C++ Logo


Advanced search

Re: [Tooling] Modules feedback

From: Matthew Woehlke <mwoehlke.floss_at_[hidden]>
Date: Thu, 14 Feb 2019 15:56:29 -0500
On 14/02/2019 13.55, JF Bastien wrote:
> I'll remind you the context: your complaint that you have to maintain
> headers. I give you the obvious solution, which EWG discussed quite a few
> meetings ago.

Aside from it not really being "my" complaint, it doesn't seem to have
gone anywhere; it just turned up again yesterday on ext¹.

> If you want to call headers "PMIR" and claim novelty, go for
> it. Or are we playing some sort of game now?

Um... Let's start over.

Some folks that think modules will "replace headers"¹. We also know that
a) not all builds are monolithic, b) BMI's are non-portable and c)
shipping complete source code is not desirable.

As I have been attempting to explain in the last several messages, a
consequence of this is that, in order to package and distribute a
library which uses IS modules, I also have to package and distribute
something (a "PMIR") which compilers can consume when they see an 'import'.

The two main questions I've been trying to address are:

- Do we agree that "libraries which can be packaged and distributed in
non-source form" and "unified interface and implementation" are, absent
some PMIR (which *may be* "headers"), contradictory goals?

- Is there any consensus on what this "PMIR" should look like?

Now, I'm fine with the answer to the second being "header files". If
that's the answer, though, and if we also want to make the "unified
source file" crowd happy, we'll need a way to generate them from the
unified sources. (I'm not sure if that's "novel", but at least I am not
aware of any existing implementation that already does this.)



Received on 2019-02-14 21:56:33