This patch augments the XML dumps with a 'directivelist'
subnode which lists all raw preprocessor directives met
while reading the source code in each configuration.
Also, the addons/cppcheckdata.py file has been extended
to give easy access to the list of directives and to
provide Python support for the --template (or short -t)
option.
Finally, an new addon, addons/y2038/y2038.py, is created
to detect when a glibc symbol might be Y2038-sensitive,
based on whether and how _TIME_BITS and _USE_TIME_BITS64
are defined when meeting the symbol.
With multiple configurations, option --dump only dumps
the last configuration.
Fix it to dump every configuration.
Also update all Python addons so that they can handle
multiple-configuration dumps.
Additionally run autopep8 on addons/*.py.
The results of 'make test' before and after applying
this commit are identical.
Reorganize cppcheck.cpp to separate actual checking from
tokenizing and XML dumping. This implies splitting checkFile()
in three parts, checkRawTokens, checkNormalTokens() and
checkSimplifiedTokens().
syntaxErrors were not suppressable, because they were treated as
InternalErrors which were thrown and catched during the checking,
and normal suppression rules were not applied for those.
We fix this by calling the normal reportErr() function that does
suppression matching.
Introduce a new bool setting jointSuppressionReport
that will be set by the analyseWholeProgram() code path.
When the flag is enabled, unmatched suppressions are
collected after running the final whole program analysis
to prevent false positives for the unusedFunction check.
The check functions in the unit test
for single / multi file suppressions were unified.
The underlying problem was in lib and is fixed by changing the behaviour of CppCheck::check(). Checking is performed on empty content without attempt to load from file.
- Don't abort checking (reverts 42140b6488)
- Modified error message: New Id unhandledCharacter, removed redundant line information, improved message text
- Made Tokenizer::getErrorMessages static - Avoids creation of unnecessary instance of Tokenizer
- Changed Tokenizer::removeExceptionSpecifications to common style for simplification functions. In contrast to the comment, this function doesn't call itself recursivly - fixed comment.
- Made Tokenizer::IsScopeNoReturn static.
Currently the information severity messages are outputted as error
messages with Severity::Information. This causes constant confusion
as people think it as mildest error severity (and rightfully so).
When it was meant to be for printing messages about the checking
procedure itself (like missing header files etc).
So I'm adding a new function for the ErrorLogger for printing these
informative messages. This makes clear the distinction of errors
found from the code and messages related to the checking itself.
It also makes it easier for clients to handle these separately.
- Added support for pointers in self assignement check
- Removed redundant for loop in checknullpointer.cpp
- Fixed warning about signed/unsigned mismatch in cppcheck.cpp by making Settings::_maxConfig unsigned
- Added forgotten initialization of Settings::_relativePaths
- Some PCRE-Rules specific code hidden behind HAVE_RULES
- Use initialization list in ErrorLogger::ErrorMessage::ErrorMessage() and CppCheck::CppCheck
- Avoided unnecessary copies of std::strings in cppcheck.cpp
- Moved "// Alert only about unique errors"-code to make it work in debugFalsePositive mode
Detect sign extension problems when variable is a reference (#3637)
Refactorizations:
- Tokenizer::getFiles returns a reference instead of a pointer, because its guaranteed that no nullpointer is returned
- Remove signed/unsigned in one step for "%type% signed|unsigned"
- Fixed recently introduced compiler warning in symboldatabase.cpp
- Fixed#3628
- Added support for friend
Improved symbol database:
- friend scopes are now set
- Added findScopeByName function
Refactorizations:
- Removed some unnecessary "virtual" keywords
- Removed unnecessary _filename member variable, pass it as argument instead
- Made CppCheck::replaceAll static, since it is independant from a specific CppCheck instance, Pass string to be modified by reference
-- Use MathLib::toLongNumber for conversion in tokenizer (Fix#3610)
-- Handle octal numbers in tokenizer
- Refactorizations in MathLib::toLongNumber and Settings
- Previously MathLib errors did not provide a filename, but after this change at least source file name should be printed
and if token is provided, also line number should be printed.
- Change also Token to use InternalError
- Modify Cppcheck-class to catch InternalError instead of Token
- Run dmake to update Makefile
Whitespace was removed from the append data for every configuration
and every file: n(configuration) * n(files).
Removing whitespace immediately after the append data was read can
significantly reduce the execution time and memory usage.
It also allows further improvement because copies to the temporary
object appendCode are no longer needed.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
This is ment to be used for debugging false positive errors in Cppcheck.
Current implementation tries two alternatives. Without all headers or with all headers and prints out the option with
less code. In future versions this could try with individual headers or group of header files.
The "ExtraVersion" can be used for things like Git commit Id,
release tag (version control), release date etc. If the string
is empty, nothing is printed.
When user wants to see only errors printed (--quiet in CLI) we
must obey that. And not print unmatchedSuppressions list.
Ticket: #2895 (Cannot suppress unmatchedSuppression reports)
Daniel Marjamäki
amai2012
Albert ARIBAUD (3ADEV)
Lauri Nurmi
PKEuS
Miika-Petteri Matikainen
Simon Shanks
Thomas Jarosch
PKEuS
Sam Truscott
Reijo Tomperi
Edoardo Prezioso
Daniel Marjamäki
Kimmo Varis
Simon Martin
Stefan Weil
Stefan Weil
Joshua Beck
Greg Hewgill