Naming of native-inputs/inputs/propagated-inputs

Naming of native-inputs/inputs/propagated-inputs

[Ben Sturmfels]

[-- Attachment #1: Type: text/plain, Size: 852 bytes --]

Hi Folks,

On the rare occasion I dip into Guix packaging, I rely on Pierre's
tutorial[1] because it has the least confusing definition of
"native-inputs" "inputs" and "propagated-inputs". Before that tutorial
existed, I don't think I properly understood the difference.

I know this is superficial, but I wonder if these names could be refined
to better communicate their use. What about:

| Currently         | Possible new name |
|-------------------+-------------------|
| native-inputs     | build-inputs      |
| inputs            | runtime-inputs    |
| propagated-inputs | profile-inputs    |

Just a thought. May even "build-dependencies"? Still doesn't explain
that profile-inputs are also included as build and runtime inputs. Maybe
"build-only-dependencies"?

Regards,
Ben

[1] https://guix.gnu.org/blog/2018/a-packaging-tutorial-for-guix/

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

Re: Naming of native-inputs/inputs/propagated-inputs

[Ricardo Wurmus]

Hi Ben,

> I know this is superficial, but I wonder if these names could be refined
> to better communicate their use. What about:
>
> | Currently         | Possible new name |
> |-------------------+-------------------|
> | native-inputs     | build-inputs      |
> | inputs            | runtime-inputs    |
> | propagated-inputs | profile-inputs    |

The difference between “native-inputs” and “inputs” is not about whether
they are needed at build time or runtime.  The difference only really
comes into play when cross-building.  “native-inputs” must be of the
host architecture, while “inputs” are for the target architecture.

It is correct that “native-inputs” often happen to be only used during
the build, but that doesn’t need to be the case.

Ultimately, runtime inputs are determined by whether they are referenced
in the output.  This is orthogonal to whether they are listed in the
“native-inputs”, “inputs”, or even “propagated-inputs” field.

I empathize with the desire to remove sources of confusion, but changing
the names would blur this distinction for a superficial orthogonal
property and give the wrong impression about how these inputs are used.

--
Ricardo

Re: Naming of native-inputs/inputs/propagated-inputs

[Ben Sturmfels]

[-- Attachment #1: Type: text/plain, Size: 774 bytes --]

On Tue, 23 Jul 2019, Ricardo Wurmus wrote:

> The difference between “native-inputs” and “inputs” is not about whether
> they are needed at build time or runtime.  The difference only really
> comes into play when cross-building.  “native-inputs” must be of the
> host architecture, while “inputs” are for the target architecture.
>
> It is correct that “native-inputs” often happen to be only used during
> the build, but that doesn’t need to be the case.
>
> Ultimately, runtime inputs are determined by whether they are referenced
> in the output.  This is orthogonal to whether they are listed in the
> “native-inputs”, “inputs”, or even “propagated-inputs” field.

Thanks for the clarification Ricardo!

Regards,
Ben

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]