<thomassgn>If I installed guixsd on a usb stick, could I easily make it work as a live USB? <buenouanq>as I understand it, the installer \\emph{is} a live image <baconicsynergy>Gonna submit my first bug report too, attempting to install tuxguitar gave me a segfault ***Apteryx_ is now known as Apteryx
<Apteryx>sneek: Later, tell alezost that it doesn't seem to be a pinentry issue, as I tried the following: http://paste.lisp.org/display/331729. It seems that buenouanq was right that my keys were made with GnuPG v1 and I was trying to use those with GnuPG v2. The error message is misleading. <sneek>Apteryx, you have 1 message. <sneek>Apteryx, alezost says: re "No pinentry" error: the problem is gpg doesn't know where pinentry is placed, so gpg-agent has be started with --pinentry-program option. One way to do it is to put "pinentry-program /home/<you>/.guix-profile/bin/pinentry" to your ~/.gnupg/gpg-agent.conf <Apteryx>sneek: Later, told alezost that the problem was two-folded: 1) I needed to re-import my keys into gnupg v2 after exporting them from gnupg v1. 2) To get gpg v2 happy, I needed to point it to my pinentry program as you recommended. Thanks :) <Apteryx>sneek: Later, tell alezost that the problem was two-folded: 1) I needed to re-import my keys into gnupg v2 after exporting them from gnupg v1. 2) To get gpg v2 happy, I needed to point it to my pinentry program as you recommended. Thanks :). <Apteryx>buenouanq: You were wondering how to migrate from gnupg v1 to gnupg v2, right? I did this just now, it was a matter of exporting the key material from v1 and reimporting it from v2, as apparently the keys are not stored in the same format. <ecraven>davexunit: well, why should I not set state with the config? <Apteryx>buenouanq: Also, you'll want to configure the pinentry-program option in ~/.gnupg/gpg-agent.conf to something like /home/youruser/.guix-profile/bin/pinentry. I believe gpg-agent should be smarter and just look for "pinentry" instead of some hardcoded path, but this does the trick for now. <ecraven>isn't everything state on a guixsd system? why should I *not* be able to set ssh keys, create user accounts, and do whatever else? <ecraven>I don't (necessarily) want them in the store, I want them deployed to the user's home (or in whatever other way so that they are honoured) <ecraven>I do understand that putting files in users' homes is kind of a no-go, but I don't really see how adding the ssh key is really different from creating the user account in the first place <ecraven>also, sorry if I didn't explain this properly, but I'm not talking about private ssh keys, but about authorized_keys, so I can actually login to the machine <buenouanq>Apteryx: I didn't have any issues with pinentry - But I installed pinentry-gnome and am just using it with claws-mail. <efraim>it turns out my build-slave had the wrong guile-load-path or something set from before the last merge <efraim>also, is the staging branch also for ungrafting? <rekado>today I'm trying to speed up Guix on NFS by not using NFS. <rekado>writing locally, using inotify to detect changes, then push these changes to NFS. <rekado>I don't like how hackish this sounds but the performance is just too bad here. <rekado>for almost every operation one of these pesky .nfsXXXXX files is created and modified which indicates that the NFS server is lagging <rekado>inotify tells me that I might need to act on MOVED_TO, CLOSE_WRITE, and ATTRIB syscalls <rekado>guess I can ignore all the other syscalls on the store. <jmd>rekado: What kind of NFS are you using? <rekado>we have a ZFS black box which exports a file system via NFS. <jmd>Yeah is it NFSv2 3 or 4 ? <jmd>and did I understand correctly - you are NFS mounting /gnu/store ? <rekado>all of /gnu is on NFS, this is /gnu/store and /gnu/var (aka localstatedir) <jmd>Presumably you're running a foreign distro, not GuixSD? <jmd>If there is any chance of moving to NFSv4 that would probably work better. <jmd>Otherwise, I can only suggest tweeking the mount options. <rekado>NFSv4 is feared by the admins here for some reason. It's very unlikely they'll switch until v3 is no longer available. <rekado>they tried once before and broke things in the process, so they are wary of a second attempt <rekado>I already tweaked the mount options to the extent possible. <rekado>At this point I think I need to hack beyond NFS and take advantage of the fact that the store is append-only (ignoring GC for now) <rekado>with inotify I can detect when the daemon is done working on a store item, so I could move /gnu to local storage and then push only the new store items to NFS <civodul`>rekado: which Guix operations are slow on this NFS store exactly? <civodul`>(sorry, i keep forgeting the details, my memory must be filtering out things that cause bad feelings ;-)) <rekado>civodul: I can't pinpoint just one operation. <rekado>which I think is bound by disk IO <rekado>pretty sure that's misconfiguration on the side of the NFS black box <rekado>but being that it's a black box we can't fix it without involving Oracle <rekado>*at the MDC are a big pain point <civodul>the thing is, derivation builds take place on /tmp, except the final copy operation when the derivation succeeds <civodul>oh but yeah, RPCs that add things to the store may be slow <civodul>and there are lots of reads are from the store, of course <rekado>grafting is terribly slow, despite all the performance improvements in guix <rekado>it's pretty acceptable without NFS <jmd>Silly question: What is the use case? Why do you want to NFS mount the store? <jmd>Generally speaking, NFS is less reliable than local filesystems. <rekado>we want to use central IT services for storage instead of exporting /gnu from our tiny virtual machine <rekado>the file server is used everywhere, so if it goes down nobody can work anyway. <rekado>but if our central Guix server goes down and with it the whole store then running jobs on the cluster will fail. <jmd>So far as I know, it's not possible to share a store between multiple installations. <rekado>because actually we only have one installation <rekado>that's our VM where the daemon runs <rekado>it's the only node with write access to the store. <rekado>every other cluster node treats the store like just another shared file system (read-only) <rekado>the profile directories are writeable on all nodes <rekado>so that people can manage their profiles <rekado>but all access to the daemon is central <rekado>(we use a wrapper that forwards RPCs to the daemon on the VM) <rekado>what's good about this is that we don't need to maintain multiple independent Guix installations on countless workstations and nodes <rekado>users manage their profiles across all machines by going through the central daemon <rekado>so they can use the very same software on the cluster as on their workstations <jmd>I thought that user profiles are simply simlinks to the store. <rekado>to create a new profile generation a new store item has to be created <rekado>and then Guix changes the target of a symlink in the profiles directory <rekado>with this central approach users can still use their own version of Guix; we just centralise storage. <jmd>This is an experimental branch? <rekado>all of this already works out of the box; only a little plumbing is needed <rekado>(like providing a wrapper script for Guix) <civodul>the days of the wrapper script are counted :-) <rekado>civodul: in the inotify log I see that near the end of building a package there are a lot of syscalls like this: <rekado>/gnu/store/ CREATE .tmp-link-24757-1804289383 <rekado>/gnu/store/ MOVED_FROM .tmp-link-24757-1804289383 <rekado>several pages full of .tmp-link creation and moving <rekado>and right before deleting the chroot and the lock file <civodul>did you try --disable-deduplication? :-) <civodul>so if you have enough disk space, just disable it <civodul>or disable it temporarily just to see how it affects performance <thomassgn>derp, so guixsd doesn't currently work with gpt it seems? <jmd>thomassgn: I have never tried, but I don't see why it shouldn't work. <thomassgn>keep getting ext2 embedding errors from grub-install <sneek>Welcome back ng0, you have 2 messages. <sneek>ng0, galex-713 says: yeah solved the gnupg-2.1.15 install problem long time ago <sneek>ng0, galex-713 says: then I tried to see how could I use the same keypair for both gpg and ssh (and tls) and failed (to understand and find the correct informations about gpg-agent and stuff like key formats-keygrip)… so I’ll try another time <ng0>kay.. whatever the context was. <ng0>all the 0ad dependencies are merged. I can give packaging it another try, leaving out nvidia-texture-tools and using the bundled one it needs anyway because of applied bundled patches. <thomassgn>ng0: oh, huh. since I allready restarted to build the config I'll leave it as is for now. Might look into it if needed later on though- do you have anything gpt related in your configs? <ng0>the system should really not care at that low level <ng0>i have grub_boot, /boot , / , /home as partitions <iyzsong>when using GPT, grub need a special BIOS boot parition. <thomassgn>ok, I was struggling finding a way of setting the device for grub-configuration. It's a system with several disks. I thought I had a grub boot partition, but maybe I didn't configure it properly <iyzsong>there is no configure for that in guix, I think grub will pick the BIOS boot partition automaticlly if you have one. <thomassgn>wtf, now I get the same error on an mbr setup... <thomassgn>I'll get ssh running and paste the config soon <iyzsong>thomassgn: the '(device ..)' in grub-configuration usually need to be a disk device like "/dev/sda", not a partition. <thomassgn>ofcourse. derp. For some reason I've had a comment from the generated grub.cfg in the back of my mind telling me it should be a fs... :P ***GrimKrie- is now known as GrimKriegor
<paroneayea>had the not fun experience this morning of trying to do a webrtc call through either epiphany or icecat on guixsd <paroneayea>it would be nice to test some VoiP type stuff in GuixSD and do a community #guix call in it to verify that we get far enough <ng0>webrtc requires features which are not present in icecat <ng0>tox (the software) is not really good for videocalls <ng0>we don't have mumble so far <ng0>i'm worried about etheureum servers for signups for ring <ng0>blockchain is horrible and olgarchic <ng0>but for the moment it could be better than skype <ng0>where I have my 2nd "you are sending spam" verification in 4 weeks now <paroneayea>a lot of things are called "blockchain" these days which aren't very similar to bitcoin, but yes <paroneayea>I didn't know ethereum was required for ring though <ng0>ethereum is similar to bitcoin <ng0>no, they have plans for this <ng0>bitmessage uses blockchain but scales (so far) okay'ish <civodul>taylan: did you eventually solve that FileChooser schema issue with Conkeror? <civodul>i'm experiencing it now so i sympathize a lot more ;-) <ng0>icecat doesn't allow editing of search engine keywords.. :/ <ng0>about:preferences#search you can edit the urls and properties in normal firefox <ng0>so when I add searx.me and change the url to an .onion it works, but in icecat i don't even get to the edit dialogue <ng0>should be in about:config <ng0>but in about:preferences it is normally <ng0>no, it's not in about:config <rekado>there's keyword.URL in about:config --- doesn't that do what you want? <taylan>civodul: for now I've installed the gtk+ package into my profile manually, then set XDG_DATA_DIRS=~/.guix-profile/share:/usr/share:/usr/local/share <taylan>civodul: maybe gtk+ should be propagated by icecat... <ng0>in firefox, you can add searx.me through me about page of it, and then edit it to use another url <ng0>this is not possible in icecat, for whatever obscure reason <ng0>what i mean is, you can edit the property of an added 'one-click search engine' <civodul>taylan: i'm looking into it, but it seems it's really a matter of using glib-or-gtk-build-system <civodul>which would mean that icecat has the same problem since the switch to GTK+3 <taylan>civodul: FYI I also had another issue with icecat 45-gnu1-beta: it crashes after a while when (I think) I leave Twitter open. maybe a memory leak. I'm back to 38 for now... <ng0>maybe there's some lines in the icecat code we need to remove to get the functionality to edit search engines back <civodul>yeah, 45 is known to be crashy, but that's a different story... <taylan>civodul: I don't use the guix conkeror package by the way, I just run "icecat -app ~/src/conkeror/application.ini" where ~/src/conkeror is a git checkout of conkeror <ng0>paroneayea: from news of ring beta 2: A global account identification mechanism based on blockchain implementation on an Ethereum technology. <paroneayea>ng0: hm, too bad... I thought the fingerprints as id system was brilliant <rekado>ng0: does this work in firefox at the same version as the one our icecat is based on? <ng0>i currently have no old firefox. <ng0>I know that it used to work for a long time ***d__ is now known as Guest45177
<ng0>paroneayea: which features of firefox are needed for webrtc videochat to function? i think some of the defaults of icecat disable features needed <ng0>or whatever is patched out <ng0>h264 is disabled in icecat <ng0>so we can build --with-daala or however that would work? <Apteryx>ng0: IceCat 45 is much more stable for me that 38 was. <paroneayea>with sites / clients supporting daala, it should work (assuming --with-daala is what's needed to enable that) <ng0>i didn't talk about crashes.. <Apteryx>ooh, sorry, that reply was meant to civodul. <ng0>using outdated browsers is no solution to crashing not-so-old ones though <ng0>oh, read it the wrong way around <ng0>what about the tor description patch? can someone review and maybe apply it? <ng0>one of the remaining problems about 0ad is the bundled mozjs. it wouldn't be so much of a problem if they would not apply patches <ng0>i think it would be easier to package mozjs-31.2.0(...) and somehow make 0ad use that <ng0>hm.. but even if there's a system package, how to apply the patches which only live in the source of 0ad? just assume that mozjs-31 will just work despite the note that it needs these exact changes for the 0ad engine? <bavier>ng0: what are the nature of the patches? <ng0>how am I supposed to treat this: ERROR: Throw to key `gnutls-error' with args `(#<gnutls-error-enum A TLS warning alert has been received.> handshake)'. <ng0>bavier: let me read them again, it's been a while <ng0>there's just the mozilla tls failure before it gets downloaded to store <ng0>it's because of a recent commit ludovic made, but which one? <ng0>can I add some "ignore tls error" in the package? if it fails here, it will fail on hydra <ng0>well the version of 0ad I package has 31.2.0.rc0, but there was a new 0ad release <bavier>ng0: I get the same tls handshake error with an older guix for that url <ng0>works in the webbrowser <ng0>can we make guix ignore the tls error? <ng0>otherwise we can not package 0ad until mozjs 38 lands on ftp, which can take months, years, whoknows <ng0>I could use fedora mirrors <ng0>or debian, or whoever provides unchanged tar.bz2 <bavier>we point to debian mirrors for many other source <davexunit>sad, there's a decent amount of resistance to reproducible builds among GNU maintainers. <ng0>bavier: fedora is so far the only distro carrying version 38 <ng0>but i don't know if/where they store tarballs <ng0>i could get a source from nixpkgs when they update 0ad <ng0>an update of 0ad would require them to include mozjs 38 <ng0>are freebsd distfiles compiled for bsd or upstream sources? <ng0>which is not optimal.. but it's difficult to find sources <ng0>it has the same size as the source <ng0>would diffoscope tell me if it's 1:1 the same? <ng0>or, how can I diff tar.bz2 files? <bavier>ng0: 'guix hash' would produce the same value <ng0>i will comment that, the freebsd.org link worked. fedora provides no tarballs, debian doesn#t have it, gentoo only keeps stable releases, nixos doesn't have version 38 yet, archlinux only has it in aur so no tarball, ... <ng0>should I make (source (uri a list which includes the original source in case it gets functional one day? <jmd>How do I set a new file-system-service? The manual doesn't explain this. <thomassgn>how do I find what package module a package is in? <jmd>thomassgn: I use grep. <efraim>packages build in order now? `guix build foo bar baz' used to build baz, bar, foo <ng0>where does the python module "xpt" come from, which package provides it? <ng0>could be that this is something which needs to be bundled <jmd>efraim: I thought they built in parallel <thomassgn>I keep having 'ip addr' and ifconfig hang and lock my console. kill -9 has no effect. I have got networking running on one ethernet port earlier, but not now. Any tips? <chatter>planets are not doing allah is doing <chatter>galaxies are not doing allah is doing <chatter>mountains are not doing allah is doing <chatter>medicine is not doing allah is doing <chatter>customers are not doing allah is doing <chatter>you can not get a job without the permission of allah <chatter>you can not get married without the permission of allah <chatter>nobody can get angry at you without the permission of allah <chatter>businessess are not doing allah is doing <chatter>fire can not burn without the permission of allah <ng0>so I take it no one is around with op for this channel <chatter>knife can not cut without the permission of allah <chatter>governments are not doing allah is doing <ng0>bavier: do you have op? or do you know who does? <chatter>food does not take away the hunger allah takes away the hunger <chatter>water does not take away the thirst allah takes away the thirst <chatter>seasons are not doing allah is doing <chatter>animals are not doing allah is doing <chatter>the best amongst you are those who learn the quran and teach it <ng0>or davexunit. someone has to have ops... otherwise I recommend to give more people in other timezones op <chatter>one letter read from book of allah amounts to one good deed and allah multiplies one good deed ten times <chatter>hearts get rusted as does iron with water to remove rust from heart recitation of quran and rememberance of death <chatter>when a person commits one sin a black dot sustains the heart <ng0>i dropped a message in #freenode about it <alezost>ouch, someone wanted to make allah great again <ng0>spam is spam, no matter the content <Fuchs>it's a known issue, I took care of it, poke me if it returns. Thanks :) <davexunit>argh GNU maintainers are some of the most stubborn people I've ever encountered <alezost>Apteryx: I've read the logs, glad to see you figured out those gpg/pinentry problems. BTW to make sneek leave a message for someone, you need to write "later tell" not "Later, tell" <davexunit>should've known better than to chime in on g-p-d <albertoefg>we play every week at a different time so any time zone can join <alezost>albertoefg: thanks for the link! :-) <ng0>does someone know why openssl is not reproducible yet? <ng0>or