<eric23>I get an error running guix pull ***Piece_Maker is now known as Acou_Bass
***Piece_Maker is now known as Acou_Bass
<rekado>Is there a variant of download-to-store that is cached? <civodul>the new M-x guix-system-generations is really nice <civodul>well it's probably not that new, but i had not seen the latest changes <rekado_>for the big storage we got for berlin.guixsd.org I’d like to try to connect it via glusterfs <mb[m]1>rekado: sweet! What made you choose it over CephFS? <rekado>we had installed glusterfs on another cluster, so I remembered glusterfs and didn’t think of CephFS. <rekado>I’ll take almost anything that’s not NFS :) <mb[m]1>Oh, right :) Ceph imposes a certain storage architecture, so for small/simple setups it may not be the best choice. <mb[m]1>I evaluated Gluster some five years ago to replace an OCFS2 setup (ugh), but found it was pretty fragile under stress. Hopefully it has gotten better. <mb[m]1>Speak of the devil, the last message over at #debian-security is CVE-2017-15096/glusterfs. <htgoebel>last week "make check-system TESTS=installed-os" worked without any problem <brendyn>maybe the opennic dns server im using doesnt have it for some weird reason... <brendyn>ok it goes to localhost what the hell <brendyn>because its in my /etc/hosts file... i wonder why i did that <efraim>armhf and aarch64 cant use regular grub <mb[m]1>I'll be experimenting with building a kernel for Rockchip Veyron this weekend. And then embedding u-boot in libreboots "depthcharge" loader. <civodul>rekado_: regarding the big storage you got for berlin, it'd be interesting to have long(er)-term archival of nars <civodul>like currently 'guix publish' advertises a TTL of 45 days <civodul>we could make that 12 months or something :-) <enick_2>Hmm, I think I'm missing something about how Debbugs works. I just replied to this bug, and it now says modified today, but my reply isn't shown <enick_2>Do bugs lock themselves after a period of time? <civodul>enick_2: it takes a while if it's your first message to the list <enick_2>No, that can't be right, as there were other messages CC'ing debbugs. It does appear if I click "Toggle" at the top of the page. For some reason it's not considered a first-class message. <enick_2>civodul: sorry, saying no to myself, not you you :) ***enick_2 is now known as sturm
<sturm>It's not my first message to the list though, so I guess that's not the cause <Achylles>Why should I guix instead of "apt"in my debian box? Trying to understand this package... <sneek>Achylles, you have 1 message. <sneek>Achylles, rekado says: Actually, the mailing list is help-guix@gnu.org, not guix-help@gnu.org <Achylles>Why should I guix instead of "apt"in my debian box? Trying to understand this package... <civodul>mb[m]1: looks like we don't have substitutes for 'glibc-final', i.e., /gnu/store/3h31zsqxjjg52da5gp3qmhkh4x8klhah-glibc-2.25 <civodul>i think that's because 'glibc-final' is private <civodul>so the Cuirass/Hydra stuff doesn't pick it up <sturm>civodul: you're right, I was being too impatient - the message is showing up properly now <htgoebel>civodul: Running "make check-system TESTS=installed-os" hangs again. SO I can't verify the overhauled "overlay instead of unionfs" patches. How to proceed? <civodul>htgoebel: could you try to see where it hangs? does it hang in the same way on master? <rekado_>htgoebel: does it only hang sometimes when running it multiple times? <rekado_>or does this only happen for specific versions? (e.g. only with your changes) <htgoebel>civodul: Each turn-around takes 2 hours :-( <htgoebel>This means I need to change the vm's memory and do another cycle. <htgoebel>I'm doing this for a week now. This is really frustrating. <Achylles>I mean that I use debian admin package ""apt/apt-get". Why should I use guix? <Achylles>Which is the advantage of using an external admin package over a native one? <rekado_>Achylles: Guix is a different package manager with different properties. I suggest browsing the website for a list of features. <civodul>htgoebel: if you test on master, hopefully there are substitutes available for it <civodul>but anyway, if the other tests pass, that's