<chipb>if I remember the detail correctly -- in effect instead of only using tramp's concept of what the remote search path should be, supplement it with what gets set through your .profile during shell init.
<chipb>I can see why they don't default to that, but it behaves poorly on less "regular" systems that don't install everything in, say, /usr.
<nckx>If my memory serves me, I was able to reproduce it only on a borrowed laptop with 2 wi-fi cards, but I had to give it back. The problem went away when I took either one out. You're not dual-wielding wireless chips by any chance?
<dissoc>im sure it's something i did wrong. not sure what tho
<cnmne>chipb: setting tramp-remote-path seems to be working; still getting a failed insertion, and the debug is saying there's no file notification program; I installed inotify-tools, but i guess that's not what it means
<cnmne>also the Messages say there's a "wrong method specification for 'ssh'"
<chipb>cnmne: hm. I'm not sure I tested tramp shell as I generally do that direct from a non-emacs terminal.
<chipb>you did clear the tramp connection cache for the host, yes?
<chipb>out of curiousity, what does 'getconf PATH' return for your guix host? I no longer have one running at my disposal.
<cnmne>yes, deleted `.emacs.d/tramp'. and getconf isn't a command i have or a package. is that supposed to be a builtin ?
<cnmne>all I find is the python package, is that it ?
<chipb>huh. well, I guess that'd probably do it. it's apparently used by emacs to seed the remote $PATH setting.
<cnmne>lol i thought it would add, but i see now what you're saying
<cnmne>I'm not even sure the path is an issue. tramp is letting me peruse the filesystem, but not open anything
<cnmne>setting tramp-remote-path by appending tramp-own-remote-path seemed to do the trick
<chipb>I'd bet the easiest path forward is to add the static string for your user and possibly system profiles to tramp-remote-path in emacs configuration. or the 'tramp-own-remote-path thing I ended up doing.
<romulas>It says it supports it but not the model with R2.0 (no info in the wiki)
<bUst4gr0>is there a way to browse your packages by category or i search a replacement for iceweasel i used on arch freedom supported os, i used icecat (really hardened) + iceweasel (for nowaday websites)
<jamesb>Greetings, I looking to install Guix system on my Lenove E595 laptop. Problem, failing to boot from USB with error: driver 'pcspkr' is already registered, aborting... any pointers or help appreciated. Thank you
<jamesb>apteryx, it's only to do with the speakers lol! There are some workarounds but they're for already installed systems. Not sure if I'm keen on editing an iso file! A kernel cmd would be helpful though
<romulas>Yeah it is a bug that it tries to load it when it is already loaded, lol.
<jamesb>I'm going to try and blacklist it at the kernel boot line thingy. I'll be back :)
<romulas>guile: warning: failing to install locale
<jamesb>A support forum for guix would be helpful, most answers would be there
<romulas>GNU Linux-Libre 5.6 kernel deblobs three new drivers that have been included in the Linux 5.6 kernel series, namely AMD Trusted Execution Environment, ATH11K WiFi, and Mediatek SCP remoteproc Additionally, it also cleans up the Nouveau, AMDGPU, and AMD PSP drivers.
<manzerbredes>Hi! I a trying to build my own library package, however I am facing to the error during the "validate-runpath" phase. I have a bunch of .so file generated during the compilation phase which depends on each other. Thus there are not in the RUNPATH, which raise me an error. What is the best way to solve this ? Should I skip the "validate-runpath" phase ?
<xelxebar>manzerbredes: Jumping in sideways on the discussion here, but $ORIGIN/../lib means that the executable must exist in that exact filesystem relationship with its libraries to work. An absolute path would let it work even if copying the executable to $HOME or wherever.
<civodul>efraim: i think we can keep the same module names as in Guix
<zimoun>civodul: yes, it could be very cool to have a collection of "old stuff". In the same time, something is laking to navigate in the history of version already in Guix. I mean I use "git log" which is not fine.
<civodul>in some places i used specification->package to make it less sensitive to API changes
<efraim>civodul: that's generally what I am to do with guix-bioinformatics, octave-3.4.3 is in (gn packages maths)
<efraim>I should also send a patch for tensorflow-native, with #:substitutable? #f it's a big speed-up if you have the hardware to take advantage of it
<zimoun>civodul: and speaking about old version, it raises also the ability to change the building toolchain. I mean, if for example Python 2.5 is in guix-past, but then the version of python-numpy I need is in old Guix, I should be able to rebuild all the graph with specific version. It is already possible but it requires a bit of Guix knowledge.
