c++ code behaved
slightly differently when `var` is NULL. This seemed to not have an
impact as the code was written prior to 8513fb81d2,
but when the same code was used for conditions, FPs were introduced.
The introduced FPs were memleak warnings when there should have been an
information message about missing configurations for code like
void f() {
char *p = malloc(10);
if (set_data(p)) {}
}
Fix this by always returning true if varTok->Variable() is NULL for
both c and c++ code.
* Improve function name
This will diagnose more issues such as:
```cpp
void f(std::vector<int> &v) {
auto v0 = v.begin();
v.push_back(123);
std::cout << *v0 << std::endl;
}
```
* Improve configuration of g_try_realloc and g_try_realloc_n
* Mark g_realloc and similar functions as realloc functions
* Remove g_new, g_new0, etc as <alloc> functions, these are defined as
macros and handled as the functions they're expanded to.
* Add tests. TODO tests will be resolved by using the library
configuration in the checker.
* Fix adding unescaped slash token when splitting gcc case range.
Construction like case '!'...'~' converted to a list of separate case
tokens. When slas '\' symbol appears as a part of this list it was added
"as is", but it should be escaped like '\\' to be valid c++ code.
* Add test for switch-case range with slash
* Fix#9097 (Crash on thousands of "else ifs"s in gcc-avr package)
* increase recursion count maximum to 512 because cppcheck was hitting the 256 limit
* 512 was too much for windows
* Refactor Tokenizer::simplifyUsing to use continue to reduce indentation
added function findTemplateDeclarationEnd to skip template declarations
to reduce duplicate code
* fix travis build
This switches to use lifetime analysis to check for assigning to non-local variables:
```cpp
class test
{
public:
void f()
{
int x;
this->ptr = &x;
}
protected:
int *ptr;
};
```
* Partial fix for #9191 (simplifyTypedef: Problem when namespace is used)
This fixes simplifyUsing which has the same problem as simplifyTypedef.
simplifyUsing was designed to support using namespace but it was never
implemented. The changes are minor to add it.
simplifyTypedef requires much more work to support using namespace.
* reduce scope of variable
* make idx const
* Allow to configure realloc like functions
* memleakonrealloc: Bring back tests.
The old memleak checker was removed, and the tests for it was removed in
commit 9765a2dfab. This also removed the
tests for memleakOnRealloc. Bring back those tests, somewhat modified
since the checker no longer checks for memory leaks.
* Add realloc to mem leak check
* Add tests of realloc buffer size
* Configure realloc functions
* Add test of freopen
* Allow to configure which element is realloc argument
* Fix wrong close in test
cppcheck now warns for this
* Update manual
* Update docs
* Rename alloc/dalloc/realloc functions
Naming the member function realloc caused problems on appveyor. Rename
the alloc and dealloc functions as well for consistency.
* Change comparisson order
* Remove variable and use function call directly
* Create temporary variable to simplify
* Throw mismatchError on mismatching allocation/reallocation
* Refactor to separate function
* Fix potential nullptr dereference
As pointed out by cppcheck.
* Overlapping sprintf, improve handling of casts
If there is a cast of the argument buffer, cppcheck would print out the
expression including the cast, which looks a bit strange to talk about
Variable (char*)buf is used as...
Instead, only print the variable name without the cast.
Also, handle arbitrary many casts (the previous code only handled one).
Multiple casts of the input arguments is probably an unusual case in
real code, but can perhaps occur if macros are used.
* Fix printing of variable
... and add a test.
* Simplify testcase
* Update symbol database such that the override keyword implies that the function is also virtual
* Add test case for implicit override
* change isVirtual to hasVirtualSpecifier
* fix method documentation for getVirtualFunctionCalls and getFirstVirtualFunctionCallStack
* Fix isImplicitlyVirtual to consider the override keyword and document logic
* Fix getFirstVirtualFunctionCallStack and getVirtualFunctionCalls to use isImplicitlyVirtual instead of isVirtual so new test case passes
* Fix#9047 (c-style casts before malloc)
Note that there are still no warnings for c++-style casts
* Fix memleak check with casts of assignments in if-statements
* Fix possible null pointer dereference
As pointed out by cppcheck.
* Add check of astOperand2 when removing casts
This is similar to how it is done in other checks.
Further to pull request #1938. Changes were missed in previous commit.
Resolve warnings `warning: zero as null pointer constant` in code by
using C++ 11 recommended `nullptr`.
Building with enhanced clang warnings indicated a large number of
instances with the warning:
`warning: zero as null pointer constant`
Recommended practice in C++11 is to use `nullptr` as value for
a NULL or empty pointer value. All instances where this warning
was encountered were corrected in this commit.
Where warning was encountered in dependency code (i.e. external library)
no chnages were made. Patching will be offered upstream.
* fix adding instantiation of first argument to an instantiation
* add support for function pointer template variables
* fix more cases where templates ending in ">>" are changed to end in "> >"
* fix travis build
* standard types can't be a template parameter name
* remove redundant level == 0 checks
* fix lambda in template variable
* fix a test
This reverts commit 2a4be5ae1c.
When I look at daca@home now there are still lots of false negatives. So this bailout did not cause as much false negatives as I thought.
strdup() allocates the string length plus one for a terminating null
character. Add one to compensate for this.
Fixes false positive buffer out of bounds on code like this:
void f() {
const char *a = "abcd";
char * b = strdup(a);
printf("%c", b[4]); // prints the terminating null character
free(b);
}
Also, add a testcase for valueFlowDynamicBufferSize() and add tests for
strdup(), malloc() and calloc().
* Add non const version of some methods of Token
The aim is to reduce the (ab)use of const_cast.
