On Monday, 23 December 2024 at 18:07:45 UTC, Timon Gehr wrote:
> On 12/19/24 17:52, Atila Neves wrote:
>> On Thursday, 6 June 2024 at 15:47:50 UTC, Timon Gehr wrote:
>>> [...]
>> 
>> Why do we need to compile all of them? Do we need something like `__traits(callerEdition)` maybe?
>> ...
>
> It may not be known at library compile time what edition the user will be using, and there may be multiple different users on different editions.

Right, but not all edition-specific branches will need to compile in any one given compiler pass, right?

> Maybe, though it does invalidate memory safety guarantees. Another way to deal with it is to require something like `@trusted import` for such modules or to otherwise treat all `@safe` imports from that module as `@system`.

I like both of those ideas.

>> Maybe there will be changes we won't be able to make.
>> 
>
> Maybe, though I think one thing that will often work is to make such calls `@system`.

Right.

On 12/26/24 10:55, Atila Neves wrote:
>>
>> It may not be known at library compile time what edition the user will be using, and there may be multiple different users on different editions.
> 
> Right, but not all edition-specific branches will need to compile in any one given compiler pass, right?

Well, I think if a library wants to branch on the edition of the caller, all of those branches would need to be compiled when the library is compiled. (Think druntime.)

Of course, the case where a library wants to branch on its own edition is separate. There, not all branches will need to compile on any given compiler invocation, though I am not sure that this is an important case (as usually every piece of code just targets a specific edition).

On Saturday, 28 December 2024 at 02:03:57 UTC, Timon Gehr wrote:
> On 12/26/24 10:55, Atila Neves wrote:
>>>
>>> It may not be known at library compile time what edition the user will be using, and there may be multiple different users on different editions.
>> 
>> Right, but not all edition-specific branches will need to compile in any one given compiler pass, right?
>
> Well, I think if a library wants to branch on the edition of the caller, all of those branches would need to be compiled when the library is compiled. (Think druntime.)
>
> Of course, the case where a library wants to branch on its own edition is separate. There, not all branches will need to compile on any given compiler invocation, though I am not sure that this is an important case (as usually every piece of code just targets a specific edition).

Ok. What do you propose we do, then? Because AFAICT none of the current mechanisms will work for this.

On 12/30/24 16:49, Atila Neves wrote:
> On Saturday, 28 December 2024 at 02:03:57 UTC, Timon Gehr wrote:
>> On 12/26/24 10:55, Atila Neves wrote:
>>>>
>>>> It may not be known at library compile time what edition the user will be using, and there may be multiple different users on different editions.
>>>
>>> Right, but not all edition-specific branches will need to compile in any one given compiler pass, right?
>>
>> Well, I think if a library wants to branch on the edition of the caller, all of those branches would need to be compiled when the library is compiled. (Think druntime.)
>>
>> Of course, the case where a library wants to branch on its own edition is separate. There, not all branches will need to compile on any given compiler invocation, though I am not sure that this is an important case (as usually every piece of code just targets a specific edition).
> 
> Ok. What do you propose we do, then? Because AFAICT none of the current mechanisms will work for this.

I am not sure.

One simple idea that came to mind would be to add an `__EDITION__` special identifier that is evaluated in the caller's scope, similar to `__LINE__` and `__FILE__`, but that seems like only a partial solution.

Perhaps what would work better would be a visibility attribute that says "only visible from this edition/these editions".

Then you can e.g. have different declarations and alias them using aliases that have the correct visibility attributes.

Of course, interoperability is a concern, but that would be up to the library author.