<civodul>zimoun: you mean the implicit inputs, right?
<civodul>we won't solve everything today though :-)
<zimoun>civodul: yes implicit inputs. Somehow be able to specify which toolchain is used by <foo>-build-system. And not today, sure! Tomorrow morning? ;-)
<civodul>janneke: i see you updated the 'guix' package but yesterday mothacehe reverted the update
<civodul>due to test failures, though i don't know what exactly
<mothacehe>civodul: i think it was an error on my side (corrupted .go), I ended up updating guix just before janneke update
***Kimapr_ is now known as Kimapr
<janneke>civodul, mothacehe: yeah -- i noticed; so guess we're fine
<mothacehe>I just created a strace service that logs one process trace to /var/log/strace-pid.log, hope it will help me understand what I suspect to be a very hard to reproduce Shepherd race-condition.
<txt-file>but I just ran "guile --version" and got 2.2.6
<civodul>txt-file: in 1.1.0, Guix uses Guile 3.0, but the "guile" package itself was still 2.2
<civodul>you'd have to install "guile-next" to get 3.0
<civodul>this changed ~1 month ago, and now the "guile" package is 3.0
<civodul>try "guix pull" to get the latest & greatest!
<txt-file>top also shows that guile 2.2.6 is running when I execute "guix pull"
<blendergeeksimpl>I am trying to install guix. Unfortunately, I cannot get the network working. When I use `ifconfig -a`, it only shows the local loopback. I ran `lspci` and I have two network cards. One is an 'Intel Corporation Wireless 3165 (rev 21)'. That one requires nonfree firmware and I thus expect it to not work. The other one is 'Realtek Semiconductor Co., Ltd. RTL810xE'. This one works in other fully free GNU/Linux distros. I ran d
<blendergeeksimpl>…ealtek.ko not loaded, maybe it needs to be added to initramfs?' Any ideas on how to enable my network card?
<txt-file>Am currently running "guix pull" (and waiting as it takes ages on my test-netbook
<nckx>You need the r8169 driver, right? It's definitely available as a module on Guix Systems.
<mothacehe>blendergeeksimpl: the relevant module is "r8169" you can check if it's loaded with lsmod command. Sharing the output of the "dmesg" command would also be helpful here
<nckx>I'm not sure about that. I have an r8169 card (Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 0c), not quite the same) and it prints nothing to dmesg. Works though.
<rekado_>civodul: I also have a few old things in guix-bimsb:/bimsb/packages/variants.scm
<nckx>All right, I'll upload the ISO with realtek added to the initrd as suggested above. If you trust me you can try it. If you don't, well, probably a good habit but I'm out of ideas (apart from setting up your own Guix check-out to build one, or trying different kernel versions which also requires the former).
<manzerbredes>Hi again, I am able to compile my own package for now. But I wanted to have access to the gcc used to build the package afterward and thus I added gcc to the "propagated-inputs". However, the package failed to build now, and It seems that it is using a different gcc (I just add gcc to the propagated inputs). Then, after investigation, it seems that during the build phase, a different gcc is used (the
<NieDzejkob>I solved this by returning each machine-system combination separately
<NieDzejkob>you might try setting the system field to a list, though
<manzerbredes>NieDzejkob: My package is building a bunch of shared libraries. Then package users are expected to used the shared libraies during the linking of their own softwares
<apteryx>NieDzejkob: thank you. That'd be most elegant.
<NieDzejkob>manzerbredes: why does that imply having to propagate gcc?
<manzerbredes>NieDzejkob: I did "guix pack" of my package, tranfert it on another machine (which do not have guix). I extract it on the root of the file system. Then I tried to compile a small app and linked it to my shared libraries under /gnu/store. And I have linking problems related to my .so file like undefined reference to `pow@GLIBC_2.29'
<apteryx>ci.guix.gnu.org is repeatedly timing out.
<NieDzejkob>manzerbredes: it would be more conventional to use "guix pack your-package gcc-toolchain", I think
<nckx>dftxbs3e: AGPL3+ doesn't change how Cuirass would interact with GPL3+ software, right?
