<brendyn>I guess that makes sense in it's own way, since there aren't really any forks of Linux as far as I know <ng0>there are other kind of kernels.. <lfam>brendyn: One has to be sanguine. If I found this situation absolutely unacceptable, I simply would not use computers. <lfam>brendyn: I don't think it's specific to Linux. <brendyn>Yes but that's irrelevant to Linux specific bugs <ng0>well you find some situation unacceptable, otherwise you would not contribute to guix :) <brendyn>Do you think it's worth while announcing security bugs to the world? <lfam>ng0: Indeed :) I find existing methods of putting operating systems together unacceptable when Nix and Guix exist <mark_weaver>the issue is: it would be a *lot* of work to consider each bug and try to determine if it constitutes a security flaw. in practice, a large percentage of them probably could. <brendyn>ng0: That is true, but I don't think anyone else is working on Guix for the reason I am <ng0>I recommend reading 'a bughunters diary' for some small input on bugs :) <lfam>brendyn: It's worthwhile, but I agree with Mark. <lfam>They don't hide the bugs, but they don't usually try to pick out the security related bugs <mark_weaver>brendyn: what's your unusual reason for working on Guix? <lfam>When bugs are publicized, it's usually by a 3rd party <ng0>lfam: or because upstream did not respond at all <ng0>I'd do the same, though I've read that in germany one can no longer publish the exploit code. then again i have not been in the position of having upstream which did not respond <lfam>If we like computers (I do!), I think we have to keep trying to improve, even though the current situation is really bad <lfam>ng0: Hm, sounds like a bad law that will force security researchers to other countries, or underground <lfam>But, I have a strong free speech bias <ng0>there was a recent law change, so it could be fixed again.. i don't know <mark_weaver>in fact, I think it's safe to assume that for almost any project that announces security flaws, a great many of the other bugs they fix can also be exploited <lfam>"Inconsequential" bugs can be used together in creative ways <lfam>It's also fun when a bug fix introduces a new bug. I see that often on oss-sec <ng0> /me plays 'Jesu - we all faulter' <ng0>what we can do is where discussion was picked up again, things like hardening etc.. it does not fix all the problems in the world, but the right techniques will prevent a good amount of exploits <brendyn>mark_weaver: I want to build a libre distro for schools, I suppose that can be done with any other distro but when I look at Trisquel I just see an impoverished Ubuntu. Guix is foremost libre, so it has that momentum. Guix system definitions seem to facilitate easily making preconfigured .iso files for a variety of purposes. I'd like to create a system definition that give students network logins with their <brendyn>guix profile so they can install whatever they want but there would be no pressure on the school admin to update the root instance <brendyn>Also, another experiment is to setup configurations with xorg_multiseat or whatever wayland equivalent so that single computers can be setup in computer labs with multiple monitors, keyboards and mice <ng0>that's cool :) I'm trying to do something similar, though with a different purpose, the live-system is the same. <brendyn>I have lots of ambitions but other people usually don't care so I don't know if these are good ideas <brendyn>Current Edu- distros seem to just be Ubuntu+ random "educational" software <lfam>brendyn: I think that a declarative system like GuixSD is a great improvement over traditional imperative systems for mass deployment <lfam>It should reduce the workload of the administrator <ng0>that was one reason why I picked GuixSD and NixOS for the live-system we create <brendyn>Yeah. Ultimately I dream of convincing schools to switch entirely such that we (Australia) would create end up maintaining our own base system definition for schools that could then be setup and customised freely, but people think it is foolish to think we could ever get rid of Windows <brendyn>Anyway, 1.0 is still a fair way off I suppose, so I intend to spend that time learning to program more <lfam>brendyn: It's not foolish. It will be hard but it can happen. <brendyn>lfam: Yeah, I just can't convince people by pointing out "obvious" ethical reasons, because people rationalise the status quo. Probably using the Indian schools example would be good. <lfam>brendyn: It's worth it to spend a lot of time learning how people are persuaded, in general. I think that ethical arguments are ultimately not that powerful. <ng0>brendyn: I am preparing packaging for SecuShare, which is still taking time for any public release candidate, and I only pick NixOS in addition to GuixSD because one option should be a live-system people can boot into a ready-to-use program on all possible platforms <brendyn>lfam: I have read a fair bit on pyschology. I think theoretically we can design presentations better than RMS can. <brendyn>I don't care much for spending time hating on people, but I think showing people ways they've been screwed over my M$ might be effective. People have alot of trouble understanding the being able to edit code benefits them, but I think everyone has benefited from the creation of Firefox, else the web would be ruined by now <ng0>not to play devils advocate... but the web is ruined :) it wasn't firefox' fault though <lfam>The company I work for uses a 3rd party service implemented in Flash to do all their inventory and sales tracking. It really boggles my mind. <lfam>Adobe doesn't even develop Flash any more <brendyn>The Taiwanese government is pretty bad too. Most of their sites have Flash <lfam>That's the thing about computer systems. Even when they are implemented extremely poorly, they still have a huge value. <lfam>We were just saying the the state of computing is terrible but, on the other hand, I don't think one can deny the tremendous impact computers have had on our lives <lfam>My response is, "Well, I better try to improve things, because I don't have a choice to not be affected by computers." <ng0>yay. thanks for pushing the perl patches :) <brendyn>Hmm, do you have any ideas for community development? I was thinking eventually we could have a sort of Planet Guix where we create and share all sorts of system definitions for fun, like we Could create a boot-to-emacs/spacemacs setup up for instantly getting started hacking on Guix <lfam>It could be a good place to start <lfam>It has a list of 3rd party package repos that is by now way too short <brendyn>I'd like to eventually build an installer that helps guide through installing GuixSD, maybe it could be loaded with an absurd amount of community made configs :p <lfam>I hoped for a GuixSD installer Google Summer of Code project this year, but it didn't happen <brendyn>For basic setup, You just need to decide how to format your hard drive, and then the rest is just installing packages with config basically <lfam>That person is in the channel <mark_weaver>brendyn: the libre distro for schools project sounds like a very worthwhile initiative! <brendyn>lfam: Do you know what the exact command was used to make the GuixSD iso? <ng0>there is no iso at the moment <lfam>brendyn: `guix system disk-image --image-size=1G gnu/system/install.scm` <lfam>See the manual, section 7.1.7 Building the Installation Image <brendyn>I wonder if other GUI installers can be pillaged from <lfam>We were going for an ncurses installer <lfam>But, any installer would be a welcome improvement in my opinion <ng0>I could volunteer to setup an test instance of gitlab... but I'm still exploring gitlab myself. i have no traffic limitation on my server. <brendyn>lfam: I think front-end agnostic is the way to go <lfam>brendyn: What do you mean? <mark_weaver>an installer for guix should be written in guile, I think <brendyn>lfam: I mean, you just write the code to take care of constructing a system definition based on certain choices presented to the user, then add a GUI on top?? <brendyn>I mean, I've never designed a program before but thats how I guess it'd be <lfam>Agreed that it should be in Guile. It had been suggested to use guile-ncurses <brendyn>0.o what's this all about?: ** Message: Cannot create profile directory /homeless-shelter/.config/inkscape. <mark_weaver>brendyn: in the build container, HOME is set to /homeless-shelter, which doesn't exist <brendyn>mark_weaver: running make after having run ./configure --localstatedir=/var <brendyn>That was running `guix system disk-image --image-size=1G gnu/system/install.scm` <mark_weaver>we currently use 'inkscape' to convert our grub background image from SVG to PNG, and that's done as part of building any GuixSD system. <brendyn>Looks like I need to reboot to use KVM <mark_weaver>when it runs, it tries to create $HOME/.config/inkscape <mark_weaver>I've just recently posted a patch to