IRC channel logs


back to list of logs

<kurisu>but it doesn't says so here : and here :
<kurisu>my second issue is a non functional gajim executable, i got a python error
<simonsouth>Anyone else having trouble accessing (and
<Aurora_v_kosmose>simonsouth: Works from Tor for me.
<simonsouth>Aurora_v_kosmose: Thanks. I think it's a problem on my end.
<rekado>kurisu: what does “type guix” say?
***fever is now known as Guest34566
<kurisu>rekado_: /usr/local/bin/guix
<rekado_>kurisu: that’s not the Guix that “guix pull” creates
<rekado_>that would be ~/.config/guix/current/bin/guix
<rekado_>try adjusting your PATH and things should work fine again.
*rekado_ —> zzzZ
<kurisu>thx, i'll try that
<Guest34566>I'm having trouble setting up mpd. I want it to run as my user and so in config.scm i have put
<Guest34566> (user "fever")
<Guest34566> (music-dir "/mnt/home/fever/Music")))
<Guest34566>after reconfiguring, mpd fails to start, printing "Service mpd is currently disabled"
<Guest34566>Apologies for spamming the chat, is there a way to debug why the service is disabled? the man page for "herd" gives no such indication.
<leoprikler>That's actually a known issue. You should still be able to start mpd manually, though.
<TBC_x>Hello, how is LVM support in Guix?
<Guest34566>Ah thanks.
<stikonas[m]>TBC_x: lvm is used for some things, so it works. E.g. rust...
<TBC_x>I remember, last time I've checked, it wasn't there.
<TBC_x>I want to use LVM device for root partition, will see if that works today
<Aurora_v_kosmose>stikonas[m]: I think he means LVM devices, rather than LLVM
<stikonas[m]>Oh, indeed, my mistake...
<iyzsong-w>i think a public guix hurd instance would be fun, in the spirit of
<snybajl[m]>Does GuixSD unlock luks volumes before or after GRUB? AFAIK there is no AES-acceleration in GRUB so it's a bit slower. Evil-maid attacks aren't really part of my threat model, so I would prefer to have it unlocked after GRUB. Is it possible to set that up?
<mroh>TBC_x: a lvm root partition works w/o trouble (even with lvm-cache...). What might be problematic is booting from it, because I'm not sure if our grub has the lvm module configured. (my grub config is an organic giant hack, grown over decades with lots of distros...;)
<apteryx>snybajl[m]: you're asking if Guix System supports a non-encrypted /boot. The answer is no, not at this time.
<apteryx>(it supports a fully encrypted drive, that prompts once in GRUB and once when the kernel boots)
<CarloSvarga>Hi, Guix noob here.  I am considering building an internet-facing router with Guix instead of Ubuntu, but I don't really see any security stuff for either Guix servers or internet-facing machines.   Are there any security issues unique to Guix in such a case?
<snybajl[m]>@apteryx gotcha, does it support lvm+luks? (for unlocking multiple drives using the same password)
<sturm>Hi folks, is it possible to run the Guix System graphical installer on top of a running Guix System? This would be useful to allow you to partition and install onto a second hard drive rather than messing around with a bootable USB.
<apteryx>snybajl[m]: I don't use lvm, it used to not support it, but ISTR it was added recently (to verify)
<apteryx>support for it was added with commit a9a2fdaabcc78e7a54d9a6bcfa4ee3de308e9a90; there's a news entry for it (guix pull --news)
<apteryx>lvm-device-mapping can be used to declared mapped devices for LVM.
<apteryx>this was 2 months ago
<apteryx>anyone knows the differente in (guix depreciation) between define-deprecated vs define-deprecated/alias? Both seems to support declaring a variable as obsolete with a replacement.
<apteryx>the difference*
<apteryx>I think the later supports deprecating macros as well, while the former is only for variables or procedures, if the docstring is accurate
<apteryx>also the former accepts the full procedure signature to be passed as arguments, while the later is just old-binding -> new-binding
<raghavgururajan>Currently, while building tdesktop, I get this error (
<raid5atemyhomewo>So... ZFS on Guix. What else can I do to help get it in?
<raghavgururajan>New diff file is the following, which is to be applied on top off master (efa773f94a18b40f2c63795f364ae87dade76f60).
