[bug#54216] [PATCH staging] gnu: shepherd: Build it from git, and clean up shepherd-for-guix.

[Maxime Devos <maximedevos@telenet.be>]

[-- Attachment #1: Type: text/plain, Size: 2205 bytes --]

zimoun schreef op wo 02-03-2022 om 10:14 [+0100]:
> Hi Attila,
> 
> On Tue, 1 Mar 2022 at 20:42, Attila Lendvai <attila@lendvai.name> wrote:
> 
> > unfortunately, it won't help us much here, because one snippet modifies
> > Makefile.am, while the other Makefile.in.
> 
> Guix does not have a clear line for patching (or I am not aware of the
> update); what is going to source+snippet vs arguments+phase.  Pros and
> cons for both; basically the question is what "guix build --source"
> should return?
> 
> Option source+snippet means it returns the source of what Guix really
> builds

Seems like a pro for source+snippet (or source+patch, though that
would be a bit more verbose) to me.

>  -- so many packages would not respect this rule of thumb.

In that case, it seems like there are plenty of package definitions to
improve!

> Option arguments+phase means it returns the real unmodified upstream
> source (modulo removal of non-free) -- so "guix shell -D foo" would
> break for many packages.

I assume you meant "guix build --source foo"?
This seems like a con for "phases+arguments" to me.
Sometimes, to hack on software, I download the source code with
"guix build --source the-package", unpack it and do
"guix shell -D the-package".

I don't see much value in returning the unmodified upstream source.
Especially since in this case the modified source fixes a bug
(well, works-around a Guile bug). As long as it's source code, it
builds, it doesn't do things like bundling, including binaries or
non-free things, and it avoids being Guix-specific and fixes known
bugs, it seems good source code to me.

Also, most packages don't modify upstream code, so I don't see
the ‘would break for many packages’ here ...

> Difficult tension. :-)

As implied from my explanations above, I don't see any tension here.

> That's said, personally, in this case, instead of having the Makefile*
> patch in 'source', I would do the patching using a phase.

It's ‘merely’ setting some compilation flags, so maybe.
Both options seem fine to me here but I don't see a point to _moving_
from the snippet-shed to phase-shed.


Greetings,
Maxime.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 260 bytes --]