avoid using 'inkscape' for this job, and after some revisions it will be pushed soon. <brendyn>mark_weaver: How do you deal with git noticing all the .po files changing? <mark_weaver>brendyn: when that happens, I avoid using things like 'git commit -a' and instead 'git add' the files I changed explicitly. <mark_weaver>in practice, I don't seem to run into this problem very often. I guess civodul periodically checks those in, but I'm not sure. <brendyn>I'm trying desperately to get my own changes into a branch <brendyn>So far I've figured out I can stash everthing and was able to rebase to a new branch <brendyn>Ok. I didn't realise these were files that were meant to even be in the repo, just like other outputs of make <mark_weaver>the .po files contain the translated strings. they are not merely auto-generated, although I guess 'gettext' includes tools for automatically adding new skeleton translation entries as needed, by scanning the source code for new messages. <brendyn>Hmm, it seems the 7zip patch was submittied with Windows line ending encoding. Does that matter? ***sprang_ is now known as csprng
<brendyn>Font packaging looks like a terrible mess. I'd assumed it was just a matter of downloading a file <brendyn>ACTION 1KiB/s 08:16 | 512KiB transferred <rekado>brendyn: font packaging can be as simple as copying a file somewhere, but when possible we try to build the fonts from source. <rekado>brendyn: when it's just about copying a file people often do without pre-defined build systems and write a builder themselves. <brendyn>I didn't realise fonts needed to be 'built' <efraim>that was a suprise for me too when I first learned it <efraim>gst-plugins-bad has issues when I add qtdeclarative and qtx11extras, so they don't get added yet <ng0> /me updates golang .. yet another sidequest. <brendyn>I'm currently looking through Spacemacs code trying to figure out if it's easier to use Spacemacs since it fixes a few Emacs problems or use Emacs because Spacemacs breaks a few things that work fine in Emacs <ng0>I'm just picking many things because I want to learn more. <Petter>I have sent in a patch updating Go to 1.7 <ng0>want to package gogs, figured I want a go build-system, saw that go 1.7 was released... <Petter>There is this error: "gcc: error: -fsanitize=thread linking must be done with -pie or -shared" <jmd>The Cgroups implementation that guix uses. Is it v1 or v2 ? <ng0>done. thanks for pointing it out to me Petter :) <ng0>do you have more knowledge on go? so far writing the go build-system looks easy enough to me, only difficulty is applying it :) <Petter>Go is my preferred language. I'm not well experienced with building the language itself it though. <ng0>go-build-system would be more about using (build-system go-build-system) in packages, because I need to package 30+ of them for gogs <Petter>Oh, i have no experience with this. <ng0>well... okay. gogs comes with a Makefile which calls go. I'm not sure about the dependencies, those I need to check. but a go build-system is a nice idea. <quigonjinn>ng0: I'd like to learn as well about go dependencies. If they are compiled from source or not. <quigonjinn>I'm using some go software myself, but never looked into the build system <ng0>i must say with the a9 search added the surfraw dependency graph is big... <ng0>i like gogs. their source uri string fits on one line with 89 characters including indentation :) <ng0>i think we need go build-system.. I need to look at gogs some more, but it maybe would make things easier <brendyn>Ok well I accidently commited with that <lostcoffee>also, I think those utilities are for contributing to unifont <ng0>surfraw is now functional. someone interested in reviewing? <OrangeShark>hey ng0, I will try working on a patch for utox when I can. I think one issue is that it should support changing avatars which wasn't working. <sneek>OrangeShark, you have 1 message. <ng0>imo changing avatars is a minor bug which does not prevent functionality <efraim>I'll try to take a look at surfraw soon <ng0>for testing it needs to be in $PATH, straight from /gnu/store/$out does not work. I'll try the same for pbpst soon <efraim>that sounds similar to the 'wrap-binary extra phase I had with bambam <efraim>i almost have grub fixed from the qemu update ***marxistvegan_ is now known as marxistvegan
<efraim>... one of the comments in a patch i'm preparing: ;; ncurses5 was sooo last year <efraim>wow, I really have been sitting on that patch for forever, time to take out that substitution ***cehteh_ is now known as cehteh
