weird code analysis behavior for c++ module

Created October 14, 2025 08:33

I created a sample project that uses traditional forward declarations in a module interface unit with its implementation. The project builds and runs without issues, and Visual Studio's built in code analysis tools report no errors. However, ReSharper’s code analysis reports an ‘ambiguous symbol’ C++ compiler warning (or error).

ReSharper’s IntelliSense also detects two ‘InClass’ types, resulting in ambiguity.

If I switch the IntelliSense engine to Visual Studio’s and turn off ‘code analysis’, the ambiguity issue disappears.

I hope to hear some good news about resolving this issue.

attached sample project at:

Upload ID: 2025_10_14_7bz7i1jrGTPNn8oSMfEdHf

version:

9 comments

Igor Akhmetov

Created October 14, 2025 12:57

Hello,

R# found a bug for you. All declarations of an entity should be attached to the same module, see “Module ownership” on this page. You have InClass attached to the global module fragment and to Module. GCC also reports ambiguous class error - https://godbolt.org/z/r3v41M48a. See also this post - https://abuehl.github.io/2025/03/10/modules-forward-declarations.html.

Jovial92

Created October 14, 2025 22:56

The ambiguity seems to come from the use of the ‘export’ keyword. What is the correct way to forward declare a class declared in a ‘traditional header file’ from a module interface file?

Andrey Davydov

Created October 15, 2025 10:18

Although everything my colleague wrote in the start comment is correct, and this is the reason why the current version of ReSharper gives the error, I am not absolutely sure that the code in the goldbolt example is wrong. I'd say that this question is underspecified in the C++ standard, that's why GCC and ReSharper on the one hand and MSVC and Clang on the other hand treat this code differently. I decided to change the behavior in ReSharper, so in a future version it will match MSVC and Clang behavior.

The ambiguity seems to come from the use of the export keyword.

The `export` keyword is certainly needed, without it all compilers (Clang since the last version 21) agree that the code is wrong: (with export: https://godbolt.org/z/z9c3hTEMP, without export: https://godbolt.org/z/sWG59TE68).

What is the correct way to forward declare a class declared in a ‘traditional header file’ from a module interface file?

Export forward declaration should be wrapped into extern “C++”: https://godbolt.org/z/x74Yo76z8, in the such case declared `InClass` will be attached to the global module and so there won't be any ambiguity.

Jovial92

Created October 22, 2025 03:46

Hi Andrey,

Thank you for the tip about using the extern keyword for forward declarations, which resolved the issue when I forward declared a class within a header file from a module.

However, if you forward declare each other's classes in both a module and a header, ambiguity still occurs.

We use this pattern frequently in our in-house code, so it’s very important that there are no issues when using ReSharper++.

You mentioned, “so in a future version it will match MSVC and Clang behavior.”
Can I expect these changes to also resolve this ambiguity?

Here’s a new sample project located at:

Upload ID: 2025_10_22_MDF1Hqpi3LUf7EsGaKd7VC

Thanks in advance