IRC channel logs

2021-11-25.log

back to list of logs

<zimoun>hi!
<zimoun>for French speaker: https://reproducibility.gricad-pages.univ-grenoble-alpes.fr/web/
<civodul>o/
<civodul>too bad it's on Zoom!
<rekado_>at the MDC everything is Zoombified.
<civodul>terrible
<civodul>it would cost all these orgs very little to run a BBB instance
<zimoun>yeah, here at Univ Paris, it is THE official recommendation: use Zoom period.
<civodul>refuse, resist!
*civodul managed to never use it
<zimoun>I did and am doing. Now for team meeting, we use Renater rendez-vous. :-)
<zimoun>And the recent Jami thing done by apteryx seems doable to setup for team meeting. ;-) Argh why the day has only 24h? :-D
<civodul>heh :-)
<zimoun>civodul, if you have some CPU to burn, I would be interested if you build Julia from core-updates-frozen. Since you had an issue, I would like to be sure it is completely fixed. Because it had been a mess. :-)
<civodul>there are substitutes, so everything's fine? :-)
<civodul>i can give it a spin though
<civodul>build started, talk to you in 12h :-)
<civodul>good job anyway!
<zimoun>normally, the long check phase is strongly reduced. On Berlin 5x ;-)
<zimoun>and substitutes was also available previously ;-)
<civodul>oh how did you manage to speed up tests?
<zimoun>yep, I guess.
<zimoun>civodul, apteryx reads the doc, find Nix patch. I adapted and did all the estings.
<rekado_>I’ve never used the zoom application. Only ever through the browser, so I guess that’s using webrtc…?
<zimoun>civodul`, hum is it zoom? https://videos.univ-grenoble-alpes.fr/live/gricad/
<civodul`>dunno, but https://reproducibility.gricad-pages.univ-grenoble-alpes.fr/web/programme_251121.html mentions Zoom
<zimoun>live lost…
<zimoun>Interesting “quote” (roughly translated), by Olivier Richard: it is easier to think about complex system when using functional management. It is not the case for Spack.
<zimoun>«technical scope are vaster than Spack, Flatpack or Snap…»
<zimoun>«60k packages is maybe too much; at least we should ask to keep consistent system»
<zimoun>arf! it takes ages on the cluster for “guix search” by Pierre-Antoine demoing. )-:
<civodul>bah, another thing we should finally tackle
<zimoun>yes and I am still puzzled why my last attempt was slower https://issues.guix.gnu.org/39258#119
<zimoun>Especially that https://issues.guix.gnu.org/39258#127
<zimoun>BTW, I think it would really useful to write down the difference between Nix and Guix; specially about the language.
<civodul>yeah
<zimoun>Excellent answer by Pierre-Antoine: Intel compiler, no! Because it does not make sense in the context of reproducible science. Done. ;-)
<civodul>good :-)
<civodul>our work is mostly to help change social norms
<zimoun>On this seminar, it is about Reproducible Environment, which is already nice. But no question about preservation of source code… weird. Am I biased? ;-)
<zimoun>Oula, the Nix mechanism for channel seems complicated for reproducing later.
<civodul>for Nix i think you can always just grab the Nixpkgs revision you want
<civodul>but you also need a version of Nix that can actually interpret it
<civodul>(since the Nix language evolves over time)
<civodul>this is the main difference compared to Guix
<zimoun>civodul, yeah. Well, from the demo, it appears to me that “guix time-machine” is just missing. And it is hard to guarantee that the Nix code used for building is *exactly* the same. Well, it appears to me they miss inferior, right?
<civodul>i think (but i haven't checked in a while) that (1) the Nix user interface is less convenient and describe+time-machine
<civodul>and (2) the false decoupling between Nix and Nixpkgs may complicate things more often than not
<zimoun>civodul, julia should have finished to build. Does it?
<drakonis>zimoun: they kinda have something like that
<drakonis> https://nix.dev/tutorials/towards-reproducibility-pinning-nixpkgs
<drakonis>you have to fetch a tarball of the repository at a specific commit
<zimoun>drakonis: thanks! It is indeed more complicated than Guix. :-) Well, from my understanding.
<zimoun>How do you have a guarantee that is work? https://github.com/NixOS/nixpkgs/archive/3590f02e7d5760e52072c1a729ee2250b5560746.tar.gz
<zimoun>it can disappear.
<zimoun>ok, probably stored on SWH. But then one need SWHID to effectively look up.
<drakonis>(please note that i'm not even quoting the new way to do it, because its still experimental)
<drakonis>flakes is pretty complicated and uses lockfiles
<drakonis>its a band-aid for existing issues
<zimoun>Often I complain about Guix missing X. Today, I praise The Right Design guix has! :-)
<drakonis>hah
<drakonis>i mean, it is certainly a universal improvement over nix
<drakonis>guix might not have everything (yet) but it is looking to improve itself in meaningful ways