<efraim>oh my hack was to give it ncurses6 instead of ncurses5 <paroneayea>I wonder if we should build emacs with imagemagick support. But then again, probably nobody would want this but me. <paroneayea>it gives you the ability to scale down and up images, which might be useful in the mode I'm writing <paroneayea>but then again, I could probably call out to imagemagick to do the same, if I really wanted to. <ng0>why not if we default to 'all options' like always... <jmd>What does it use for image-rendering now? <paroneayea>jmd: I'm writing up a little social networking library. It would be nice to have avatars displayed, scaled down as needed. <rekado>paroneayea: isn't it built with imagemagick support? <rekado>I'm looking at images in Emacs all the time. <paroneayea>rekado: the function imagemagick-types does not seem to be defined <paroneayea>just to have some extra nice resizing/etc features available to elisp. <paroneayea>including some discussion about civodul's "The Emacs of Distros" slogan (which I've given a hard time to before, but is quite compelling) <paroneayea>wingo: also references your writing on "the emacs thesis" :) <kalashkash>I have a question!... there are some difference between login as root on TTY and login with SU command in terminal on GUIXSD ? <paroneayea>kalashkash: but in general, it's not a good idea to log in as root <paroneayea>kalashkash: instead, you should use su for short periods, or even better, sudo for short periods as needed <kalashkash>because when I do "echo $PATH" as root, I get different values, depending on whether I login on a tty or use the command "su" <paroneayea>kalashkash: maybe will do closer to what you want <kalashkash>for example, in TTY to login as root, and excecute 'echo $PATH' I get "/root/.guix-profile/bin:/run/setuid-programs:/run/current-system/profile/bin:/run/current-system/profile/sbin" but with su command, when I excecute "echo $PATH" I get "/sbin:/bin:/usr/sbin:/usr/bin" <kalashkash>then in TTY as root I can excecute the "ls" command, but in a graphical terminal, to login as root with su command i can't excecute "ls" <kalashkash>paroneayea: That must be (my english is not very good) <kalashkash>"su -" login root with their own home! that its solved <ng0>or is line 43 bsd-4 clause? looks like it <ng0>well every license is clear to me but not the one I linked <ng0>2 basic server applications left to package (and write service for) and I can setup a new server with guixsd :) <snape>ng0, what applications are left? <nckx>ACTION please say postfix please say postfix <snape>ACTION was planning to work on postfix, but if ng0 does... <ng0>opensmtpd, git service (I'm writing and debugging that atm), searx, and that's what I can recall atm <ng0>we have dovecot already, right? <davexunit>I need to fix our mysql service sometime. it's quite useless currently. <davexunit>of course, I would be happy if someone beat me to it <ng0>i will not work on postfix, i don't even need it <snape>so you send emails with opensmtpd? does it work well? <snape>guess its easier than postfix to configure <ng0>send with opensmtpd, receive with dovecot <snape>nice. this discourages me from getting into postfix <ng0>someone told me to try it when I was ranting about postfix config <ng0>of course postfix is a contribution worth making :) <snape>yes, yes, of course, but motivation matters <snape>ng0, could you setup something like backup smtp server? <snape>like one that could receive emails when the primary server is down <snape>and put them in a queue, while waiting for the primary <ng0>idk.. my motivation for administration is low. look at the documentation, this should be possible. <snape>that's what I use with Postfix <snape>this important feature is necessary to sleep well :) <ng0>so far I hope to replace $current-server-systems with guixsd in 2017 <snape>ng0 thanks for the link, sounds pretty nice <snape>ng0 did you start working on opensmtpd? <ng0>compiling as I'm writing this <snape>(btw, 'backup mx' is the name of the thing I was talking about 10min ago) <ng0>searx isn't even a requirement.. there are many other searx instances out there. it would just be nice to have it <snape>well, and there is gitolite already packaged <ng0>idk.. i don't like gitolite. that's why I wrote the git daemon service <ng0>I would even use gogs when I could disable web-view for some repositories entirely. The work I mirror to notabug, gitlab, github, etc