IRC channel logs

2020-05-04.log

back to list of logs

<cbaines>ryanprior, given that Guix is a tool for software deployment, there's plenty of opportunities for that. I especially like Guix Jupyter
<rekado_>as a tool I don’t see the need to wrap Guix. As an operating system I think there’s value in stronger opinions.
<dissoc3>i just discovered guix jupyter last night. havent really used it but it looks pretty awesome
<janneke>hah, sizeof-ifconf is wrong
<ryanprior>Learning guix is reminding me of ~2009 when I was using debootstrap + fakeroot to manage environments, then I learned to use lxc and that was cool but still very low level, then lxd and docker came along
<janneke>now, where's that defined?!
<rekado_>cbaines: the cleanup-cuirass-roots job failed because we’re using an older commit with a bug!
<rekado_>we never reconfigured with the follow-up commit that included the missing module import
<rekado_>so it never actually finished cleaning up all dependencies
<rekado_>I get a 504 trying to fetch from the maintenance.git repo
<rekado_>jonsger: can you get in touch with Adrian about the license of their contributions to the icedove package definition?
<jonsger>yes
<rekado_>I only just noticed that it is supposedly under AGPL
<rekado_>I should learn to read before resuming other people’s work
*janneke -> zZzzz
<mbakke>pinoaffe: you need to use screen-locker-service-type to make screen lockers work on Guix System
<mbakke>civodul: adding the previous libc locales by default makes a lot of sense, thanks
<anadon>Python dependencies are supposed to be native-inputs, right?
<pinoaffe>mbakke: aight, I did not know that
***dingenskirchen1 is now known as dingenskirchen
***catonano_ is now known as catonano
<anadon>During building, how to I add programs to $PATH? In particular, when trying to build watchman, it wants python to be findable via $PATH.
<anadon>warning: no binary for interpreter `python' found in $PATH
<pkill9>anadon: you can use (setenv "PATH" ... to set it
<pkill9>so something like (setenv "PATH" (string-append (assoc-ref inputs "python") "/bin"))
<pkill9>put that in a phase before the build phase
<pkill9>and make sure it has access to 'inputs' e.g. (lambda* (#:key inputs #:allow-other-keywords) ...
<apteryx>mbakke: finally got to try core-updates with my manifest
<apteryx>I had a failure for spacefm, which depended on fakeroot that failed to build
<anadon>pkill9: https://paste.debian.net/1144807/ has the error "error: (#:allow-other-keywords): invalid keyword argument list"
<pkill9>oh woops, it's #:allow-other-keys
<anadon>pkill9: error: (%modify-phases phases* (add-after (quote build) (lambda* (#:key inputs #:allow-other-keys) (setenv "PATH" (string-append (assoc-ref inputs "python") "/bin"))))): source expression failed to match any pattern
<pkill9>you need to add a name for the phase after 'build
<pkill9>e.g. (add-after 'build 'set-path ...
<anadon>pkill9: Now this makes even less sense! Package file: https://paste.debian.net/1144826/ CLI: https://paste.debian.net/1144827/
<pkill9>i don't know
<anadon>Ah crud. I'll pop in tomorrow about this.
<raghavgururajan>Hello Guix!
<anadon>Hello
<anadon>EST here, almost midnight...where are you?
<raghavgururajan>EST. At work now.
<raghavgururajan>Toronto
<anadon>Almost CMT
<anadon>What job has you this late on a Sunday?
<raghavgururajan>Concierge
<raghavgururajan>I work night shifts.
<raghavgururajan>8PM to 8AM
<anadon>I would not have figured someone in this channel would work concierge
<pkill9>why?
<edxsa>In nix and guix these dmenu-like programs start slowly. How did you solve it?
<anadon>pkill9: I'd expect an extreme bias towards paid programmers.
<str1ngs>what the heck, free software is not against being paid. heck you can sell free software even.
<str1ngs>raghavgururajan: hello fellow Canadian :)
<raghavgururajan>str1ngs 0/ :-)
<raghavgururajan>anadon So am I XD
<raghavgururajan>So if a package hard-code /usr/bin directory and I want to patch it. Should I use ~/.guix-profile/bin or /run/current-system/profile/bin ???
<pkill9>raghavgururajan: i would remove the absolute path and have it just reference the binary, so it automatically searches in $PATH
<raghavgururajan>pkill9 But here the application is not looking for specific binary. It searches the dir to show available programs.
<raghavgururajan>Like "open with" option that lists available progs.
<pkill9>i would have it search all directories in $PATH then
<pkill9>but if not
<raghavgururajan>So should I replace "/usr/bin" with "$PATH" ?
<pkill9>then i would use ~/.guix-profile/bin because that's the closest mirror to /usr
<pkill9>nah, it sounds like it wants a directory
<pkill9>i dunno what this program does, that would probably involve patching it
<pkill9>anyway, /usr is for user programs, so i would use the user's guix profile
<raghavgururajan> https://bin.disroot.org/?ac93f5a506fd15d2#5A64z7y2FkkAb7itHxeXxVeqeMdL7KYSBoeBpvkCiNwF
<pkill9>as that consists of 'user programs'
<raghavgururajan>Cool!
<raghavgururajan>Which is the correct syntax "~/.guix-profile/bin" or "$HOME/.guix-profile/bin" ?
<str1ngs>raghavgururajan: I'm PST Vancouver now though.
<raghavgururajan>str1ngs Ah I see. We should meet up sometime. Wdyt?
<str1ngs>raghavgururajan: I'm pretty far away. maybe if I'm Toronto though :)
<raghavgururajan>Yeah, that's what I meant by sometime XD
<str1ngs>I'm originally from that area, so sometimes I do travel there. :)
<raghavgururajan>I have not visited vancouver. So might come one day.
<raghavgururajan>Ah I see.
<str1ngs> It's nice here.
<efraim>perhaps not suprisingly, enlightenment doesn't like it when using enlightenment built against efl-1.23 and apps built against efl-1.24
<janneke>perhaps not ;)
<efraim>alright, it is suprising. I don't think that happens with most other libraries
<efraim>not really worth raising a bug upstream, they'll say it's not supported. I think I've seen it before in their IRC channel
***apteryx_ is now known as apteryx
<gf>hi, would it be ok to ask a question? I just installed guix,following guide, but guix install hello fails...
<gf>it shows ;;; In procedure git_libgit2_init: Function not implemented
<gf>guix/ui.scm:1936:12: In procedure run-guix-command:
<gf>error: canonical-package: unbound variable
<gf>but I appear to have all the dependency installed, like git, libgit2, guile-git ... what is missing?
<bricewge>gf: Probably guix wasn't properly build
<bricewge>Whihc instructions did you followed, from source, binary ?
<gf>Hi <bricewge>, its from script guix-install.sh
<gf>binary install,
<gf>its was on the manual and it automagically downloaded the binary with wget
<bricewge>gf: I can't really help you there I never used it.
