cppcheck failing for version 3.1.2

[psalvaggio]

• What is the issue you have?

The "make cppcheck" test is failing on version 3.1.2

• Please describe the steps to reproduce the issue. Can you provide a small but working code example?

Download the .tar.gz for the version, run "make cppcheck"

• What is the expected behavior?

Passing

• And what is the actual behavior instead?

[json-3.1.2]$ make cppcheck
cppcheck --enable=warning --inconclusive --force --std=c++11 single_include/nlohmann/json.hpp --error-exitcode=1
Checking single_include/nlohmann/json.hpp ...
[single_include/nlohmann/json.hpp:72]: (error) syntax error
Checking single_include/nlohmann/json.hpp: JSON_CATCH_USER...
Checking single_include/nlohmann/json.hpp: JSON_NOEXCEPTION;_CPPUNWIND;__EXCEPTIONS;__cpp_exceptions...
Checking single_include/nlohmann/json.hpp: JSON_THROW_USER...
Checking single_include/nlohmann/json.hpp: JSON_TRY_USER...
Checking single_include/nlohmann/json.hpp: _HAS_CXX14...
Checking single_include/nlohmann/json.hpp: _HAS_CXX17...
Checking single_include/nlohmann/json.hpp: _MSC_VER...
Checking single_include/nlohmann/json.hpp: _WIN32...
Checking single_include/nlohmann/json.hpp: __APPLE__...
Checking single_include/nlohmann/json.hpp: __GNUC__...
Checking single_include/nlohmann/json.hpp: __GNUC__;__GNUG__...
Checking single_include/nlohmann/json.hpp: __GNUC__;__GNUG__;__clang__...
Checking single_include/nlohmann/json.hpp: __HP_aCC;__HP_cc...
Checking single_include/nlohmann/json.hpp: __IBMCPP__...
Checking single_include/nlohmann/json.hpp: __ICC;__INTEL_COMPILER...
Checking single_include/nlohmann/json.hpp: __PGI...
Checking single_include/nlohmann/json.hpp: __SUNPRO_CC...
Checking single_include/nlohmann/json.hpp: __clang__...
Checking single_include/nlohmann/json.hpp: __linux__...
Checking single_include/nlohmann/json.hpp: __unix__...
make: *** [Makefile:259: cppcheck] Error 1

• Which compiler and operating system are you using? Is it a supported compiler?

GCC 7.3.1 on Fedora 27

• Did you use a released version of the library or the version from the develop branch?

Released

• If you experience a compilation error: can you compile and run the unit tests?

Unit tests are all passing.

[nlohmann]

I am aware of this - cppcheck seems to fail parsing the code. I always wanted to open a ticket at cppcheck, but never got to it.

[nlohmann]

The line in question is

template<typename = void, typename = void>
struct adl_serializer;

The problem seems to be the typename = . The following file is a minimal version of the issue:

template<typename = void> struct s;

yielding

Checking bug.hpp ...
[bug.hpp:1]: (error) syntax error

[nlohmann]

Unfortunately, it is not trivial to report an issue with cppcheck, see https://trac.cppcheck.net.

I explained the issue on their freenode chat. Let's see how things develop.

[nlohmann]

(Using Cppcheck 1.83 by the way)

[nlohmann]

If a type name is added before the = void, cppcheck can parse the file.

[theodelrieu]

I guess changing the code to have typename T = void, typename SFINAE = void would be also better to understand what the second template argument is for.

[nlohmann]

I applied for a trac account to report the error at cppcheck. In the meantime, I added the type names proposed by @theodelrieu to mitigate the bug from our side.

[psalvaggio]

Awesome! FYI, This came up while I was trying to update the Fedora RPM. (https://bugzilla.redhat.com/show_bug.cgi?id=1558490)

[nlohmann]

@psalvaggio I had a brief look at the Fedora issue and there seems to be an assertion inside an STL header. I'm not sure whether this is a library issue. But if so, please open a new issue here so we can have a look.

[nlohmann]

Closing this issue as original problem with cppcheck has been circumvented with 1c6b332.