* Cleanup some more const_cast in valueflow
* Remove useless const_cast
* Remove some const_cast from templatesimplifier
* Remove some const_cast from valueflow
* template simplifier: add 2 new template parameter simplifications
int{} -> 0
decltype(int{}) -> int
This fixes reduced test cases like #9153. I'm not sure they will help
real world code that much.
It was necessary to increase the pass count to 4 to get #9153 completly
simplified.
* relax decltype(type{}) simplification to any type
Refactored simplifyTemplateAliases to iterate over template type aliases
rather than instantiations. This fixed template type aliases that were
not templates.
Don't instantiate templates in template type aliases. They will get
instantiated once the type alias is instantiated. This required
increasing the template simplifier pass count to 3 so one of the
existing tests continued to work.
Specialized member classes declared outsize the class were not
recognized. This caused the the member class to be instantiated rather
than the specialized class. We already had a test for this but it was
wrong so it went unnoticed.
With the following code
int f(int x, int y) {
if (!!(x != 0)) {
return y/x;
}
cppcheck would wrongly warn that there might be a division by zero in
"return y/x;".
* template simplifier: fix instantiation of variadic template with no arguments
* fix white space change
* add support for <class...>
* add variadic template flag
* Fix issue 8890: AST broken calling member function from templated base class
* Format
* Check for double bracket
* Add test to createLinks2
* Remove extra test
* Reduce test case for links
This will warn for cases where searching in an associative container happens before insertion, like this:
```cpp
void f1(std::set<unsigned>& s, unsigned x) {
if (s.find(x) == s.end()) {
s.insert(x);
}
}
void f2(std::map<unsigned, unsigned>& m, unsigned x) {
if (m.find(x) == m.end()) {
m.emplace(x, 1);
} else {
m[x] = 1;
}
}
```
In the case of the map it could be written as `m[x] = 1` as it will create the key if it doesnt exist, so the extra search is not necessary.
I have this marked as `performance` as it is mostly concerning performance, but there could be a copy-paste error possibly, although I dont think thats common.
A common pattern is to have a function like similar to this:
bool isFlagSet(uint32_t f) {
return f & 0x4;
}
Warning that the function returns a non-boolean in this case is too
noisy, it would be better suited for a Misra check, so remove the
warnings in the most obvious cases.
Change the astStringVerbose() recursion to extend a string instead of
returning one. This has the benefit that for tokens where the recursion
runs deep (typically large arrays), the time savings can be substantial
(see comments on benchmarks further down).
The reason is that previously, for each token, the astString of its
operands was constructed, and then appended to this tokens astString.
This led to a lot of unnecessary string copying (and with that
allocations). Instead, by passing the string by reference, the number
of temporary strings is greatly reduced.
Another way of seeing it is that previously, the string was constructed
from end to beginning, but now it is constructed from the beginning to
end. There was no notable speedup by preallocating the entire string
using string::reserve() (at least not on Linux).
To benchmark, the changes and master were tested on Linux using the
commands:
make
time cppcheck --debug --verbose $file >/dev/null
i.e., the cppcheck binary was compiled with the settings in the
Makefile. Printing the output to screen or file will of course take
longer time.
In Trac ticket #8355 which triggered this change, an example file from the
Wine repository was attached. Running the above cppcheck on master took
24 minutes and with the changes in this commmit, took 22 seconds.
Another test made was on lib/tokenlist.cpp in the cppcheck repo, which is
more "normal" file. On that file there was no measurable time difference.
A synthetic benchmark was generated to illustrate the effects on dumping
the ast for arrays of different sizes. The generate code looked as
follows:
const int array[] = {...};
with different number of elements. The results are as follows (times are
in seconds):
N master optimized
10 0.1 0.1
100 0.1 0.1
1000 2.8 0.7
2000 19 1.8
3000 53 3.8
5000 350 10
10000 3215 38
As we can see, for small arrays, there is no time difference, but for
large arrays the time savings are substantial.
Before this fix, the code:
```
class A {
A(int, int x=3){
x;
}
};
```
Was considered OK.
But explicit keyword is still needed
I'm still new to open-source contributions, so I will gladly take advice.
This fixes simplifyUsing to remove 'typename' and 'template' from type
aliases of the form: using T3 = typename T1::template T3<T2>;
This lets the template simplifier instantiate the type alias which will
then remove the using type alias.
The crash will still happen if there is no instantiation because the
type alias will not be removed. The type alias is what cppcheck is
crashing on after the template simplifier and that still needs fixing.
* Fixed#8889 (varid on function when using trailing return type.)
Don't set varid for trailing return type.
* Add a test for #9066 (Tokenizer::setVarId: varid set for trailing return type)
* Handle 'arguments' sections in compile_commands.json
Previous code assumes 'commands' exists and ill assert if t does not.
* Correct typo checking for "arguments" rather than "commands"
* Use ostringstring rather than stringstream
* Add test deominstrating graceful degradation
* Add test for parsing "arguments" rather than "commands"
This is trying to fix the issue by fixing the ast and symbol database. First, the ast nodes will be created for the init list and the symbol database will not mark it as a scope. I am not sure if this is the correct approach as I dont really understand how the AST part works.
It did change the AST for `try {} catch (...) {}` but that is because it incorrectly treats `try {}` as an initializer list.
Paul Fultz II
Sebastian
orbitcowboy
Rikard Falkeborn
Daniel Marjamäki
IOBYTE
Ken-Patrick
amai2012
Denis
PKEuS
shaneasd
Scott Furry
Armin Müller
Nicodemes
Gary Leutheuser
amai2012
bbennetts