<bricewge>What does “guix describe” returns?
<gf>ill try it, hold on a sec...
<raghavgururajan>error: failed to load 'gnu/tests/install.scm':
<raghavgururajan>ice-9/eval.scm:293:34: In procedure abi-check: #<record-type <operating-system>>: record ABI mismatch; recompilation needed
<raghavgururajan>That was during 'make' of cloned guix git repo.
<gf>it returns a dirty backtrace, should i pastebin it?
<bricewge>raghavgururajan: “make clean“ fix it for me most of the time
<raghavgururajan>Thanks!
<raghavgururajan>Do i have to run make again again after make clean?
<bricewge>gf: It can help if others try to look into it
<bricewge>raghavgururajan: Yes
<bricewge>gf: Your install of guix went wrong somewhere
<raghavgururajan>Cool!
<bricewge>gf: Which OS and service manager are you using?
<gf>ok, will pastebin, url is https://pastebin.com/DhHFc3Kw
<bricewge>Maybe try reinstalling it and looking closely to the exit codes
<gf>OS is linux 4.14.176, slackware-current, gcc 9.3.0, glibc 2.23 , exit code of command guix install hello was 1
<bricewge>gf: I mean the exit codes during the installation of “guix“ proper.
<bricewge>Not hello
<bricewge>gf: Are you using systemd?
<gf>no, slackware does not use systemd
<bricewge>You can try the manual installation of the binary https://guix.gnu.org/manual/devel/en/html_node/Binary-Installation.html#Binary-Installation (after the note section)
<gf>Ill try reinstalling guix manually, but first of all, is there a way to cleanly uninstall guix (from itself) or should I just manualy delete stuff ?
<gf>automated install with script does not seep to have any way
<bricewge>Then dug into the install script to see if it support your service management
<gf>ok, will do
<raghavgururajan>bricewge As your request been merged?
<bricewge>raghavgururajan: Nope. I'll take a look at yours though.
<raghavgururajan>bricewge Ah. Thanks!
<bricewge>How can stop guix to truncate it's error message?
<bricewge>s/can/can I/
<bricewge>raghavgururajan: Skimming trough it you are missing the entry in gnu/local.mk it should be in the commit that create the new file
<bricewge>raghavgururajan: What's the reason to replace git-fetch to url-fetch. It is my understanding that the former is prefered
<efraim>sneek: later tell civodul don't forget to add a news entry!
<sneek>Got it.
<efraim>sneek: botsnack
<sneek>:)
<civodul>Hello Guix!
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, efraim says: don't forget to add a news entry!
<civodul>efraim: a news entry for locale-libcs?
<efraim>civodul: for the authentication changes
<civodul>ah!
<efraim>assuming its ready for prime-time
<civodul>no!
<civodul>:-)
<civodul>it's still internal currently
<civodul>so only developers know about it
<civodul>nothing new happens when running "guix pull"
<civodul>i'd very much like more feedback before we get there, though!
<raghavgururajan>bricewge IIRC, git was intially used when release files were not published. Now they do.
<civodul>that reminds me i should add bricewge to .guix-authorizations before they push
<bricewge>raghavgururajan: girara's synopsis and description didn't changed.
<raghavgururajan>bricewge It would have been a cosmetic change :-)
<bricewge>civodul: Thanks I was waiting for it, I send https://issues.guix.info/41017 about that
<bricewge>Maybe it should by written in the commit access doc that one has to wait before pushing or that it should be done when adding the member in Savannah
<bricewge>civodul: BTW I've found out why my “guix deploy“ was hanging: remote-assertion don't support “uuid”, using a string instead.
<bricewge>But unfortunately it crash somewhere else now.
<bricewge>raghavgururajan: As I understand it 'git-fetch' is preferred over 'url-fetch' because it's content addressed which avoid build failing when an bold maintainer modify the tarball in place which end up in a hash mismatch.
<raghavgururajan>bricewge I see.
<bricewge>raghavgururajan: Actually there are no best practices on it https://lists.gnu.org/archive/html/guix-devel/2020-03/msg00084.html
<bricewge>Personally I prefer git-fetch, especially for the “guix build --with-commit” without any other modifications (dependencies or git-url).
<lprndn>Hello guix!
<bricewge>Hey lprndn!
<civodul>bricewge: oh!
<civodul>could you send the details about that uuid thing to bug-guix?
<civodul>ah you did already
<civodul>perfect
<civodul>ah no
*civodul is confused :-)
<civodul>i hadn't noticed your git-authenticate patch, i'll apply it
<bricewge>civodul: I'll open a bug for i t
<brendyyn>There were some recent patches to (guix openpgp), and i just got this: guix/openpgp.scm:138:0: error: (define ((port openpgp-invalid-signature-error-port) c) (condition-has-type? c &openpgp-invalid-signature-error)): source expression failed to match any pattern
<brendyyn>building with make clean, bootstrap,... coincidence? I think not!
<civodul>brendyyn: ooh, thanks!
<civodul>a bug
<civodul>you're on 2.2, right?
<brendyyn>yeah i think so
<civodul>3.0 doesn't catch that error, weird
<civodul>will push a fix
<brendyyn>thank you
<raghavgururajan>bricewge I see. I wanted it to be consistent with other pwmt packages. :-)
<gf>OK, I installed a new system, installed guix manually, and got a new error. is ci.guix.gnu.org down ?
<gf>substitute: guix substitute: warning: ci.guix.gnu.org: host not found: Servname not supported for ai_socktype
<gf>command following was guix package -i glibc-locales
<gf>I can ping google and others just fine
<bricewge>gf: Great. No it's up
<bricewge>gf: Same issue as you https://lists.gnu.org/archive/html/help-guix/2019-06/msg00111.html looks like it was a network missconfiguration
<gf>Humm... Pastebin of log is https://pastebin.com/raw/DykeQUZY
<gf>Ok, will lookinto network stack. but currently I see no problem in DNS resolving, host lookup ...
<bricewge>gf: Did you authorized ci.guix.gnu.org?
<kiwi_14>Hi folks, is there a known bug that installer in 1.1.0 hangs when connecting to a wifi network? In needed to reboot because system was unresponsive.
<bricewge>Normally you should get the packages binary (substitutes) from this server and not building all of them from scratch (except if it was your goal).
<bricewge>kiwi_14: Yes there is, can't find it ATM tho
<civodul>gf: if you installed manually, make sure to read https://guix.gnu.org/manual/en/html_node/Application-Setup.html
<civodul>in particular the bit on nscd
<civodul>which could explain the problem
<gf>Yes, ci.guix.gnu.org is authorized, it seems /etc/services was causing the error (I copied a good one from a different machine and now it appears to start downloading ... will report back if it works)
<bricewge>kiwi_14: It should bi fixed on master. So you can build an image yourself to workaround it
<kiwi_14>Thank you bricewge. Tethering from a phone and using wired should be faster option and I hope a working one.
<bricewge>kiwi_14: https://issues.guix.info/40682
<kiwi_14>Thank you bricewge ! That’s exactly my issue.
<kiwi_14>Fortunately tethering (wired option) worked without any issues so now installing is in progress.
<civodul>yay! https://guix.gnu.org/blog/2020/gnu-guix-maintainer-collective-update/
<civodul>welcome mothacehe, and thanks rekado! :-)
<kiwi_14>Putting btrfs everywhere also triggered the installer restart during a encrypted partitioning with a separate home.
<civodul>kiwi_14: hmm it looks like another bug
<civodul>not sure if that one is on file
<civodul>mothacehe has been working so we can provide snapshots of the ISO including bug fixes
<civodul>hopefully we can give you a new ISO that doesn't have these problems
<kiwi_14>I am only trying out for now to get a working xfce desktop. It is ok overall but installes loses progress so I need to start over after each hiccup.
<civodul>yay :-/
<kiwi_14>Also my fault for using non default options.
<civodul>well, they're here to be used
<kiwi_14>Chosing wired and ext4 should result in no hiccups I assume.
<civodul>yes
<civodul>i'm looking at clang: the clang-tidy executable takes 40M, 32M if we --strip-all, almost all is .text
<civodul>that's crazy
<civodul>looks half of LLVM gets inlined or something
*rekado plays with plymouth
<rekado>welcome mothacehe!
<mothacehe>Thanks for all your work rekado :) Hope to see you around !
<mbakke>apteryx: thanks for testing, are you working on fixing fakeroot?
<roelj>Who's familiar with running java in Guix? I get: java.security.InvalidAlgorithmParameterException: the trustAnchors parameter must be non-empty. Even with: -Djavax.net.ssl.trustAnchors=/gnu/store/...-icedtea-3.7.0-jdk/jre/lib/security/cacerts.
<rekado>roelj: you probably also need to provide -Djavax.net.ssl.trustStore
<kiwi_14>Now installation hangs on downloading gtk packages. Is there a way to resume the process?
<roelj>rekado: Should that point to <profile>/lib/security/cacerts or to something java-specific?
<kiwi_14>(Network still works. Just tested in an another tty)
<rekado>roelj: I think it should point to the cacerts directory
<rekado>roelj: also consider adding -Djavax.net.debug=all to see what it tries to do
<roelj>rekado: Thanks. Unfortunately, that doesn't seem to do the trick.
<brendyyn>thunderbird works! amazing stuff.
<rekado>brendyyn: do you mean Icedove? :)
<rekado>(it shouldn’t mention Thunderbird anywhere)
<brendyyn>i think i grabbed the wrong patch that is just a build of thunderbird
<rekado>the latest patch is here: https://issues.guix.gnu.org/40959#10
<gf>Humm... Clean reinstall again, auth added, network checked, good etc/services, and Im still getting a error...
<gf>guix package: error: while setting up the build environment: unable to make ?/? private mount: Invalid argument
<gf>is guix not supposed to work froma chroot ?
<kiwi_14>kmscon runs on 100% of CPU and installation hangs. I will start from scratch. Dunno how to recover.
<ruffni>where/how can i look up what each of the phases in build-system-gnu does?
<guix-vits>ruffni: idk. Is there anything: `ls ~/.config/guix/current/share/guile/site/3.0/guix/build-system/`?
<ruffni>there is. thanks!
<fishinthecalcula>Hi Guixers! I'm trying to run realtime JACK on the Guix System. I followed the manual instructions to obtain a suitable /etc/security/limits.conf but when I activate a profile built with manifest https://paste.debian.net/1144892/ , whenever I start Jack it crashes saying "JACK is running in realtime mode, but you are not allowed to use realtime scheduling [...]". What's strangest is that if I do `sudo jackd -d alsa` it
<fishinthecalcula>doesn't complain. Do you have any idea about how I could use realtime JACK as a regular user?
<sirmacik>I'm trying to setup name-service-switch according to current guix manual ut I het hosts: unbound variable
<sirmacik>I've got (gnu system nss) in my config
<sirmacik>but I get *
<sirmacik>sorry
<civodul>fishinthecalcula: maybe you need to add your account to the "realtime" group, no?
<civodul>sirmacik: could you paste your config?
<rekado>fishinthecalcula: you need to use the pam-limits-service to lift restrictions from users in the realtime group
<sirmacik>civodul: https://paste.debian.net/hidden/8dd203ff/
<rekado>fishinthecalcula: the Guix manual has an example in 8.8.1 Base Services
<rekado>(search the index for “realtime”)
<fishinthecalcula>civodul, rekado: I tried both to giving realtime permission on the existing audio group and on a new "realtime" group but I never managed to make it work. This is my current config https://paste.debian.net/1144891/ , I'm going to try again reconfiguring with the "realtime" group.
<fishinthecalcula>I'm supposed to log out and lo in again after reconfiguring?
<mmdev>I used guix to install sed on the new machine, but the build showed that the test failed. I checked the log and found that its selinux test failed. My machine does not have selinux. How should I ignore this error?
<guix-vits>mmdev: What distribution it is?
<mmdev>nixos
<mmdev>I tried switching to nixos, but found that nixos did not meet the needs of my development environment, so I installed a Guix on it
<guix-vits>mmdev: I'm not a developer, but: `guix describe`?
<mmdev> commit: aea6ab2f4ca060e68f8539cd612b0ce088627557
<mmdev> guix aea6ab2
<mmdev>I guix pulled the latest official channels, it seems that ci.guix.gnu.org did not have time to build a new sed?
<fishinthecalcula>So I reconfigured with https://paste.debian.net/1144895/ and rebooted my system. This is the output I get https://paste.debian.net/1144897/ . It seems that in some way the /etc/security/limits.conf is ignored by the system.
<mmdev>Sorry, I did n’t check the translation software, it ’s guix pull, not pulled
<fishinthecalcula>Anybody had the same problem? I'll keep investigating BTW
<TZander>mmdev: use `guix history sed` to see if ci built it yet
<TZander>no
<guix-vits>mmdev: strange. sed should be in %base-packages. i'm on guix a414d6a, sed works (FWIW).
<TZander>mmdev: use `guix weather sed` to see if ci built it yet
<mmdev>it display: 100.0% substitutes available (1 out of 1)
<mmdev> https://paste.debian.net/1144898/
<mmdev>It looks like ci.guix.gnu.org already exists, but why is my machine trying to build it?
<guix-vits>mmdev: did you authorized substitutes (as in Manual)?
<guix-vits>it's OFF by default for "on-top" installations.
<mmdev>I executed the guix archive --authorized < /path/to/xxx.pub command
***thelounge9924 is now known as brown121407
<gf>Im getting certificate errors, anyone know how to fix this ?
