Re: Remove sphinx from native inputs

[jgart]

On Wed, 15 Sep 2021 21:55:37 +1000 Ben Sturmfels <ben@sturm.com.au> wrote:

Hi Ben,

> celery`, since they can't see the inputs.

For now, you could just add python-celery to the propagated inputs if you're
calling the binary directly from the environment somewhere. That might be the
right approach in this case.

Raghav, wdyt?

python-celery is used as both a library and a program:

```
$ tree $(guix build python-celery)

/gnu/store/21rfamirv6kr736jps7nkgmd2zqqinw4-python-celery-4.2.1
├── bin
│   └── celery   <--- executable
```

See meld for a python application that uses all three types of inputs:

`guix edit meld`


> I think this is a problem with MediaGoblin itself though. Ideally the
> `gmg` command would serve as the single entrypoint here, eg. `gmg serve`
> and `gmg celery` but these have never been implemented.

Would you use celery as a library at that point? If you still need
`/gnu/store/.../bin/celery` then you might still need celery to be available as
a propagated-input. I would just test it out when the time comes.

> This problem is becoming clearer to me now, and I've learnt more about
> paster, so I need to hack on it for a bit.

Cool, keep us posted on it :)

> Ultimately, I think you're right though - we should use "inputs". Just
> not *quite* yet. Leave it with me.

You might still require propagated-inputs for things like celery. See above.

We're looking forward to mediagoblin's next release and our guix adventure of
writing a service. I'll let you know soon about the possibility of having a
mediagoblin deploy meetup. That would be very useful for us in this endeavour.

all best,

jgart

 
 _________________________________________ 
/ 3B1D 7F19 E36B B60C 0F5B 2CA9 A52A A2B4 \
\ 77B6 DD35                               /
 ----------------------------------------- 
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||