C++ Logo

sg15

Advanced search

Re: Exploring another alternative for Distributing C++ Module Libraries

From: Daniel Ruoso <daniel_at_[hidden]>
Date: Fri, 11 Feb 2022 18:32:33 -0500
On Fri, Feb 11, 2022, 18:22 David Blaikie <dblaikie_at_[hidden]> wrote:

> but it's not like we can actually use the linker search path resolution,
> can we?
>

The behavior of the linker is defined. In POSIX it is actually
standardized. And there is plenty of prior art of things external do the
linker reproducing the same search[1].

But perhaps more importantly, the thing that makes this proposal viable,
imho, is precisely the fact that we don't need to introduce a new search
path mechanism at all, which means we don't need to get all package
managers to converge before we do it.

Now, that isn't to say that the interface units themselves need to be in
the library directory, they can still be deployed in whatever location, and
the metadata file will point at them directly.

[1] the way we use pkg-config with CMake is by parsing the package config
files ourselves, finding the libraries and mapping them to individual
imported targets in CMake.

Received on 2022-02-11 23:32:45