<bhartrihari>Hello, I installed node package but node binary hasn't been added to my path. (npm gives me the following error: /usr/bin/env: ‘node’: No such file or directory). Is this supposed to be so?
<nckx>bhartrihari: Guix's firstname.lastname@example.org provides 3 binaries (node, npm, npx). What's the value of $PATH, and how are you trying to run ‘node’ (i.e. why is /usr/bin/env involved)? Does it work from a (fresh) terminal?
<bhartrihari>nckx, `whereis npm` gives me the following `npm: /gnu/store/37lhmlavwvaf58kkqmmw0jx73378j80n-profile/bin/npm`. I have both npm and npx commands after doing a `guix install node`, but not the node command. It doesn't work from a fresh terminal.
<bhartrihari>`whereis node` doesn't give any useful output, merely `node: `.
<lprndn>rekado, bricewge1 : On the LightDM service, How dirty would it be to use a third servicewhich would expect extentions from lightdm and greeters just to merge everything..? (My brain is melting.)
<nckx>bhartrihari: I'm genuinely confused. What does ‘readlink -m /gnu/store/37lhmlavwvaf58kkqmmw0jx73378j80n-profile/bin/npm’ say?
<nckx>Installing node 10.19.0 here, from current Guix master, gives me all 3 binaries, and ‘node’ starts a REPL.
<NieDzejkob>civodul: I took another look at the LLVM 10 issue, and I can't find the message of Pierre you're referring to. Would you mind posting a link?
<bhartrihari>nckx, The output for the command you mention is `/gnu/store/p0b64xpigk7lblw0vbk49qp66vhbjavj-node-10.19.0/lib/node_modules/npm/bin/npm-cli.js`
<NieDzejkob>bhartrihari: what does `ls /gnu/store/p0b64xpigk7lblw0vbk49qp66vhbjavj-node-10.19.0/bin` say?
<nckx>bhartrihari: Hmkay, and what's in /gnu/store/vkwbavbvnsp58xz34cz11bz63729qph7-node-10.19.0/bin? Also, is your guix up-to-date (guix describe)?
<nckx>Installing node here gives me /gnu/store/vkwbavbvnsp58xz34cz11bz63729qph7-node-10.19.0/bin/node (i.e. a different hash, I'm pulling now to make sure I'm current).
<niklascarlsson>Hi everyone! :) I want to create a system service so that I can start the tool mbpfan automatically. It requires sudo access so I figured a service that I could add to my system configuration would be the way to go. If I want to start experiment with creating such a service, does anyone have any tip on where to put it and how I could load it into my system config.scm :)
<nckx>bhartrihari: I made a pasto above (my hash instead of yours), I meant to ask what's in /gnu/store/p0b64xpigk7lblw0vbk49qp66vhbjavj-node-10.19.0/bin.
<bhartrihari>nckx, There are 2 links there, to npm and npx binaries. My guix is not current.
<nckx>The current documentation is a text file manually kept in sync with the code by hand (and, sometimes, not). That includes each service-type and default and the like. So much room for Guile magic.
<raghavgururajan>nckx: Sorry, I made a mistake. I thought my patches didn't affect openjdk and icedtea. But they have affected it indirectly. Coud you please apply following patches on current wip-desktop and then try to build openjdk and icedtea please?
<joshuaBPMan>nckx: I've never used it. I just read a bit of the guile devel's messages...looks like rendering to html is causing some issues.
<bricewge1>The thing is that function (getrpcbynumber) can be provided by 'glibc' or 'libtirpc'
<lprndn>bricewge1: the service would be private. I'm not a big fan of the added complexity but it allows the user to define either a lightdm service, a greeter one or both and always get a working service... (does it?) I'm just exploring ideas really... I don't remember why but I ended up discarding most other possibilities (and this one too soon...)
<bricewge1>I made a static version of libtirpc, but it doesn't include the function 'getrpcbynumber' because it was already present in 'glibc'
***certifiedcitrus is now known as _
***_ is now known as Guest57951
<nckx>raghavgururajan: Your patch set is missing glib-disable-failing-tests.patch.
<txt-file>Is there a way to disallow local builds and tell guix to only use stuff from ci.guix.gnu.org?
<txt-file>My netbook is running "guix pull" since 3 hours and I fear to need to do "while true; do guix pull; done;" to stay up to date.
