The tests seem to fail because they use '--std=cplusplus11'. Should I change the parsing to accept the 'cplusplus' prefix in addition to 'c++'?

The tests seem to fail because they use '--std=cplusplus11'. Should I change the parsing to accept the 'cplusplus' prefix in addition to 'c++'?

No, just turn the TODO_ASSERT_EQUALS into a regular ASSERT_EQUALS.

So this line now results in unknownMacro errors: ./cppcheck -q --template=selfcheck --error-exitcode=1 --library=cppcheck-lib --library=qt -D__CPPCHECK__ -D__GNUC__ -DQT_VERSION=0x060000 -DQ_MOC_OUTPUT_REVISION=68 -DQT_CHARTS_LIB -DQT_MOC_HAS_STRINGDATA --enable=unusedFunction --exception-handling -rp=. --project=cmake.output.notest/compile_commands.json --suppressions-list=.selfcheck_unused_suppressions --inline-suppr
I wonder what that compile_commands.json looks like.

sorry for late reply.

Ok, seems to work now. I'm still unsure about what to do with the 'gnu++11' options generated by CMake in the self check though. Would it be appropriate to sed it to c++11 in the CI script as a quick fix?

Ok, seems to work now. I'm still unsure about what to do with the 'gnu++11' options generated by CMake in the self check though. Would it be appropriate to sed it to c++11 in the CI script as a quick fix?

I think that users might have to sed sometimes. however it wouldn't hurt to handle gnu++11 in the importproject directly. using --std=gnu++11 on the command line should still be an error but if we see it an compile_commands.json imho it would make sense to handle it like "c++11".

Shouldn't we be consistent here, i.e. handle gnu++11 everywhere or not at all?

Perhaps handle it if a certain cli option is set (e.g. —allow-gnu)? Adds flexibility but prevents any misunderstandings.

Shouldn't we be consistent here, i.e. handle gnu++11 everywhere or not at all?

my thinking was that why would you really want to use --std=gnu++11 on the command line instead of c++11? I don't see the use case really.

in a compile_commands.json file it's a different story. That is not a cppcheck option but a gcc option. And I assume it's not very uncommon.

I have a commit that handles gnu++XX as c++XX in command files, should I push?

I have a commit that handles gnu++XX as c++XX in command files, should I push?

That's fine with me.

Sorry for the late reply but I have been out sick and still not fully recovered.

This is possibly all wrong. The whole knowledge about what is a valid standard already exists in simplecpp and that needs to be leveraged. Since it is handled differently in Cppcheck I added the TODO. I cannot really tell if I just didn't have the time to implement it (and then forgot) or if there was something more problematic preventing this.

This also needs tests for the project import.

The whole knowledge about what is a valid standard already exists in simplecpp and that needs to be leveraged.

hmm you mean to utilize simplecpp::getCppStdString and simplecpp::getCStdString right? That sounds pretty interesting to me however it does not seem trivial directly. Maybe we should split up the simplecpp functions but that should be done upstream of course not here in cppcheck repo.

hmm you mean to utilize simplecpp::getCppStdString and simplecpp::getCStdString right?

Yes, at least the knowledge. I did not have a look yet. IIRC there is also some inconsistencies with the later on usage. I might not have slapped a TODO on everything.

I think this should done incrementally. Start with the CLI parameter bailing out based on the existing logic. Then plumping in the simplecpp logic and afterwards address the project import part (as already mentioned that needs some test coverage - and maybe quite some of it).

I think this should done incrementally. Start with the CLI parameter bailing out based on the existing logic.

yes that is acceptable to me also. we don't have to change the standards.h or importproject at all it's enough with cmdlineparser and testcmdlineparser..

I think this should done incrementally. Start with the CLI parameter bailing out based on the existing logic.

yes that is acceptable to me also. we don't have to change the standards.h or importproject at all it's enough with cmdlineparser and testcmdlineparser..

I suppose this should be done in a new pull request?

I suppose this should be done in a new pull request?

That is up to you. You can keep this and just remove the stuff that is deemed out of scope for the first step.

I created a new pull request.