<gf>guix weather hello
<gf>guix/ui.scm:1936:12: In procedure run-guix-command:
<gf>X.509 certificate of 'ci.guix.gnu.org' could not be verified:
<gf> signer-not-found
<gf> invalid
<mmdev>I want to find a way to ignore it. Before I gave up Guix on Android, it was because some non-important errors at the time of the build destroyed the installation
<civodul>gf: see https://guix.gnu.org/manual/en/html_node/Application-Setup.html
<gf>civodul, guix install glibc-locales gives guix install: error: while setting up the build environment: unable to make ?/? private mount: Invalid argument
<civodul>gf: what does "uname -sr" return?
<gf>Linux 4.14.176
<gf>Its supposed to be lingterm kernel, is this a bad thing?
<mmdev>thanks, guix-vits, I think I found the error. I didn't use double quotes in --substitute-urls=mirror1 mirror2 , so it only recognized my publish machine, not ci.guix.gnu.org.
<gf>s/ling/long
<civodul>there are very old kernels that could be problematic but this one should be fine
<civodul>not sure
<apteryx>civodul: so, I pushed this yesterday night (early morning UTC time): https://git.savannah.gnu.org/cgit/guix/guix-artwork.git/commit/?id=ce5bf15ca2d0dc5fd65e72effaff48164253ae40 yet it hasn't appeared on the website yet. Did I miss something? :-)
<civodul>also what's "?/?"?
<civodul>apteryx: it did appear, try reloading the page :-)
<apteryx>ah, refreshing the page did it, haha
<gf>the strange thing is, ?/? is exactly as it says on screen...
<guix-vits>mmdev: ah, cool
<civodul>apteryx: that's because we still have that nginx timestamp issue: https://issues.guix.gnu.org/issue/37207
<civodul>nckx & g_bor[m] had proposals, we should give them a try
<apteryx>civodul: I see! I was thinking as well, I should probably make mention of that blog post on the mailing list. Would guix-help be appropriate?
<civodul>apteryx: maybe guix-devel + help-guix?
<civodul>thanks for taking care of all this
<apteryx>Alright, I'll post the announcement on both lists :-)
<guix-vits>gf: aside, about Kernel: i'd just tried 4.19 today, and had errors with nftables. There was some WA required to use 'type nat', and those WA not needed with newer version.
<gf>guix-vits: yea, netfilter stuff is said to change, at least the 4.14.x vanilla kernel I see no problem... had provelm with 5.x something tho so testing guix on 4.14 ..
<bricewge>Is there a way to display an guix error in a non-trucated way?
<bricewge>Or how can I debug guix? I can't manage to have a functional breakpoint in “guix repl”
<rekado>bricewge: the obscure COLUMNS variable is how Guile decides how to truncate backtraces
<rekado>I built an initrd with plymouth (not statically linked), but creating the disk image fails as Linux fails to “write” when uncompressing the initrd.
<rekado>it is 44MiB compared to 11MiB without plymouth.
<civodul>rekado: probably you need to pass qemu -m 2048 or so
<civodul>because the uncompressed initrd must fit in RAM
<rekado>aha, thanks
<bricewge>rekado: It doesn't change anything in my case https://paste.debian.net/plain/1144904
<rekado>bricewge: oh, I assumed you meant a different kind of truncation (e.g. in backtraces)
<bricewge>Maybe it wasn't truncated, but setting COLUMNS very low doesn't truncate at all
<mmdev>The same problem, I also want to know how to debug guix, always encounter guix errors for various reasons, but it does not output the log, I am difficult to find out where the problem is
<bricewge>Then how can I create a functional breakpoint on guix code?
<bricewge>I can “,br guix-deploy”, see that the trap in enabled but executing the procedure is done as normal
<mmdev>I think guix or guile should add a --show-backtrace feature. If something goes wrong, it can display and locate a line in a file, so it is easy to find the problem
<sirmacik>civodul: did you have a chance to look at that config?
*raghav-gururajan is back from work
<gf>Ive tried pretty much everything I can think of to try to get guix running but failed... Is there like, a chroot image in a tarball or thomething, that I can try to run ?
<bricewge>sirmacik: Your 'name-service-switch' isn't using the 'name-service-switch' record
<bricewge>sirmacik: See gnu/tests/ldap.scm for an example of it's usage
<ruffni>gf: on hardware? in a vm? i installed guix for qemu and it works like a charm....
<gf>on a forgin platform, like slackware, debian, etc...
<gf>hell, even gnu hurd runs like a charm on qemu
<bricewge>gf: There is a qemu image https://guix.gnu.org/download/
<bricewge>sirmacik: s/Your 'name-service-switch'/Your 'name-service-switch' field/
<ruffni>gf: are you trying to install guix within debian (in a chroot)?
<gf>slackware, in a chroot, but if debian is in need, i can try
<gf>is the chroot route not that common ?
<gf>I use it like on many software and it runs like gold
<ruffni>not sure, nevert tried it for guix. but since you might want to enjoy guix's system-wide capabilities (which suppose root) you might just want to look around and get a feel for it in a regular (qemu) vm
<rekado>mmdev: when “guix install” fails it shows you where the logs can be viewed
<rekado>mmdev: what other failures have you noticed where Guix doesn’t spit out a backtrace?
<gf>ruffni: I need to run guix, and trace the outout of java.scm, so I can get a grip on waht its doing in how to build build javasdk from jikes
<rekado>java.scm has no output
<ruffni>so why not get it to run on it'
<ruffni>s own?
<gf>ruffni: does java.scm run in a normal scheme interpreter ?
<rekado>gf: if you want to understand how we bootstrap the JDK from jikes I think the comment at the top of java.scm is helpful.
<rekado>gf: no!
<rekado>java.scm is a module providing variables that are bound to values of the <package> type.
<rekado>evaluating java.scm gets you nowhere near a build.
<gf>So where is the meat, in this situation...?
<rekado>I don’t know what you’re asking for
<rekado>can you try to clarify what you’re looking for?
<gf>Like, exact command line arguments passed to jikes/classpath/etc for rebuilding/bootstrapping all the way to java sdk
<rekado>(I wrote most of the Java bootstrap, so you can also ask me questions about that)
<gf>great!
<rekado>all the defaults are in the respective build system definitions
<rekado>jikes, for example, is built with the gnu-build-system
<rekado>no extra flags beyond the defaults
<rekado>you can compile the “derivation” with: guix build -d -e '(@@ (gnu packages java) jikes)'
<rekado>the derivation is the low-level representation of the build
<rekado>it’s *not* how I recommend learning about how the bootstrap works
<rekado>but this is ultimately what the guix-daemon processes
<rekado>if a derivation is mentioned whose outputs haven’t been built yet it gets built first
<rekado>the build script is the thing ending on -builder
<rekado>for my version of Guix I see in /gnu/store/2qg4hqnngx5r98bjj9h21h5sblq7hc57-jikes-1.22.drv that the builder is /gnu/store/9zi26bc4w2sl4ck7yalybvqp80spg292-jikes-1.22-guile-builder
<rekado>that’s a Guile script that invokes “gnu-build” (because of the gnu-build-system)
<civodul>sirmacik: it should read: (name-service-switch (name-service-switch (hosts ...)))
<civodul>the first 'name-service-switch' is the name of the operating-system field
<rekado>gf: gnu-build is defined in the module (guix build gnu-build-system).
<civodul>the second one is how you construct and <name-service-switch> record
<rekado>but really, you can ignore most of that.
<civodul>s/and/a/
<rekado>what matters is that jikes doesn’t need any fancy configure or make flags.
<rekado>classpath-bootstrap is different in that it needs the configure flags listed in the #:configure-flags slot of the arguments field.
<fishinthecalcula>Hello again Guixers, I'm trying everything I can but I can't seem to be able to start JACK.
<fishinthecalcula>I now tried this ( https://paste.debian.net/1144911/ ) ALSA config in my ~/.asoundrc but it still says that JACK cannot be run by my regular user.
<fishinthecalcula>I also noticed that ulimits -r returns 0 both with sudo and without. Is this supposed to happen after following the manual's Base Services instructions?
<rekado>fishinthecalcula: oh, it’s a bug! –> https://issues.guix.gnu.org/issue/37380
<fishinthecalcula>My system config is https://paste.debian.net/1144895
<rekado>a fix supposedly has been applied: https://issues.guix.gnu.org/issue/37405
<gf>rekado: Big thanks for explanation rekado, I will read the said file and try to follow the content...
<rekado>gf: the take-away here is that package definitions are written in a high-level DSL that relegates build details to build system abstractions
<rekado>users and developers rarely ever need to look at the “compiled” representation (the derivations and build scripts)
<gf>yes, It was/is highly abstracted (and hard to read)...
<fishinthecalcula>rekado: OMG thanks, I didn't think about that :D. Also sorry I have some network issues and I get all messages with like 3 minutes of delay :( I'll give a look at the bug report and the solution. Thank you again
<sirmacik>civodul: thanks!
<rekado>gf: easier to read than the derivations ;)
<rekado>(and it gets easier with practise)
<kmicu>rekado: Thank you for your amazing work and glad to see you are still here! ʕノ•ᴥ•ʔノ ︵λ𝛌𝚲𝝀
<bricewge>rekado: Yes, thank you for all your work on Guix!!
***nly` is now known as nly
<happycorsair>I've installed and initially configured Guix on my server for the first time. Seems very nice. Thank you, all the devs and users!
<sirmacik>now printing and scanning over the network works better than on my previous distro (:
<bricewge>Hooray! “guix deploy“ is finally working for me.
<bricewge>Time to fix/report those pesky bugs.
<sirmacik>anybody hosting guix on ovh vps machines? or mostly vultr users here?
<mothacehe>janneke: read your email, could you 'pk' the value of target in qemu-image?
<mothacehe>this is the one passed to gexp->derivation in expression->derivation-in-linux-vm, that should make everything cross-compile
<civodul>bricewge: yay! thanks for persevering, hopefully it'll be better as a result
<raghav-gururajan>Folks! I have a source that has "/usr/bin/cut" and "/usr/bin/ls" hard-coded. I want to patch them. Should I replace them with "/run/current-system/profile/bin/program" or just "program" ???
<guix-vits>raghav-gururajan: (which "ls")?
<raghav-gururajan>guix-vits But for that, programs should be present as inputs right?
<guix-vits>idk
<raghav-gururajan>I saw 'which' examples in our code base. What ever programs mentioned with 'which', were also present as input.
<guix-vits>raghav-gururajan: games.scm, 1828
<raghav-gururajan>So I believe 'which' uses PATH defined with-in build environment.
<raghav-gururajan>So 'that PATH' will not be same during run-time.
<guix-vits>in roguebox-adventures, bash isn't included in inputs
*raghav-gururajan checks games.scm
<janneke>mothacehe: ah, good idea
<janneke>mothacehe: ;;; ("qemu-image target=" "i586-pc-gnu")
<raghav-gururajan>guix-vits Yeah, It wasn't included. I wonder how that works.
<janneke>so, yea that looks alright
<guix-vits>raghav-gururajan: maybe like with services? Like, i've no tor installed, but tor-service-type works.
<raghav-gururajan>guix-vits I am confused.
<lprndn>guix-vits: not sure but bash (or ls) might be part of the base build environment?
<nly>hi janneke, In Emacsy, i'm cherry picking some commits off wip into master, hope it's alright :)
<raghav-gururajan>Ah yeah ^
<raghav-gururajan>part of build-system
<raghav-gururajan>*build env
<efraim>i586-pc-gnu isn't a qemu-target
<efraim> https://git.qemu.org/?p=qemu.git;a=blob;f=configure;h=6099be1d848c7f52ea02694d6d86583093de440f;hb=b0ca999a43a22b38158a222233d3f5881648bb4f#l7576
<efraim>i've been looking at qemu for a bit to smallerize qemu-minimal
<ruffni>i've successfully built my first package! yay! but the binary doesn't seem to be in my PATH. do i have to add it after install phase?
<janneke>nly: go for it!
<guix-vits>raghav-gururajan: but so long, as some package refer to "this bash" it will not be `gc`ed...
<raghav-gururajan>hmm
<guix-vits>lprndn: idk.
<guix-vits>ruffni: what package?
<lprndn>ruffni: `guix build only builds the package and outputs its store path. If you want it to available in your profile, you need to `guix install it ;)
<lprndn>*to be
<sirmacik>can someone remind me how to export/import list of packages for guix user profile? Is there a way to configure it like /etc/config.scm?
<guix-vits>sirmacik: 4.2 Invoking guix package (--manifest)
<sirmacik>does it synchronize profile with manifest file or only add missing ones?
<guix-vits>"overwrites", i rather say
<guix-vits>"Create a new generation of the profile .... This option can be repeated several times, in which case the manifests are concatenated."
<sirmacik>so if it's concat than it adds without removal of missing list elements
<ruffni>can i install the locally built package with 'guix install'? does "guix install" even do more than just set the PATH to the freshly generated derivation?
<mothacehe>janneke: ok, thanks. The problem is deeper then, can't see where though.
<raghav-gururajan>lprndn Do you know what should I do with my situation?
<raghav-gururajan>I have a source that has "/usr/bin/cut" and "/usr/bin/ls" hard-coded. I want to patch them. Should I replace them with "/run/current-system/profile/bin/program" or just "program" ???
<lprndn>ruffni: yes! If the package is already built and available locally, guix install will set up a few symlinks and update PATH. maybe more (man, stuffs..) but I don't know exactly what.
