Re: guix-based installation of pigx-rnaseq - works
Am 26.02.21 um 22:25 schrieb zimoun:
> Hi,
>
> On Tue, 23 Feb 2021 at 21:28, Steffen Möller <steffen.moeller@uni-rostock.de> wrote:
>
>> sudo apt-get install guix
>> guix install pigx-rnaseq
> Thanks Vagrant! :-)
> And if I may, I would recommend this Debconf track:
> <https://debconf18.debconf.org/talks/99-my-crush-on-gnu-guix/>
Watched it! [hence the delayed reply]
>
>
>> We keep mentioning conda also the time, while guix is somewhat left
>> aside, but https://guix.gnu.org/packages/ is truly impressive. Somewhat
>> annoying, if we ever decide to also reference guix packages in
>> d/u/metadata, then there are only versioned web pages like
>> https://guix.gnu.org/de/packages/bc-1.07.1/, so we would only have
>> moving targets point to.
> What do you mean exactly?
In d/u/metadata we give references to the bio.tools registry,
SciCrunch's RRIDs and conda. But we should also give one to guix if we
know about an "ortholog" package. These references point up as URLs on
our task pages for the various blends. There is however no page for a
package in guix that is not versioned if I get this right.
>> I had tried in vain to create a dockerfile with this setup which gets a
>> permission error in the moment that the installation of the packages
>> starts:
> To create a Docker image, instead of going through Dockerfile, you could
> use:
>
> guix pack -f docker pigx-rnaseq
>
> (plus maybe some other options, for reference:
> <https://guix.gnu.org/manual/devel/en/guix.html#Invoking-guix-pack>)
Will try that. Many thanks!
> Well, from my point of view, the issue with this snippet
>
>> FROM debian:unstable
>> ENV TERM=xterm
>> RUN apt-get update -qq
>> RUN apt-get install -y guix
>> #RUN (/usr/bin/guix-daemon --build-users-group=_guixbuild & ) && guix
>> install pigx-rnaseq
>> #RUN (/usr/bin/guix-daemon & ) && guix install pigx-rnaseq # too big to fail
> is that it is not reproducible, first because of ’apt-get update’ then
> second because the Guix version is not tracked. Other said, it will be
> hard nor impossible to rebuild the exact same image.
This was only a pilot that mimics the install instructions for
pigx-rnaseq provided by upstream. Also, a controlled variation is
intentional as we will be using it all together with snaptshot.d.o. The
idea is to have the RNA-seq workflow as a template for other (also
non-biological) workflows to come and then see how results change while
we travel through time. So we want to get away from "does it build and
autotest" to a quantitative comparison - and maybe find other
benchmarks. We'll see.
The fun part is, and this is where this shot on guix comes in, that we
can compare a Debian based implementation with an image that upstream
provides. Or compare with conda/brew/...guix. We are currently team
building and collect ideas what we want to achieve.
> Instead,
>
> guix time-machine --commit=12345 -- pack -f docker pix-rnaseq --save-provenance
>
> seems better, IMHO. However, going via debuerreotype would be another
> option. :-)
>
>
> About the ’vim’ fails, see <http://issues.guix.gnu.org/issue/46642#2>.
> The Guix CI does not have a concept of “stable” or “testing” and time to
> time a Sid commit breaks master. :-)
Hm. Not so sure. My interpretation was some general incompatibility of
above docker-approach with GUIX, so I happily adopt the "guix pack".
Many thanks!
Steffen
Reply to: