IRC channel logs


back to list of logs

<zimoun>rekado: about GWL, I am doing ‘guix environment -C -L channel --load=.guix.scm’ then ./bootstrap && ./configure && make. And I do not understand why "/usr/local/share/guile/site/3.0" appears in guix/extensions/workflow.scm. Any idea about what I am doing wrong?
<rekado_>‘/usr/local’ is the default prefix.
<rekado_>when you use ./configure without any arguments that’s the installation prefix.
<rekado_>guix/extensions/ contains placeholders for @guilemoduledir@ and @guileobjectdir@
<rekado_>these are defined in terms of the prefix
<rekado_>since the pre-inst-env script sets GWL_UNINSTALLED these locations are ignored, though
<rekado_>when the GWL is installed, however, the (source and compiled) modules from these locations are put on the load path.
<zimoun>rekado_: thanks. These ’configure’ are and will always be mysterious for me. Thanks for the explanations, make sense.
<zimoun>civodul: thanks for Docker. Now I am trying to put the initial motivation in a blog post or cookbook recipe. Let decide once written down. :-) I have started yesterday envening then hit bug#46390 and fix it with patch#46391.
<civodul>hi zimoun!
<civodul>hey, i guess i haven't seen those yet :-)
<zimoun>civodul: well, ‘guix lint -L example foo’ fails uglyly but not ‘guix lint -L /tmp/example foo’ or ’guix lint -L . foo’ with ’foo’ a custom package with “bad” Git source. Mainly because (search-path %load-path "example/my-pkg.scm") return #f i.e., %load-path contains ("example” "/gnu/store/…"). Anyway, in
<zimoun>short, another corner case. These days, I «have the chic» to find them. ;-)
<civodul>looking at the "Perspectives" section now
<civodul>not super motivated but hey!
<civodul>oh we didn't write about the R.install/guix module you worked on, rekado_
<zimoun>civodul: thanks for the better wording. :-)
<zimoun>one perspective mothacehe discussed yesterday (and we are discussing it since last year ;-)) is Docker images as a service. User provides a manifest+channel and the instance returns a link you can use for “docker load”, for example.
<zimoun>I do not know.
<civodul>yeah pack-as-a-service is something we keep talking about
<civodul>but it's too "low-level" IMO for the "Perspectives" section
<civodul>we want to dream, to aim high! :-)
<rekado_>the guix.install thing for R is pretty much ready, but it isn’t part of our R package yet.
<rekado_>I could write something about it, but where would it go and where would I publish that file?
<civodul>i don't know :-)
<civodul>i thought about it as i was writing the perspectives in the annual report
<civodul>as an example of a tool that pushed reproducible deployments "in the hands" of scientists
<rekado_>I’ll put it up on first; we can move it later if needed.
<rekado_>is it okay if I do this in an hour or two?
<rekado_>I’ll also add the GWL announcement to the HPC blog then
<civodul>sure, there's no rush after all
<civodul>my plan is to publish the report early this afternoon
<civodul>(i have a meeting at 4PM, so i'd do this before)
<civodul>so... your call
<civodul>we can also keep it for next year's report ;-)
<zimoun>I am back. :-)
<zimoun>civodul: well, for me pack-as-service is a dream :-)
<rekado_>the R thing is now here:
<rekado_>what do you think of this:
<rekado_>I’d add it to the bottom of the Packaging section
<civodul>rekado_: LGTM!
<civodul>you need to use a literal em dash instead of "---" tho
<zimoun>rekado_: I am always confused by “trivial” and I generally prefer “naturally”. Anyway, French pernickety guy :-) (I open the dictionnary for pernickety ;-))
<rekado_>I use “trivially” here to mean “like, super easy, don’t need to think about it”
<rekado_>I think it’s appropriate here because the feature here is to use importers in the background without involving the user at all
<civodul>rekado_: thanks!
<civodul>i'll go ahead and publish
<civodul>hmm pdfbook is not putting pages 1 and 22 on the same page
<civodul>alright comrades! we did it!
<civodul>(this "we did it" is not quite comparable to janneke's tho)
<rekado_>civodul: woo, thanks!
<civodul>chances are too few people will notice and read
<zimoun>civodul: what do you mean? Do you mean we should send an email to guix-science?
<zimoun>rekado_: yeah, it remembers my math teacher saying: this proof is trivial; but with my classmates we had not any idea how to prove it. And later, we often answered: that’s trivial because it were so obvious and the teacher said: no, it’s not! Well, since then, I am always suspicious when I read “trivial”. :-)
<janneke>that's quite an impressive achievement, folks!
*janneke also likes the (past autotools); been pondering on something like that for the bootstrap
<rekado_>civodul: I already forwarded it to all MDC cluster users
<civodul>zimoun: oh right, do you want to do it?
<civodul>i posted it on Inria's reproducible-research Discourse (!) instance
<civodul>speaking of which, there are clever things in Discourse
<civodul>you get "badges" when you follow the etiquette for instance
<civodul>people can see how long you've been around, etc., so they know "where you come from"
<civodul>(on a mailing list that's insider knowledge, structurelessness and all)
<zimoun>civodul: well, as people already suggested, that’s just the GNU mailman is really old. New mailman looks like that:
<zimoun>maybe one solution should to push for an upgrade of the GNU mailman instance and address the blockers. I do not know. I am not enough annoyed to dive in. :-)
<zimoun>rekado_: I have seen that GWL had not been announced on guix-science. Maybe, it is worth. I do not know who is reading but who knows. ;-)
***ChanServ sets mode: +o civodul
***civodul changes topic to 'GNU Guix in high-performance computing | this channel, it is logged: | activity report: | FOSDEM: |'
***ChanServ sets mode: -o civodul
<rekado_>zimoun: I’ll prepare the post for the HPC blog; then we can mention it on guix-science.
*rekado_ dusts off Guile-DRMAA code
<rekado_>looks like submitting a job actually worked
<rekado_>but there are also very weird other behaviors
<rekado_>like part of an error message being interpreted as a command to the scheduler… worrying
<rekado_>segfault with run-job but success with run-bulk-jobs
<rekado_>it’s a good start :)
<civodul>heh :-)
<civodul>sounds promising!
<rekado_>fixed the segfault. Still have that weird error-message-as-command thing going on, but perhaps I’m just using the wrong buffer variabl somewhere.
<civodul>it reminds me of that error that people now get (rarely, but still):
<civodul>error: substitute: Backtrace:
<civodul>(that's it)
<rekado_>oh, I remember that
<civodul>i can't help but read it like "guile drama"
<civodul>looks pretty cool already