<guix-vits>raghav-gururajan: Aren't "/run/current-system/profile" is a system-wide things? I see no emacs in it's ../bin.
<guix-vits>but nano is there (which is in %base-packages)
<raghav-gururajan>guix-vits Yeah system-wide.
<lprndn>raghav-gururajan: I think (which "ls") is probably the right thing to do anyway.
<janneke>mothacehe: np, thanks!
<mothacehe>I might have an idea, answering your email!
<janneke>mothacehe: i have been "pounding" on this for quite some time (not knowing it was here-abouts), from within the qemu-image area
<janneke>ah!
<lprndn>raghav-gururajan: About adding `coreutils (is `cut part of it?) to inputs, I don't know what's prefered in guix. Are those links used at build time only?
<guix-vits>at run-time
<ruffni>lprndn: guix install pkg-name and guix install pkg-path both return: "guix install: error: pkg-name: unknown package"
<lprndn>ruffni: how did you build your package? Using the same command but switching from `build to `install should be ok.
<ruffni>guix build -f filename
<ruffni>guix install -f : unrecognized option
<guix-vits>ruffni: cd where-filename; guix package -L . -i pkg-name
<lprndn> ruffni: sorry thought -f was an option of install too. try `guix package -f
<guix-vits>*if that is module*
<Misha_B>I'm installing guix right now and I gotta say, the new graphical installer is pretty slick.
<lprndn>raghav-gururajan, guix-vits : I would add `coreutils to inputs then, with a comment why. Worst case, you'll have to remove it at review time...
<ruffni>aaaand it works! i moved ibniz.scm to a new folder and did what guix-vits told me to. is it necessary to create such a directory?
<guix-vits>ruffni: idk, but i try to not run with '-L', if in subdirectories there is a lot of other .scm.
<ruffni>that's what i observed, and that's why i wondered :)
<sirmacik>guix-vits: thanks (:
<ruffni>is there a guide or some instruction to what part of a description is the synopsis and which part the description? is there a length limit to them?
<raghav-gururajan>lprndn I added coreutils to inputs. :-)
<raghav-gururajan>lprndn No, not at build-time. Used at run-time.
<bavier>ruffni: the manual has a brief summary of what's appropriate for each
<ruffni>found it, thanks!
<NieDzejkob>Yeah, adding coreutils to inputs is correct here.
<NieDzejkob>Hmm, now I'm not sure anymore... I think I see (which "ls") more often
<guix-vits>sneek: should we add coreutils to inputs?
<guix-vits>sneek: seen coreutils?
<sneek>Sorry, no.
<g_bor[m]>Hello guix!
<user_oreloznog>o/
<sirmacik>is emacs-next package the code that already uses native code presented at European Lisp Symposium?
<g_bor[m]>Congratulations to our accepted Outreachy interns.
<g_bor[m]>Now this is time for celebration :)
<guix-vits>+1
<bandali>sirmacik, i don't think so. i think the native comp stuff is still sitting on a separate branch on the emacs repo
<sirmacik>oh, so still some more time we have to wait :/
<bandali>yes :-)
<sirmacik>I see that now, it's on feature/native-comp branch
<g_bor[m]>I would also like to encourage the applicants not selected this round to stay around, as they well might be selected for the next round.
<bandali>apteryx, any news about the issue you were following up on with the emacs devs?
<g_bor[m]>Currently the Outreachy site is down for me. I got notified via mail.
<apteryx>bandali: yes, a workaround was suggested and a commit made to Guix master to fix the use case we had with Emacs 27.
<bandali>apteryx, awesome! in that case i will soon bump emacs-next to a recent commit on the emacs-27 branch
<raghav-gururajan>YAAAAAAAAAAAY!!!!!!!!!!!!!!!!!
<bandali>:-)
<apteryx>sounds good.
*raghav-gururajan got selected as intern
<bandali>i'm still undecided as to whether emacs-next should follow emacs-27 or master
<efraim>congrats raghav-gururajan!
<raghav-gururajan>Thank you g_bor[m] :-)
<bandali>technically speaking, emacs-27 is the "next" emacs release
<raghav-gururajan>efraim Thanks!
<sirmacik>bandali: it sounds like some kind of emacs-git package should follow master
<Misha_B>I'm getting this error: 'GC Warning: pthread_get_attr_np or pthreadd_attr_getstack failed for main thread\n GC Warning: Couldn't read /proc/stat'
<Misha_B>when booting up
<apteryx>bandali: I think emacs-27. In my view, it's a playground to migrate from our current emacs version to the next.
<bandali>sirmacik, apteryx, ha. then maybe emacs-trunk or something for master
<bandali>raghav-gururajan, congrats!
<raghav-gururajan>bandali Thanks!
<bandali>cheers :-)
<apteryx>bandali: We usually only package released software in Guix. I'm not sure why Emacs should be different. Users can always 'guix build emacs-next --with-git-url=emacs-next=[url-of-git-repo]'
<sirmacik>wow, I didn't know that
<sirmacik>does it work also for some stable software built normally from tarball?
<sirmacik>so I would add that argument and build it from git repo
<bandali>apteryx, that's a fair point, i guess. as a contributor and maintainer of part of emacs, i personally the master branch as my daily driver, but i guess not many others do
<raghav-gururajan>How do I specify quote with-in quote? I want to add 'one double-quote' inside string-append. Like (string-append """ out).
<bandali>s/personally/personally use/
<apteryx>sirmacik: I guess it hould. info '(guix)Package Transformation Options'
<raghav-gururajan>So is it (string-append "\"" out) correct?
<Misha_B>image of boot screen https://imgur.com/BnKI4Rd .
<apteryx>s/hould/should/
<bandali>apteryx, i think once emacs-27 is officially out, i will point emacs-next to master and periodically update it, until the branch for the next release (e.g. emacs-28) is cut, and so on
<sirmacik>apteryx: omg, that's huge. thanks, trying it out right now
<apteryx>bandali: I think we should work torward moving our base package emacs package version toward emacs-27 after it's released officially.
<janneke>mothacehe: thanks for the $/+ and (with-parameters stuff; it seems to be unrelated to the hurd cross build error i am seeing, though
<apteryx>I started testing that way, some package breaks, but at least for those who are maintained they'll get patched
<mothacehe>janneke: yw! Oh I hoped it would solve it :(
<bandali>apteryx, agreed. i was talking more about the fate of emacs-next once emacs-27 is out and guix's emacs is switched to 27
<bricewge>Congratulations raghav-gururajan!
<mothacehe>janneke: I'm grepping 'hurd' to search new culprits :p
<raghav-gururajan>bricewge Thank you!
<janneke>mothacehe: i /just/ learned about +/$ from civodul two days ago and we found a bug in system/vm.scm
<janneke>i didn't look at this code after that, it sure looked fishy
<Misha_B>how do I envoke --repl for initrd repl?
