IRC channel logs

<rekado_>just read this in a presentation for our lab: “patent is critical for further grants”

<civodul>that's wrong though? (i hope?)

<rekado_>civodul: it’s sobering.

<rekado_>it’s really uncomfortable to feel the need to change a project so that a patent can come out of it just to be allowed to continue working on it.

<rekado_>it’s disappointing to see this in the context of a pure research institute.

<civodul>it's even outrageous

<civodul>but do EU FP calls even encourage that?

<civodul>or is it just misunderstanding on the part of the folks speaking?

<zimoun>what appears to me bad, it is that I am earing similar “message” here in France.

<civodul>i haven't noticed (yet?)

<zimoun>maybe it is when “medical doctor” are around :-P

<rekado_>yes, I think it’s something to do with whatever touches on clinical uses

<rekado_>this project is about detecting and classifying heart attacks from blood samples; and it looks like funding sources are exhausted unless it produces “IP”

<rekado_>the MDC has a department for “technology transfer” (= commercialization), but I always assumed I could ignore them.

<efraim>zimoun: I'm checking out your julia patches but I ran 'guix gc' recently. I forgot how slowly julia packages can build sometimes

<civodul>rekado_: it's a general trend (we have an almighty "technology tranfer" dept., which pours public money in startups, including startup projects that have nothing to do with the institute!)

<civodul>i think we should regain control of our institutes

<zimoun>efraim, thanks. Now, the Julia build-system honnors --max-jobs.

<zimoun>so it should be a little faster.

<zimoun>Although the series implies a Julia-world rebuild. :-)

<zimoun>efraim: thanks for the quick Julia review.

<efraim>zimoun: it was either that or auditing my own go packages :)

<zimoun>:-)

<civodul>zimoun: had we agreed that we can remove the deprecation warning for "guix hash -r", and keep "-r" as an alias?

<civodul>i vaguely remember we discussed it, but i'm not sure what the outcome was

<zimoun>civodul, yes we agreed. :-)

<zimoun>the short alias only, IIRC.

<civodul>alright

<civodul>i'll make that change, then!

<zimoun>civodul: you splitted the Café Guix by year, for my imformation, why?

<civodul>as mentioned on mattermost, it's mostly that i realized i was sending an announcement for a 2022 event at a 2021 URL

<civodul>that looked confusing

<civodul>(and since nobody was acting, i did what sounded good to me :-))

<zimoun>heh! It makes sense. I missed that the URL contains the year. :-)

<zimoun> (sorry I do not follow too much mattermost :-)

<zimoun>civodul: a beautiful example of broken world :/ https://lists.gnu.org/archive/html/guix-devel/2022-01/msg00164.html It means that ~2000 of packages from 2020 are lost…

<drakonis>ah, its all being memory holed isnt it

<civodul>zimoun: uh, maybe not all hope is lost

<zimoun>civodul, if you have a plan, I am all ear :-)

<civodul>but yeah, the chances of losing source keep decreasing

<drakonis>try pulling from swh?

<civodul>zimoun: the plan in this case is to look for the original tarball on machines we have access to

<civodul>no magic!

<zimoun>civodul, nckx already did, I guess.

<zimoun>nckx is the magic ;-)

<civodul>without success?

*civodul tries

*nckx no magic :(

<nckx>I'm trying to think of old forgotten machines I might have access to, but no.

<drakonis>can't you fetch it from a git mirror?

<nckx>Git mirror of tarballs?

<drakonis>attempt to acquire the tarball off a commit

<nckx>I am not going to even try making a bit-identical .tar.xz based on the supposed git checkout that produced the original.

<drakonis>the particular file server that imagemagick's tarball only stores the latest minor release

<drakonis>up to the latest

<nckx>drakonis: That's not what I thought: https://download.imagemagick.org/ImageMagick/download/releases/

<nckx>Unless they have some very specific versioning scheme.

<drakonis>it got relocated to releases

<nckx>I don't understand what you're trying to say.

<drakonis> https://sunsite.icm.edu.pl/packages/ImageMagick/

<drakonis>oka

<drakonis>one sec i'm being distracted

<nckx>Sunsite is just a mirror (if it ever goes away I will be sad and the Internet will have officially died, but still).

