IRC channel logs
2018-02-08.log
back to list of logs
<jsierles>Just getting back into guix again, but forgot why installing python3 won't install a symlink for 'python'. I vaguely remember there's a package to make that happen. <Apteryx>eh, thanks. For some reason I was looking at guix/guix/tests and there was only 2 files there :) <Apteryx>I'm looking at wrap-program in build-utils.scm. <lfam>Apteryx: There are also GuixSD system tests in gnu/tests <lfam>Heh, looks just like the Boston red line <Apteryx>Eh! I should visit Boston again. Libreplanet seems a valid excuse. <lfam>I should try to go this year <Apteryx>I'm not sure if I'll be able to make it *this* year, but if not, I'll go next year for sure :) <Apteryx>for a software named Mocker.el, do we name it 'emacs-mocker.el' or 'emacs-mocker-el' in Guix? <atw>probably emacs-mocker, as that looks like it follows the existing naming convention. The manual might have something to say on the topic <efraim>U-boot seems to be working fine, I'm not sure about having it load grub though <janneke>...emacs memory exhausted -- wasn't that a trivial bug? <efraim>I copied rEFInd aarch64 to /boot/efi/EFI/boot and I got a boot loop since it couldn't find the framebuffer <mbakke>Great, the scipy tarball on pypi does not have "runtests.py". <mbakke>Oh wait, `scipy.test()` still works. <Apteryx>civodul: when building X, is X always implicitly added itself to the 'inputs'? <mbakke>Allright, numpy and scipy updated, now only pandas left. <Apteryx>sneek: later tell atw thanks (will use emacs-mocker) <mbakke>Heading out for a bit, but will continue later today :) <balder[m]>mbakke, did you make -march-native easy to do for those? :) one of those factor of a 100in difference packages <civodul>i've heard good things about this blog post <balder[m]>gcc6 should get more attention! news to me its not only intel doing fat binaries :D <balder[m]>or what seems to be happening, usable defaults not limited to one-off uses and examples for most users <civodul>balder[m]: you mean defaults for compile flags? <balder[m]>civodul, not just that, but can generally expect the need and the existence of exotic hardware that libraries won't always be able to take full advantage of without hand-optimization. <rekado>mbakke: ISTR that these Python packages have a smaller set of tests that are to be used by packagers (whereas the full set is used by developers). <rekado>it may make sense to use the smaller test sets in the future. <rekado>not all test failures are bad enough to justify a failing bulid. <civodul>balder[m]: i think that's an overstatement, but the blog post above summarizes the situation, if you haven't read it <civodul>given that we more or less have that now :-) <rekado>yes, I think we won’t need that project anymore <civodul>good thing that we can remove items :-) <civodul>did you push that Bash implementation somewhere already? <jlicht>rekado: why did Rutger work on gash? Was it part of some project or just for fun? <jlicht>and gash seems very far along, actually. Impressive stuff :o <janneke>jlicht, rekado: Rutger took up work on gash's pipeline this week, expect some progress soon <civodul>janneke: that's awesome, say "hi" and "thanks" to Rutger! :-) <civodul>i just tried it and it seems to work well, even things like jobs <civodul>it doesn't seem to support $? and $$ though, but i wonder if this is bad interaction with readline or something <janneke>ACTION wonders about potential users 0..46 <Inops>Hi all. I'm trying to install GuixSD in a qemu VM, and I've got to the point in the manual where it tells you to run "herd start cow-store /mnt". The machine replies "herd: service 'cow-store' could not be found". Any idea of what is causing this? <jlicht>Inops: what does `herd status' show? <Inops>jlicht: A list of the services enabled, about 35 of them. "user-homes" is stopped. <jlicht>Inops: how did you boot the VM? Because it might be easier to just use guix to build a vm image via `guix system vm-image' <Inops>OriansJ: yes, looks very similar. I used the QEMU image listed on the Downloads page though though, rather than the straight ISO, following the instructions about installing it in a VM (which links you to follow the general installation instructions). I imagine this is something to do with my problem? <OriansJ>Inops: the QEMU image is already setup. You need only update the configuration.scm and apply it to the system <Inops>OriansJ: ah...ok. The maunal (which is linked underneath the VM image) isn't clear: it says "You’re now root in the VM, proceed with the installation process. See Preparing for Installation, and follow the instructions." <OriansJ>Inops: I believe that is a bug we need to correct to prevent future similiar confusion. Thank you for bringing this to our attention Inops <efraim>I think I got the pine64 booting, I copied /boot/efi/EFI/grub/grubaa64.efi to /boot/efi/EFI/boot/bootaa64.efi and I got the grub launcher and the start of booting, and then the screen resolution changed to a black screen <snape>It seems like the certbot service can't handle several domains within the same certificate. <snape>Yes. I'm going to have a look <pkill9>is there a flag for `guix build` that let's you add a propagated input? (not replace, like with --with-input=package=replacement) <rekado>pkill9: a propagated input only makes sense in the context of installing a package to a profile. <rekado>when using guix package you can install more than one package in the same transaction <rekado>that’s almost the same as installing a package with a propagated input. <pkill9>this is more of a guile related question, but when writing a module, how do i tell guix to append an input to the existing list of inputs of an inherited package? <bavier`>pkill9: usually something like (inputs `(("foo" ,foo) ,@(package-inputs bar))) <pkill9>would i need to import the modules that all the original package's inputs are in? or would they be automatically imported? <apteryx_>Hello! which package provides 'top' in GuixSD? <bavier`>pkill9: you do not need to import the modules, since you wouldn't be directly referencing those packages <rekado>ACTION just wrote a long reply about Guix HPC in response to questions by a journalist for a German science magazine <civodul>were you interviewed specifically on this topic? <civodul>or was it more of a gratuitous plug? <rekado>they wrote me an email asking specifically about the Guix HPC project <rekado>I think they may have found us through the 6-months GuixHPC tweet… <civodul>speaking of which, should we make a mastodon account and have the other one "mirror" it? <civodul>i think that's what dustyweb and others do <rekado>the magazine appears to focus on wetlab / life sciences in general; I hope that my explanation isn’t too technical. <civodul>so their interest was reproducibility in the context of life sciences? <dustyweb>ah yeah I do mirroring... manually currently, but there's no reason you can't do so via a script <rekado>they wanted to get an explanation of what HPC is and what it is used for in life sciences; what problems there are with reproducibility in this context; how Guix solves these problems; and what I can say about the progress the Guix HPC project has made in the past six months. <bavier`>rekado: do you recall why our julia package requires the "objconv" source? <bavier`>afaict it's only used building an internal openblas <rekado>It’s been over a year since I had a look at the julia package. <bavier`>my current issue with it is just that the zip file is unversioned, and was recently updated <civodul>rekado: looking forward to reading that article <rekado>I hope it will actually lead to an article :) <bavier`>weird, our aarch64 gcc doesn't recognize -mfp16-format=ieee ? <bavier`>julia assumes so though, need to patch that out I guess <civodul>i suppose you'll also have those on aarch64 <bavier`>I don't even know what julia is trying to do with this __fp16 in their code; it just seems all wrong <bavier`>or maybe thay write their code with clang <bavier`>that seems like it; they declare a __fp16 return type and function parameter, which gcc didn't support until version 7 <mbakke>Phew, took a while, but I've think I've sorted the pandas tests now. crosses fingers <mbakke>Note to self: Add helper method to get the Python build directory in the next core-updates.