probably good already <rekado_>Achylles: please check out the list of features on the website and the documentation. <civodul>rekado_: i've sent an email re FOSDEM <rekado_>Achylles: if you’re happy with apt that’s great. We won’t try to convince you to use something else. :) <Achylles>can I "apt install guix" to have a try on it? <rekado_>Achylles: there is no “guix” package in the Debian repositories. <roptat>but you can still install the "binary" version, it's pretty easy <roptat>and then guix gets managed by guix :) <Achylles>Have you contacted debian devs to add guix to the official repos? <roptat>yes, but it doesn't comply with their packaging policy <roptat>we need to create a /gnu/store directory, which is not standard <Achylles>I am making a guix usb to try it upon reboot <jonsger[m]>Achylles: I use apt as package manager as well as guix :) <Achylles>jonsger[m], what are the advantage of using guix instead of apt? <rekado_>none of these implement “functional package management”. I can only recommend reading the descriptions of Guix or watching the talks on the website. <civodul>still, hydra.gnu.org lacks substitutes for /gnu/store/399spz1vczfqzvyx352ln9bwsjm2ykwh-glibc-intermediate-2.25, meaning that people typically end up building it by hand currently :/ <civodul>transactional upgrades/rollback as non-root is not something apt & co. provide <rekado_>civodul: would it help to define a set of core packages and evaluate and build those more frequently? <rekado_>currently, cuirass creates an evaluation for all packages and it takes a long time before it can start a new evaluation. <civodul>rekado_: that's a good idea, but the problem here is different i think <civodul>rekado_: another problem is that Cuirass does nothing while evaluating, and does not evaluate while building <Achylles>civodul, ok I will have a look on the video <civodul>Achylles: well again, we won't argue :-) <civodul>it's up to you to see whether it's of any use to you <jonsger[m]>Achylles: I use guix _not_ instead of apt. I use it as addition so I can easily install the most recent version of gcc for example :) <rekado_>I’m using it on a Fedora workstation instead of yum/dnf so that I can get reproducibility, roll-backs, ad-hoc environments, … <efraim>The ad-hoc environments are great <htgoebel>Using more memory for the VM help continuing the test. Now I get an error regarding my last change. <htgoebel>Is "canonicalize-path" a string-only operation, or does it work on the file-system? <mb[m]1>civodul: I've noticed this too, and also that the ungrafted glibc (I think) is prone to garbage collection(!), so people may end up building it multiple times. <efraim>I noticed that if I have a grafted substitute then guix will prefer that over downloading an ungrafted one and then grafting it <efraim>Also gnupg builds and installs gpgscm, but online documentation on using it is sparse <civodul>that was supposed to be a hack for internal use <efraim>I didn't check 2.2.1, but 2.2.2 it gets installed <civodul>who could guess gpg would compete with Guile? ;-) <htgoebel>civodul: For overlay you recommended using "canonicalize-path". This fails with "No such file or directory". Looks like its accessing the file-system, like "realpath". I need a string-only operation. <bavier>have we thought about substitute servers serving up package deltas? <htgoebel>civodul: BTW: I could not find the definition of "canonicalize-path" anywhere. <htgoebel>ACTION loves the well structured Python standard library using spelling module names <civodul>htgoebel: it's implemented in C, in libguile <civodul>but yeah, i agree that there's a whole lot of stuff in the default module <civodul>would be nice to make it more modular, but it's hard to do at this late stage <htgoebel>civodul: Anyway, I a string-manipulation-only version of it. Is it available? <mb[m]1>Any chance we could squeeze in a Guile 2.2.3 update (and release!) this rebuild cycle? <civodul>wingo: what do you think? would it be doable for you? <htgoebel>No "normalize-path" in guile? Seriously? <htgoebel>Do I really have to implement such a basic function myself? <civodul>htgoebel: how would it differ from