Re: should gnome-desktop-service really provide all of this to a profile?

[Ludovic Courtès]

address@hidden (宋文武) skribis:

> address@hidden (Ludovic Courtès) writes:
>
>> Pulseaudio also propagates a few things.
> Yes, I think it should be splited into “out” and “dev”, if we can make
> only the “dev” output propagating libcap and gdbm.

Yeah.

>> Two things here:
>>
>>   • Back in the day, I couldn’t think of a situation where it would make
>>     sense for ‘propagated-inputs’ to be different from
>>     ‘propagated-user-env-packages’ (at the time, the latter was little
>>     known so in practice people had to install dependencies by
>>     themselves.)  So in Guix I chose to have just one.
> In nix ‘propagated-build-inputs’ is propagations and ‘inputs’ for building
> while ‘propagated-user-env-packages’ is propagations for profile.
> In guix ‘propagated-inputs’ is propagations for both building and profile
> and at the same time is ‘inputs’ for building.

Yes.

> I agree that propagations for building or for profile is the same thing,
> and there’re some runtime only dependencies worthing treating as
> non-inputs (eg: adwaita-icon-themes, plugins for gstreamer).

Yes, though I would not worry about these “non-inputs” for now.

> How about seperating propagations from ‘propagated-inputs’?
> This requires listing some packages twice, but I think it will be more
> clear considering ‘inputs’ are for building the whole package while
> ‘propagations’ are (should be) per-output.
>
> eg:
> (package
>   (name "pulseaudio")
>   (outputs '("out" "include" "lib" "dev"))
>   (inputs
>    `(("libpcap" ,libcap)
>      ("gdbm" ,gdmb)
>      ...))
>   (propagations
>     #:output "dev"
>     `(("pulseaudio:include" ,pulseaudio "include")
>       ("pulseaudio:lib"     ,pulseaudio "lib")
>       ("libpcap" ,libpcap)
>       ("gdbm" ,gdbm)))
>   ...)

Sure.  There’s a years-old example in ‘TODO’, even.  :-)

The syntax cannot be exactly like what you suggest, but it could be
along the lines of what’s in ‘TODO’:

  (propagated-inputs
    `(((("i1" ,p1 "o1")
        ("i2" ,p2))
       => "include")
      ("i3" ,p3)))

>>   • I found the ‘nix-support’ trick (that is, having high-level
>>     information on the build side) kind of hacky, which is why this
>>     information is solely on the build side in Guix.  This is what
>>     allows higher-level functionality such as --search-paths to be
>>     implemented on the host side.
>>
>>     OTOH, things like <http://bugs.gnu.org/22138> could be more easily
>>     addressed if all the info was already on the build side.
> Agree :-)

With which part?  :-)

>> Anyway, what do you think would be the best way to avoid “profile
>> pollution” with the GNOME meta-package?
> I think, for now:
>
> From 7206310f7320ed99eabd7f774a083c7b1f78c81f Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?=E5=AE=8B=E6=96=87=E6=AD=A6?= <address@hidden>
> Date: Wed, 9 Mar 2016 13:17:48 +0800
> Subject: [PATCH] gnu: nautilus: Don't propagate gtk+.
>
> This reduces "profile pollution" of the 'gnome' meta package.
> See <http://lists.gnu.org/archive/html/guix-devel/2016-03/msg00283.html>.
>
> * gnu/packages/gnome.scm (nautilus): Move gtk+ from propagated-inputs to 
> inputs.

Please move the sentence “This reduces …” to a comment in the code.
Otherwise LGTM!

Thanks,
Ludo’.