at least why it is reported as not deterministic? <janneke>sudo: /home/janneke/src/verum/.profile/bin/sudo must be owned by uid 0 and have the setuid bit set <janneke>...and i was so happy with my upgrade ;-) <jmd>davexunit: He does have a point. Dropping timestamps from ar makes dependencies <davexunit>so that needs fixing, but I think he is looking for ways to dismiss reproducible builds <janneke>we really need a make replacement in guile and get rid of this timestamp madness <jmd>Things which have evolved over 40 years take time to change. <janneke>davexunit: oh, i should look into that <davexunit>I don't fully understand that particular problem, all I know is that it doesn't seem to be a problem in Guix <davexunit>but it's been made clear to me over the few years I've been involved with GNU that GNU maintainers *hate* change. <jmd>Well we have had *some* problems. <davexunit>or at least there are a group of particularly vocal maintainers that attempt to block any progress that GNU makes. <davexunit>it was a constant source of stress during the year I worked for the FSF <jmd>GNU is conservative. Sure. <davexunit>for example, I gave a good effort to get gnu.org out of a CVS repo and into a git repo <bavier>there is some benefit to consistency <bavier>but at the same time opening avenues for progress <janneke>bavier: agreed, i'm really bad at seeing that, i know <bavier>janneke, OriansJ: I was wondering if it might be nice to have a bootstrapping discussion at LibrePlanet <bavier>the theme is "Roots of Freedom", so it would seem like an appropriate topic <janneke>bavier: that would be nice, however it would be a bit of a challenge to go there <bavier>janneke: understood, I was thinking about making the trek <bavier>but some collaboration in a talk proposal and slides perhaps <davexunit>civodul: unfortunately, many maintainers feel there are significant downsides to reproducible builds. <davexunit>I think perhaps you could put to rest some of those fears <bavier>is there an archive of the discussion? <davexunit>civodul: hah, and now your mail has entered my inbox <civodul>thanks davexunit for monitoring that thread, BTW :-) <civodul>it's ok that people have questions about repro builds, we just need to explain it more <bavier>summary of the opposing arguments? <davexunit>something about gnu ar being reproducible breaks things <davexunit>and something about recording the CPU that was used to build the software being useful <ng0>mozjs-38.2.1.rc0 .pc file is named "js.pc", can this be the reason why pkg-config does not find it? <bavier>ng0: depends on what other packages expect it to be called, I guess <bavier>ng0: is it installed in the proper directory? <ng0>the build system of 0ad is, well spock would say "fascinating." <ng0>so i have to search a while for the right makefile <ng0>js.pc lists no version <ng0>I don't know much about pkg-config <bavier>no, I would expect that might break something <ng0>ok, so i need to patch mozjs <jmd>Does anyone have an example configuration of file-system-service-type ? <bavier>idk how `pkg-config --exists js >= ...` would be expected to work in that case <jmd>or even a description - it's not mentioned in the manual. <baconicsynergy>everything seems to be going rather swimmingly with guixsd, but i have an interesting issue <baconicsynergy>I attempted to install libreoffice, and I can confirm that its placed within /gnu/store/* <bavier>baconicsynergy: you installed with `guix package -i libreoffice`? <baconicsynergy>its the last little issue before I can use GuixSD as my main work distro! <bavier>baconicsynergy: I think the executable might be called "soffice" <bavier>baconicsynergy: it's unconventional in that we tend to stick with what upstream uses. Most other distros change the libreoffice binary names or include symlinks, etc <civodul>we could include that symlink if everybody else does <jmd>All other distros I've used also call it "soffice". <civodul>me too, but i haven't used that many distros ;-) <ng0>we should symlink it.. <bavier>baconicsynergy: good luck with guix! thanks for asking <jmd>If we do decide to do that, perhaps we can also have some other common ones, like lex -> flex, yacc -> bison, cc -> gcc etc. <ng0>soffice is pressumably from "Star Office" whic hit used to be once upon a time <ng0>sometimes upstream sucks, so we have to make the changes. <ng0>i'm just guessing based on the applkications history <ng0>and I think pybitmessage should use python3 and pyqt5, but they don't think it's an important issue (the pyqt5), so either upstream fixes it at some point, or I will when I make space for that <ng0>i think i will finish the 0ad before the end of the year <ng0>and then I will leave games to those who like to package games ***kelsoo1 is now known as kelsoo