canonicalize-path? <htgoebel>civodul: It must nor access the file-system, but work just on the string <htgoebel>Which is quite an important difference if the paths not yet exist <civodul>ok so what would it do? strip extra slashes? <htgoebel>convert "foo/bar/../bla/./x./y/.." into "foo/bla/x" <htgoebel>But maybe I'll just use string-append and leave this to the remaining code. <bavier>bms_: I just read your mail on guile-devel <ng0>civodul (and everyone): A follow up to my comment in the awesome-wm update: bug#29196: upstreaming of reproducibility related patches <civodul>htgoebel: problem is that ".." resolution is not lexical in Unix, so you cannot do what canonicalize-path does without accessing the file system <civodul>there were discussions in GNU/Hurd to switch to lexical-dot-dot, but that never materialized, although it would have been much simpler <dustyweb>is anyone using the guile-commonmark package from guix? <efraim>openmpi and libfabric built on aarch64 before psm was added <bavier>efraim: when was that change made? <bavier>efraim: oh, ok, we should fix that then <efraim>tagged psm as intel only, marking its use in libfabric and openmpi as (match (%current-system)... <bavier>efraim: great, that's just what I was thinking <efraim>i should figure out something like (match (%current-system) ((supported-systems psm) '("psm"... <bavier>efraim: (member (%current-system) (package-supported-systems psm)) ? <efraim>i'll try that, then if the supported systems change we won't have to chase it down over the other packages <bavier>with appropriate splicing and such, yup <janneke>dustyweb: it looks like i got guile-emacs resurrected, but it needs to be compiled with -g (or without -O2/-O3) does that ring a bell? <dustyweb>janneke: interesting! it doesn't, but... I don't really know much about gcc tbh :) <eric23>I am having trouble setting the time. "sudo ntpd -s" gives ntpd: unknown user _ntp <janneke>*woot*, ran my first Guile Scheme inside Guile-Emacs ;-) <taohansen>janneke: how is Guile-Emacs for daily use? could i switch my EXWM config over to it seamlessly? <janneke>taohansen: I just ran it for the first time. While it's impressive how well everything seems to work, I don't think it's ready for daily use yet. <taohansen>also: i have a HiDPI screen. both the visual elements of SLiM and my cursor are tiny. i can deal with SLiM but my cursor frequently gets lost. i've made the appropriate .Xresources change, `Xcursor.size: 48` but my cursor remains tiny. this is on GuixSD. anyone know a solution? <janneke>taohansen: same problem here, tiny cursor. <taohansen>hmm yeah, all my other .Xresources settings work fine. HiDPI is surprisingly workable with just EXWM but the tiny cursor is bothering me. <janneke>also rather happily running exwm-x since start of summer <civodul>nowadays everyone replies from their toaster or fridge, which appear to enforce top-posting <bavier>I'm guilty. haven't found a reliable way to inline-reply at $dayjob <janneke>on an on, boasting how user friendly their computer (sigh) is, yet cannot have it do any of the most simple or sensible things -- "this is what the computer does, everyones computer does" <janneke>better to see the weird/funny side of it <civodul>i just stumbled upon a series of such messages in a row <rekado_>ACTION downloads texlive-texmf to build numpy :( <bavier>ACTION needs to fix texlive on aarch64 to build hypre :( <mb[m]1>ACTION just started building binutils for some reason, when building out all dependents of "bdftopcf" :( <mb[m]1>Oh wait, that's because I removed Hydra from substitute-urls. <mb[m]1>There's a bug here somewhere, that when using multiple (or many, even) substitute servers, Guix sometimes "gives up" immediately after getting a corrupt substitute even with --fallback. <rekado_>I really want to remove the big texlive from matplotlib and the like. <rekado_>I just couldn’t get it to work, and rebuilds take a very long time. Hard to iterate. <mb[m]1>The --fallback problem seems related to offloading. Will file bug. <bms_>What's there to not be too good about? <mb[m]1>Oh, so many things :) but really, I'm fine. Just challenging conventional idioms.