<bricewge1>lprndn: I don't remember about talking about the 3 service-type implementation
<bricewge1>lprndn: If it is private it seems fine to me, but I found such services in the codebase yet.
<Guest57951>txt-file: probably You can use something `if [ $(guix weather) != '' ]; ...`. IDK, though.
<nckx>txt-file: If your machine is too weak to pull, use guix pull --commit=<an older commit>. I'd say you can check ci.guix.gnu.org for good commits (a few hours old will suffice) but it's down. 😒 Here are all commits: https://git.savannah.gnu.org/cgit/guix.git/log/
<lprndn>bricewge1: yeah we didn't discuss about it yet. I'm trying it right now. I'll probably send a draft soon with maybe some other possibilities I'm not satisfied with.
<txt-file>nckx: ci.guix.gnu.org normally shows how much is already compiled?
<nckx>Yes. There's a jobset for guix pull, I forget what it's called.
<nckx>gnutec: Thank you for blogging about Guix (with pretty pictures)!
<nckx>Glib did indeed build fine the second time, by the way.
***boegel|quassel is now known as boegel
<nckx>raghavgururajan: No, I'm building different hashes again. I started from commit 9c0156a5040683843d594c9b02441f3a1b299d3f (upstream wip-desktop), applied your patches.tar.gz, then downloaded your glib patch directly with curl (sha1sum 8adc5bb6daa2951097eccbe0c0595058692ad349): https://paste.debian.net/plain/1152724
<nckx>raghavgururajan: I suggest we abandon this error-prone method, and you host your own wip-desktop repository at e.g. gitlab or whatever, and I pull from there.
<nckx>raghavgururajan: If it's a proper mirror, yes, otherwise it might stop tracking upstream altogether (I've not used Gitea myself).
<raghavgururajan>nckx: Locally, I have work-dir "wip-desktop" that tracks origin/wip-desktop (guix upstream). This dir was used for guix bootstrap. Also, this work-dir has a branch called "temp", where I perform my work. From time-to-time, I destroy and create temp.
<raghavgururajan>nckx: I am gonna create new repo "guix-wip-desktop" on git.disroot.org. Now, can I push to this remote repo from my local branch 'temp'?
<raghavgururajan>Or should I set the 'temp' branch to track the remote "guix-wip-desktop"?
<apteryx>hmm, guix deploy is taking a looong time, without telling me what is happening (guile is using CPU on the remote, but the local guix deploy script last output was 'sending 9 store items (4 MiB) to ...'
<nckx>raghavgururajan: What I would do, since nckx would really like to start building before they have to leave: check out your ‘temp’ branch, get it in order—exactly how you want me to build it, then ‘git remote add disroot https://disroot/blah/blah.git’, then ‘git push disroot HEAD:wip-desktop’. That will simply ‘send’ the tip of your local temp branch to the ‘wip-desktop’ branch @ disroot. You can decide if/what you want to track later.
<nckx>I wish I had more time to talk you to the git part but I don't.
<nckx>raghavgururajan: Thanks! Right now I just want something I can pull/download/whatever in the next ~10 minutes and build the exact thing you need, because I won't be around to babysit the build much.
<lle-bout>nckx, gnutec: also GNU Guile's startup time and execution speed
<drakonis1>its one of the primary reasons why nix is great
<nckx>raghavgururajan: Oh, I see what likely happened. Try ‘git remote set-url --add --push disroot <the SSH URL>’ and try pushing again. The SSH URL is the one that looks like email@example.com, it does not contain git:// or https://.
<raghavgururajan>nckx: You may still have to use glib-disable-failing-tests.patch separately, as it was leftout in glib patch.
<leoprikler>That's just a claim with nothing backing it up, really. For one, you'd have to show why "packages" are too high an abstraction unit, and also how Nix does does something substantially different.
<leoprikler>All I see here is less bracketed code, which is fair enough if that's your preference, but has the same semantics otherwise.
<mbakke>txt-file: there are two problems: you need append instead of cons* in the packages field, because (map ...) returns a list, and you need an extra ) after the gdm-configuration line, while removing one ) from the %base-services line
<apteryx>civodul: I think was (gnu machine ssh) is doing at line 214 is serialize the uuid object to a string, and attempt to reform it by refeeding that string to string->uuid (the only way to make an object pass through a Gexp strata)