<mothacehe>there's a somehow similar pattern in system-qemu-image
<apteryx>bandali: OK. Then we are in perfect agreement :-)
*janneke looks
<apteryx>raghav-gururajan: congrats :-)
<bandali>awesome
<janneke>ah, yes the (with-parameters -- i had that /all/ over the place before i gave up and re-started :-(
<mothacehe>Misha_B: oh too bad, I think you can pass add --repl to the kernel command line from Grub
<mothacehe>Misha_B: but I'm afraid it won't help you much
<mothacehe>Misha_B: what's the configuration of your machine (GPU?)
<Misha_B>RX 560
<Misha_B>it does say that that fails when booting up the installation medium
<Misha_B>amd gpu fails that is
<janneke>hmm, maybe extra-directives should be cross-built later, instead of adding this kludge in system-qemu-image
*janneke looks for a + deeper-on
<Misha_B> https://paste.debian.net/1144956/
<Misha_B>here's the dmesg for gpu
<mothacehe>We are using uvesafb to work around faulty amd firmware. But it mail fail for you.
<mothacehe>You are using the 1.1.0 ?
<Misha_B>yes
<mothacehe>Ok, thanks. We are also passing "modprobe.blacklist=radeon" to the kernel. Could you boot, hit 'e' in Grub, then remove this bit and hit 'Ctrl-X'?
<mothacehe>If this does not work, it would be really nice if you could file a bug at "bug-guix@gnu.org".
<mothacehe>With the dmesg output attached :)
<mothacehe>janneke: I think extra-directives with (file-append hurd) is fine but I'm really not sure, it's maybe a good suspect.
<mothacehe>janneke: one way I debug this kind of cross-build mess, is to grep the wrong derivation in the whole store
<mothacehe>(the one that should be cross-compiled but is not)
<Misha_B>ok, it seems like it didn't work, so I'll send the bug report
<mothacehe>thank you! don't you have another display port using the motherboard graphic chip instead of the one of the AMD GPU?
<Misha_B>I do have an integrated gpu, I'll try setting that as the primary gpu and booting.
<mothacehe>oh cool, this way we'll know if the AMD GPU driver is to blame
<str1ngs>sneek: later tell nly, thanks for merging emacsy.
<sneek>Okay.
<ruffni>i'd love to submit the package to the repo, but i don't know which scm fits best... it's a minimalistic programming language/vm for audio video live-coding...
<Misha_B>seems to be working, the installation was in the normal size font instead of the large blocky sized font
<mothacehe>Misha_B: you're restarting the installer?
<Misha_B>yeah
<mothacehe>don't you try to boot the installed system ?
<janneke>that sounds smart...so: build of /gnu/store/9s621nadr6xcj74n1dl4c7gkvn8x0179-hurd-0.9-1.91a5167.drv failed
<janneke>grep 9s621nadr6xcj74n1dl4c7gkvn8x0179-hurd-0.9-1.91a5167 /gnu/store/*.drv
<janneke>...but that's empty!
<janneke>(it works great for using any dependency of the failing drv; many hits...)
<janneke>hmm
***calher is now known as KE0VVT
<janneke>thanks for helping, and helping me think in new directions!
<janneke>...and learning new tricks
<mothacehe>strange! you're welcome :)
<mothacehe>have to go, see ya
<raghav-gururajan>apteryx Thanks! :-)
<happycorsair>Guys! How can I deal with the Guix bug tracker? My e-mails are visible at https://issues.guix.gnu.org as file attachments. Is that a problem of my email web client?
<cbaines>happycorsair, what mail client are you using to send the emails?
<happycorsair>cbaines: just the web UI provided by e-mail service
<cbaines>happycorsair, right, that web UI seems to be sending webpages (HTML) rather than text for the email
<ruffni>what do i need to set the guix git repo up?
<happycorsair>cbaines, ok, thank you. Are there any web-based tools for guix bug tracking? I've found https://issues.guix.gnu.org and https://debbugs.gnu.org. I've made a comment by the first one, but it seems to be lost. The second one provides no way to write a message. So, pure email is the only way to go?
<ruffni>whoopsie. what do i need to run after freshly cloning the guix git repo to have ./pre-inst-env and such working?
<cbaines>happycorsair, yeah, I'd suggest using email. Maybe there's an option in the web client your using to send emails with text in
<cbaines>ruffni, there's some guidance here https://guix.gnu.org/manual/en/html_node/Building-from-Git.html#Building-from-Git not all the steps are necessary
<cbaines>You can skip running make authenticate
<Misha_B>it works, the installation boots up just fine
<happycorsair>cbaines, thanks a lot
<Guixguy>Misha_B: I have a 5700XT and I have to use nonfree firmware to get amdgpu working properly. Or else I get a frozen screen at boot as it's loading kernel modules because it doesn't know how to pass the graphics stack to the next step
<bricewge>Yes!!! My GSoC proposal has been accepted!
<bricewge>We'll have Guix booting form (i)PXE in a few months.
<apteryx>bricewge: amazing! :-) and congrats!
<g_bor[m]>Congratulations!
<bavier>bricewge: great!
<janneke>bricewge: \o/
<ryanprior>Is there a way to specify different sections of a manifest file? Like suppose I have one set of packages for dev & testing, and another smaller set for running the thing. Do I need two .scm files or can I write one that covers both cases?
<jonsger>bricewge: congrats!!!
<vagrantc>bricewge: nice!
<vagrantc>bricewge: that'll be one item i can just take off my back burner that i was probably never going to get to anyways :)
<Misha_B>Guixguy: the amdgpu-pro firmware? is there any special steps you have to take to install it?
<Misha_B>ah he's gone
<bricewge>Thank y'all, I hope I'll deliver good software.
<apteryx>bricewge: I'm not too worried! I've been impressed by the quality of your work thus far :-)
<apteryx>civodul: seems we forgot to update: https://guix.gnu.org/about/
<civodul>apteryx: oops, indeed
<bavier>ooo, inkscape 1.0
<civodul>apteryx: do you want to do it?
<apteryx>civodul: sure!
<sirmacik>what is the naming convention for making package from git repo (not release tag)?
<apteryx>sirmacik: 'info (guix)Version Numbers'
<apteryx>there's an example
<civodul> https://guix.gnu.org/manual/en/html_node/Version-Numbers.html
<civodul>:-)
<sirmacik>apteryx: thx (:
<apteryx>civodul: pushed! It's neat that the website automatically rebuilds itself from Git :-)
<civodul>yup!
<civodul>thank you!
<civodul>oh i think we need to update the guix-maintainers alias, too
<civodul>lemme see
<apteryx>oh, right
<efraim>it looks like guile-3.0-static on core-updates builds with gcc-5. not sure it's worth making one for bootstrapping and one for the initrd
<lispmacs[work]>hi, I noticed what looks like new effort for powerpc support. Was just curious what compute platforms are being targetted in that effort
<sirmacik>hm, next question, what to do when ./pre-inst-env gives me error about no guile interpreter?
