I've scheduled this for creating of a core issue:

https://github.com/cplusplus/CWG/issues/324

Jens


On 24/05/2023 11.21, 许传奇(夜灯) wrote:
> Hi,
>
>> Any specific suggestions here?
>
> I feel the statements in the mail is pretty clear (maybe we need to rewrite it to make it looks like standard more):
>
>> Header units do not depend on the preprocessor state at
> the location where the "import" appears.
>
> ---
>
>> Note that this says "implementation-defined, with a minimum supported set".
>>
>> So, your first go-to point to determine that set is your vendor documentation,
> not the standard.
>
> Yeah, my logic here is that both you and Tom think the header "bar.h" should not be importable without talking about specific implementation according to a paragraph in the standard. So there should be a statement in the standard to make it clear. Also, personally, "implementation-defined" is not completely equal to "we don't care what headers can be importable". For example, [module.import]/p6 (http://eel.is/c++draft/module.import#6 <http://eel.is/c++draft/module.import#6>) says:
>
>> A header unit shall not contain a definition of a non-inline function or variable whose name has external linkage.
>
> So my interpretation is "although we don't define what importable header is, we still have some rules for header units."
>
> Thanks,
> Chuanqi
>
>
> ------------------------------------------------------------------
> From:Jens Maurer <jens.maurer_at_[hidden]>
> Send Time:2023 May 24 (Wed.) 17:07
> To:sg15 <sg15_at_[hidden]>; Tom Honermann <tom_at_[hidden]>; Boris Kolpackov <boris_at_[hidden]>
> Cc:Chuanqi <chuanqi.xcq_at_[hidden]>
> Subject:Re: [SG15] Questions about wording and design intention for header units and ideas
>
> On 24/05/2023 08.45, 许传奇(夜灯) via SG15 wrote:
> > Thanks for your explanation. Your words are self-contained. But I can't see it from [module.import]p5. I don't feel the semantics are stated clearly. If it is the intention, I strongly feel we need to reword it.
>
> Any specific suggestions here?
>
> Note that this says "implementation-defined, with a minimum supported set".
>
> So, your first go-to point to determine that set is your vendor documentation,
> not the standard.
>
> Also note that header units are intended as a transitional feature, thus
> giving a bit more leeway to the implementation in that area might be a
> good idea.
>
> > And the general headers were meant to be importable although we doesn't define it well-formed in the standard. But given your interpretation, the usage of header units become less usable.
>
> Header units do not depend on the preprocessor state at
> the location where the "import" appears.
>
> If that's not what you thought the feature would deliver, please
> review the papers leading to the current design in the standard,
> and feel free to submit a paper seeking improvements to EWG.
>
> Jens