is okay to be indexed, the rest at least 1 person disagreed on public git index <paroneayea>When they presented their research at Usenix, Qian's group also showed how the bug could be exploited to force Tor users through particular exit relays. <ng0>BEAST only affects tls v1.0 and earlier, correct? <jmd>How is it that guile often gives warnings like: <jmd>guix/build/syscalls.scm:343:12: warning: possibly wrong number of arguments to `pointer->procedure' <jmd>Why "possibly". Doesn't it know? <lfam>Seems like a question for #guile <jmd>lfam: Normally I would ask there. But just recently the devs there have made it quite clear, that user questions are not welcome. <lfam>Hm. Well, I don't recall having seem that warning, and I'm still beginning with Scheme, so I'm not sure <bavier1>jmd: I would not have expected that from #guile devs <jmd>bavier1: I was disappointed too. <mark_weaver>jmd: I think it's safe to say that #guile has a good reputation for being friendly to questions. however, you have a history of asking lots of questions that can be very easily found in the manual, e.g. asking what procedure 'foo' does, when it's right in the index of our manual. <jmd>I will resist the temptation to bite at that. <mark_weaver>it's totally unfair to characterize this complaint as "the devs there have made it quite clear, that user questions are not welcome." <jmd>Well I think both your last statements are "totally unfair" too. <mark_weaver>okay, let's start with my last statement. please provide quotes to back up your claim that "the devs there have made it quite clear, that user questions are not welcome." <lfam>There's a balance to be struck between answering user questions directly, so the user can keep doing what they are doing, and encouraging users to look for answers themselves in the manual, which is better for everyone in the long run. <davexunit>user questions are certainly welcome in #guile. many people, myself included, ask many questions there. <lfam>I often wonder when to tell new users "Okay, it's time to start looking in the manual :)" <davexunit>yeah, at what point do you "teach a man to fish"? <lfam>Whenever you do it, it's important to communicate that you are trying to help them, in a round-about way, by no <lfam>Sounds like it didn't work this time, and now there is some resentment :/ <jmd>Actually I'm starting to wonder if I should configure my irc client to automatically append to the end of my posts, the words "Yes, I have read the manual." <lfam>I wonder if you could do IRC and the manual in Emacs, and the automated string could include which manual sections have been read recently ;) <lfam>Then, others could be able to say things like, "Oh, but it's in this other section, did you look there yet?" <ng0>you could have a bot which queries the documentation and links to the sections <jmd>lfam: Part of the problem, of course, is that many manuals, the information is probably there, but for the uninformed finding it is almost impossible. <lfam>Heh, as long as the bot only speaks when spoken to. It's annoying when bots interrupt the conversation constantly <jmd>(see my last post to guix-deel <ng0>bit: ,function ,page ... or something <lfam>jmd: Yeah, I know what you mean. When I'm beginning a new subject, it's really hard to even know what question to ask <ng0>*bot: ,function ,page ... or something <lfam>You have to "bootstrap" a basic level of knowledge before you can really help yourself <mark_weaver>jmd: suppose you had configured your irc client that way when you asked this question: "I expected this (or #f 0) to give me an error. Why doesn't it?" <mark_weaver>but if you had simply looked up 'or' in the guile index, all of your questions would have been answered. <lfam>As a newcomer to Scheme, I begin all my Guile questions with the Guile procedure index. That seems to work for me <mark_weaver>and the automatically added "Yes, I have read the manual." would have been false. <jmd>mark_weaver: Ok. If you really want to discuss this, we can. But not here, I think. <mark_weaver>having said this, I fully acknowledge that there are many questions whose answers are *not* easily found in the manual, and I for one am happy to answer such questions. <mark_weaver>anyway, the reason guile says "possibly" is because top-level variables are mutable, and so the compiler can only make an educated guess about what procedure will be bound to 'pointer->procedure' when the relevant code in syscalls.scm is actually run. <OrangeShark>jmd, you can also say what specifically made it difficult to find in the manual. The manual could be improved to make finding certain things better. <jmd>mark_weaver: okay. Thanks for the explaination. ***kelsoo1 is now known as kelsoo
