[PATCH] D140927: [C++20][Modules] Fix named module import diagnostics.

[Iain Sandoe via Phabricator via cfe-commits]

iains added a comment.

In D140927#4024648 <https://reviews.llvm.org/D140927#4024648>, @ChuanqiXu wrote:

> [module.import]p1 says:
>
>> In a module unit, all module-import-declarations and export-declarations exporting module-import-declarations shall appear before all other declarations in the declaration-seq of the translation-unit and of the private-module-fragment (if any).
>
> So the following case is invalid:
>
>   //--- b.hpp
>   void func() {}
>   import a;
>   
>   //--- c.cpp
>   module;
>   #include "b.hpp"
>   export module c;
>
> I feel we'd better to address such cases in the test.

I am not sure about this - what you are suggesting above would seem to make modules useless when there is a GMF (since there is almost no point to having a GMF that has no decls).

https://eel.is/c++draft/basic.link#1

Make the declaration-see of a module unit specifically follow the module keyword.

https://eel.is/c++draft/module#import-1 (which you quoted from above) makes adds to this the PMF - but I do not see that it includes the GMF (since that is neither the declaration-seq of the module unit nor is it the PMF.

(this + 15.5 named Header Unit import is why the original bug is there .. )


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140927/new/

https://reviews.llvm.org/D140927