<drakonis>if you look into the file directory i linked, it always stores the latest minor version of the major releases

<drakonis>older releases get moved to a different directory

<nckx>But are you sure this is relevant to the problem?

<mbakke>I found a copy in a CentOS source tarball and uploaded it here: https://gexp.no/ImageMagick-6.9.10-68.tar.xz

<drakonis>its really not, sadly.

<nckx>drakonis: *You* linked the top-level directory. I linked /releases.

<drakonis>since you can't amend the time machine

<nckx>mbakke: Sweet!

<drakonis>that also helps.

<nckx>I found the source rpm but I didn't realise they contained the full .tar.xz (I don't even know how to unpack an .rpm).

<drakonis>srpms are good.

<nckx>(How do you? I'd rather not install rpm, but I will, if it's for good.)

<nckx>Can I just string together a series of obsolete archivers like cpio and ar together instead, though?

<drakonis>^ this is how you do debian though

<mbakke>nckx: I have access to a RHEL workstation and just rpm --install'd it, I assume cpio would do the job too

<drakonis>but the process is the same

<nckx>drakonis: .deb was the only one I'm vaguely familiar with. The last RPM distro I used was literally a box, filled with literal floppies, from the literal local library. If you remember all 3 things: high five, your childhood was also rad.

<zimoun>mbakke: or rpm2cpio, I do not know.

<nckx>Isn't that just part of rpm tho'.

<drakonis>its cpio, yeah.

<drakonis>rpm2cpio /location/of/package.src.rpm | cpio -d -i

<nckx>Anyway, I feel silly now, having found the srpm early and ignored it. I'll go sit in the silly corner.

<nckx>With the hat yes.

<drakonis>hah

<nckx>drakonis: There's no rpm --extract? Weird? No?

<drakonis>dont worry about it

<drakonis>no there isnt

<drakonis>i just tried to invoke it

<nckx>That's a bit odd.

<drakonis>so, i'm building a nix daemon from scratch

<mbakke>drakonis: in guile? ;-)

<drakonis>ha, i'll look into doing it in guile

<drakonis>i'm fiddling with common lisp right now

<drakonis>i'll work into the guile version first then get back to the cl rewrite

<nckx>You still have more than 2.5 months to announce a Common Lisp Guix daemon…

<drakonis>ha

<drakonis>is it still a worthy goal to pursue doing it in cl?

<drakonis>also that's an oddly specific timeframe :v

<drakonis>what's in 2.5 months?

<nckx>(1 April…)

<drakonis>ha

<nckx>:)

<drakonis>i mean, there is value to doing it in cl

<drakonis>its just going to be less maintainable i guess

<nckx>It's closer to Guile.

<nckx>I guess.

<drakonis>in a way.

<zimoun>Maxime provided the link: guix download https://git.centos.org/sources/ImageMagick/c7/ed27964d872abc81dc2388d333874766058a9b5d

<drakonis>its a lot better at handling failure

<nckx>Then someone will rewrite yours in Racket, and we'll asymptotically approach a Guile daemon.

<drakonis>there's a different route though

<drakonis>there's someone writing a scheme using cl

<nckx>Is bad exception handling in the current daemon a problem?

<drakonis>it'd make for a far more maintainable option

<drakonis> https://gitlab.com/mbabich/airship-scheme

<drakonis>best of both worlds i tell ya

<nckx>Why CL at all?

<drakonis>because its development/debugging experience is just best in class

<nckx>I don't believe C++ is any less maintainable than Guile. It's less maintainable by a bunch of ragtag Guile heads, but CL will suffer the same fate.

<nckx>OK.

<drakonis>also restarts are very useful

<drakonis>i'm not much of a fan of having to restart a build because of a error

<drakonis>being able to directly work on the failed environment and resume is very useful

<drakonis>there are other things i wanted to do with a cl based guix though

<drakonis>but i suppose it might still be doable with guile, at great cost to guix itself

<drakonis>i'll admit that i feel kinda bad to be talking about this

<drakonis>because you all have been nothing but excellent to me

<drakonis>by the way, how exactly do you daemonize something in guile?

<drakonis>do you just run it on loop and add all the signals and such?

<drakonis>signal handling, i mean.

<civodul>efraim: BTW, Python 2.4 in Guix-Past fails to build: https://guix.bordeaux.inria.fr/build/3770/log/raw