guix-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: a GUIX_PACKAGE_PATH / modules puzzle


From: Nils Gillmann
Subject: Re: a GUIX_PACKAGE_PATH / modules puzzle
Date: Thu, 3 May 2018 05:36:13 +0000

Chris Marusich transcribed 3.0K bytes:
> Nils Gillmann <address@hidden> writes:
> 
> > Hi,
> >
> > I'm not sure if guile-users / -dev liste is more appropriate. If it is, let 
> > me know.
> >
> > I'm currently still using GUIX_PACKAGE_PATH until I got my layout all set 
> > up.
> > There's an issue that I can't seem to get rid of, I'll try my best to 
> > describe it now:
> >
> > I have package definitions in 2 repositories (3 to be precise, +1 WIP 
> > repo). Both follow
> > the exact same layout that can still be understood by Guix and GPP 
> > (GUIX_PACKAGE_PATH).
> >
> > Repository 1: /home/user/src/infotropique/ports
> > Repository 2: /home/user/src/infotropique/pkgs
> >
> > The modules are in similar named subfolders, following the layout $REPONAME 
> > $CATEGORY $NAME NAME.
> >
> > I only switched to this recently, previously the "CATEGORY NAME NAME" was 
> > located in the root of
> > the reository. Both repos worked this way.
> >
> > After the transition to the new layout, ports works.
> > pkgs however is throwing errors at me for days. I've tried moving all but 
> > one module
> > out of the repo, same error. It was a simple easy, no repo-internal 
> > dependencies, module.
> > The error is always similar to https://ftp.n0.is/pub/pkgs-error.txt
> >
> > I can not share the actual repo at this moment on this mailinglist, but 
> > maybe the
> > repl message is something that someone could give me a hint, even if it's 
> > just a
> > tiny bit of a "read this manpage" or something like that would help. I've 
> > tried
> > debugging this for days and any help's welcome.
> 
> Could this error be coming from the scheme-files procedure in (guix
> discovery)?  It looks like that's the only place where fold-right is
> invoked in that module, and it seems to match up with the stack trace,
> unless I'm misreading it (which is possible).  You might find it useful
> to insert some pk statements into that code to try to figure out what is
> being evaluated as "unspecified".  My guess, which could be wrong, is
> that perhaps scandir* is returning an unspecified value unintentionally
> at some point?  Someone with more experience debugging Guile code can
> probably provide better advice.
> 
> -- 
> Chris

What I found is that: (pkgs sys-kernel) providing common logics for
(pkgs sys-kernel linux linux) [and planned, (pkgs sys-kernel linux-libre)]
and (ports sysutils firmware some-firmware) as well as (pkgs sys-kernel 
linux-headers)
created a problem. This, moving the definitions from the old linux module I had 
out to
a common module, was one of the few breaking changes.
I wasn't able to resolve this as amirouche told me on irc. I want to remove 
some last
bits of these imports tonight. I was able to update my system with the modules 
in another
place and imports corrected.
What's weird to me is that even after cleaning the ccache of guile in the 
respective
locations a package that isn't involved in the new change, like for example 
numactl,
as the only package definition in this place throws errors. In another location 
it
works. I didn't come up with the name "pkgs" for the root of the repository 
yesterday,
I have used this for a couple of months in a working state.

Anyways, in which section of the guile Manual would I find "pk"? I've done 
surprisingly
little debugging with Guile itself so far.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]