<cbaines>sirmacik, what's the specific error?
<cbaines>could it be that the guile entry in the store it's trying to use has been removed?
<sirmacik>~/software/guix % ./pre-inst-env guix package -i next --with-git-url=next=https://github.com/atlas-engineer/next.git
<sirmacik>./pre-inst-env: /home/sirmacik/software/guix/scripts/guix: /gnu/store/p2nhiskk7ykv1ba704qhk1whwc2rbwyj-profile/bin/guile: bad interpreter: No such file or directory
<sirmacik>it probably got updated, but how do I refresh my environment for test builds?
<cbaines>I believe running make, or maybe ./configure (probably with --localstatedir=/var ) will update it...
<sirmacik>configure: error:
<sirmacik>No Guile development packages were found.
<sirmacik>o.O
<sirmacik>oh, looks like I've fixed it
<sirmacik>sorry for bothering with those pasted lines
<civodul>rekado: did you see that patch for network boot? https://issues.guix.gnu.org/issue/41011
<civodul>could be the beginning of something like you had in mind, no?
<vagrantc>grub's network booting is brutally slow
<civodul>ah
<civodul>maybe good enough?
<vagrantc>good enough for?
<vagrantc>i'd recommend ipxe or pxelinux from syslinux
<vagrantc>for network booting ...
<rekado>having URL support for the initrd would be useful
<rekado>I wonder how this relates to GRUB…?
<vagrantc>#41011 [PATCH] gnu: grub: Support for network boot via tftp/nfs.
*rekado just finished reading it
<vagrantc>rekado: what do you mean by "URL support for the initrd" ?
<rekado>what I had in mind was to tell the initrd to load the rootfs over NFS
<rekado>this: https://www.kernel.org/doc/Documentation/filesystems/nfs/nfsroot.txt
<vagrantc>right
<vagrantc>a writeable rootfs?
<rekado>I didn’t think much about it. Before I could I decided that for testing we would first need an NFS service, and that’s where I spent all my time
*vagrantc has often wondered about a shared /gnu/store between multiple machines that were effectively read-only machines
<vagrantc>guess you'd need to share /var/guix as well
<rekado>sharing /gnu/store between machines is exactly how we do cluster installations
<rekado>and sharing /var/guix as writeable
<vagrantc>how do you prevent one machine from clobbering files in /var/guix ?
<vagrantc>or, they're all supposed to be identical, so they don't do updates in-place?
<rekado>the localstatedir must be writeable for symlink updates when profile generations are changed
<vagrantc>right
<rekado>generally users can only write to their own directory anyway
<rekado>with profile locking I don’t think clobbering files is a real concern. It hasn’t come up in 5+ years – and much of that was before profiles were locked
<vagrantc>interesting
<rekado>maybe the users here just weren’t inquisitive enough to attempt to break things
<vagrantc>yes, a hostile environment might be more ... exciting :)
<rekado>recent comments about NFS being slow (which I share) made me consider iscsi again
<rekado>I wonder about consistency across clients
<rekado>but maybe that’s no problem
<vagrantc>i've only worked with NFS almost exclusively read-only, and similarly with NBD
<vagrantc>i've been curious to explore a shared guix rootfs read-only, with overlay fs on tmpfs or some other transient storage for the machine differences
<rekado>by consistency I mean: when the single node that writes to the iscsi disks syncs its file system, will the other nodes see inconsistent state before that?
<vagrantc>right
<vagrantc>cephfs might answer that...
<rekado>a few years back I played for just a few hours with gluster and lustre (?) and decided that I don’t want that kind of complexity; and it didn’t improve performance over our NFS setup.
<rekado>I even played with a disgusting inotify abomination intended to keep /gnu/store in sync across machines, but that was a pretty bad idea.
<rekado>there’s got to be a distributed file system optimized for this use case: one writer, lots of readers
*rekado resends a few emails that were stuck in the issues.guix.gnu.org queue…
<cbaines>I think the optimised bit is more about taking advantage of immutability
<rekado>quick update on plymouth: nothing works :)
<cbaines>You could make some things more efficient if you assume that store items don't change
<rekado>cbaines: oh yes, that too
<rekado>that’s why I thought inotify might be nice
<rekado>but it turns out that watching big directories is kinda expensive
<cbaines>It would be nice if you could listen to the guix-daemon about changes to the store
<rekado>I can start plymouthd, and perhaps also send commands to it, but I can’t see a damn thing happening.
<rekado>yeah
<civodul>rekado: nothing works... even with -m 543545?
<civodul>the solution to many of our problems these days :-)
<rekado>civodul: I increased the default memory size for building the image from 256 to 512, so the VM could be built.
<rekado>but the effects of running plymouth are invisible
<rekado>I’ll keep playing with this and reading documentation
<rekado>maybe I’m doing something silly
<NieDzejkob>huh, why are disk images built in VMs?
<civodul>NieDzejkob: because on GNU/Linux one can't mount a file system without being root
<civodul>but this is about to change
<civodul> https://issues.guix.gnu.org/issue/40955
<vagrantc>civodul: i was wondering if it would be possible to use user namespaces to fix that issue...
<civodul>vagrantc: we could, but we don't assume they're supported in build environments
<civodul>also, what Mathieu proposes above sidesteps the issue entirely
<cbaines>There looks to be an issue with #:file-creation-mask reconfiguring bayfront
<cbaines>I think this is a new Guix with not so new shepherd issue, right?
<civodul>oooh, yes
<civodul>so if you load the new service on an old shepherd, you get wrong-type-arg
<civodul>how does that manifest exactly?
<rekado>cbaines: yes, saw that on ci.gux.gnu.org as well
<rekado>and on my i686 ‘server’
<cbaines>civodul, I don't know the specifics, just reading guix-sysadmin
<cbaines>rekado, did rebooting help?
<rekado>haven’t rebooted yet
<cbaines>I guess that issue only impacts the syslog service...
<vagrantc>wow, genimage sounds amazing.
<vagrantc>oh, I see lots of pretty UTF-8 boxes after updating to core-updates ...
<vagrantc>i tried "fc-cache -rv" and logging out, rebooting, etc... to no avail
<vagrantc>not all programs, and not even all parts of all programs...
<mbakke>vagrantc: do you have any fonts installed? if not, maybe try installing one?
<mbakke>I suspect a default font got lost somewhere, but don't know where or why
<alextee[m]> https://www.gnu.org/philosophy/saying-no-even-once.html
<bandali>^ new rms article
<rekado>with more ‘display’ debugging I can confirm that plymouth actually runs and responds to commands. It just doesn’t actually show a splash screen or progress bar at all.
<rekado>maybe something wrong with the package
<rekado>more tomorrow
*rekado -> zzz