<ng0>opensmtpd done, at least the building part. needs service + testing now. <lfam>The build of slock with the patch does succeed <lfam>Well, never mind. It's not caused by my patch. I have the same issue without the patch. <jmd>At the risk of asking something else that might or might not be explained in the manual: I'm getting the error "guix system: error: mkdir: File exists" - now the function mkdir-p (YES I HAVE CHECKED THE MANUAL) does not specify what happens if the target already exists, but the "mkdir -p" command should not complain. <mark_weaver>I don't know what's happening there. it might be from a call to 'mkdir' (as opposed to 'mkdir-p') <lfam>That's the same error I am getting while trying to reconfigure my GuixSD system <lfam>It does not produce a backtrace <jmd>lfam: Probably some recent change has broken something. <lfam>Yes, I'm checking if my addition of openssl-next is the problem <lfam>No, it doesn't seem to be the case <jmd>Oh. commit 6526d43ea4fb0cd151a0d5e9a072c651c1c963d1 perhaps? <lfam>jmd, mark_weaver: That was it <jmd>lfam: Do you want to push a fix? <lfam>Yes, writing the commit message now <lfam>I used the system tests for the first time last night. Now it's David's turn to learn about them :) <lfam>We should all start using them :) <jmd>lfam: I must learn how to use them too. <mark_weaver>I'm not sure where the problem is, but running 'git pull' is taking ages for me right now. <lfam>mark_weaver: All the gnu.org sites are moving very slowly for me <lfam>I wonder what the other screen lockers do? <davexunit>does anyone know of any single-board computers that use a qualcomm SoC supported by u-boot *and* freedreno? <davexunit>even if some auxilary hardware was useless without proprietary drivers (wifi, bluetooth) I would be interested. as long as 2D/3D acceleration works with free software. <lfam>There must be some community wiki that keeps track of this stuff <davexunit>I've been frustrated for awhile now by the lack of such a board. <davexunit>and now that guixsd is close to having u-boot support, it hurts even more. <davexunit>I packaged "emulation station" for guix awhile back when I had hopes of making a tiny game console computer that ran guixsd and used only free software. <davexunit>as a replacement for raspberry pi and raspbian <davexunit>I guess a sunxi board would work too, since there are free gpu drivers for them, too. <davexunit>at least I think there are. I find this whole class of hardware difficult to navigate <jmd>davexunit: Have you checked out boudarydevices.com <davexunit>seems like finding decent free GPU + u-boot support is hard. <davexunit>I'm so ignorant about the world of ARM and all of its kernel forks and foreign boot firmware <jmd>I dunno if any of them do that. I just point it out as a possibility. <ng0>so hiredis is different than redis? <davexunit>I appreciate it. another place to look is helpful. <ng0>i'll just leave redis support out of opensmtpd then, we don't have hiredis <ng0>i assumed they are the same. <davexunit>just leave a TODO comment that hiredis is needed for redis support <davexunit>and maybe someone will come along and package it <janneke>how do i get a package's version from a phases lambda? <davexunit>but you can pass it in from the host-side with an unquote <janneke>hmm, i was hoping to get it from %build-host-inputs or so <janneke>the package may be inherited, and i want the actual version <davexunit>that type of metadata is unavailable to builders <janneke>OK, thanks...then i'll have to do some unquoting or explicitly substitute-keyword-arguments in the inherited versions <janneke>weird...what is zero? for then... I mean, why not (= 0 .. .. ..) <wingo>what is even? why not (= 0 (logand x #b1)) ? :) <janneke>ACTION just showing scheme ignorance <janneke>i didn't know zero? before and like how explicit it is <janneke>and just wondering about the #arg restriction <sapientech>hi guys, am doing a bit of android dev, and i need some 32 bit libs. I am on arch which has multilib, but i don't see many options for guix? <sapientech>whats the state of cross architecture dev with guix? <janneke>sapientech: cross building is supported quite well in guix master <janneke>not sure about android/what compiler you need