<raghavgururajan>Any ideas on how to rectify this error (
<TBC_x>That's great! I love the concept. But I couldn't get far because I put all my stuff on LVM
<raghavgururajan>sneek, later tell apteryx: Would you be able to review and push #45721 v4 (1 to 20, as 21 is not ready yet)?
<sneek>Got it.
<Aurora_v_kosmose>raid5atemyhomewo: I'd probably start with taking a look at if DKMS & modules using it exist.
<Aurora_v_kosmose>It'd probably provide a starting point.
<Aurora_v_kosmose>Hm, my searches are coming up with nothing. That's... "fun".
<raid5atemyhomewo>the existing ZFS package doesn't use DKMS to compile, it uses guix build system to compile
<raid5atemyhomewo>I have patches to enable ZFS on Guix in, but nobody is responding
<raid5atemyhomewo>what can I do so it's easier to review and people can actually respond or merge it?
<raid5atemyhomewo>including bugfixes, a convenient way to install ZFS, integration so ZFS works the way it does on other OSs, etc
<raid5atemyhomewo>they're all there, just need review and merge
<Aurora_v_kosmose>Talking about it is a good step for getting it attention. As for review, I think it's simply a matter of timing with work/vacations & such.
<Aurora_v_kosmose>Testing kernel patches also require a bit more setup/time.
<raid5atemyhomewo>it's kernel module, not quite kernel, but I guess
<raid5atemyhomewo>so it would go into staging? core-updates?
<Aurora_v_kosmose>Not quite sure. Did you poke guix-devel?
<raid5atemyhomewo>a few teims
***iyzsong-- is now known as iyzsong-w
<mothacehe>hey guix!
<efraim>just tried updating libreoffice, wow is that frustrating
<abcdw>hi guix!)
<PurpleSym>I’m having a weird issue: After `guix gc`’ing `guix environment` is stuck requesting a store item over and over again, but the daemon always replies “path … is not valid”. If I manually `guix build` it, it moves on to the next one.
<abcdw>In ~3 hours will do a stream about Guix System Services, Shepherd and will explain what happens during system reconfigure. Will try to read this chat during the stream, will appreciate all questions and suggestions.
<raid5atemyhomewo>So. ZFS on Guix. What are the chances my patches for that will get some attention? What can I do toget those patches merged?
<Ikosit>Hi there! I recently switched from NixOS to GuixSD. One thing i miss so far are overlays. Is there a mechanism in guix to override a specific package, so that all packages, which depend on that package will use the new package? I don't like using my own guix package repository, because then overridden packages were mixed with normal ones, also i'd have to pull with git and guix to update my system
<PurpleSym>Ikosit: You can try the --with-input transformation when installing a package, but afaik global overrides do not exist.
<rekado_>Ikosit: you can automate this with the Guix API in a manifest file.
<leoprikler>raid5atemyhomewo: I think you've done a lot of work recently, which has already been noticed by many. Now give people some time to do proper reviews.
<raghavgururajan>Who's here a fan of Theadore Bagwell?
*raghavgururajan posted in wrong chat
<roptat>hi guix!
<mdevos>hi sneek! Any messages?
<raghavgururajan>roptat, o/
<roptat>so, now that we have our project at the fedora weblate, I'd like to advertise it as the new way to contribute translations. What should be the steps to do that?
<roptat>there are a few mentions in the manual
<apteryx>civodul: is there a reason why using define-deprecated/alias would require the Guile modules to be compiled, else it wouldn't be expanded where it's used?
<sneek>apteryx, you have 1 message!
<sneek>apteryx, raghavgururajan says: Would you be able to review and push #45721 v4 (1 to 20, as 21 is not ready yet)?
<roptat>there's also the Makefile that fetches translations from the TP, it should fetch from weblate now
<roptat>I guess I should also contact the TP to ask them to remove/disable the guix project
<apteryx>I have this in my tex.scm: (define-deprecated/alias texlive-union texlive-updmap.cfg). I have to run 'make' every time I touch the module, else it throws: Wrong type to apply: #<syntax-transformer texlive-union>
<Ikosit>rekado_: Where should i use the manifest?
<civodul>roptat: re translation, Florian wrote a blog post a few months ago and was waiting for the process to be streamlined to publish it
<civodul>so i guess now's a good time!
<civodul>you can find it in the patch tracker i think
<roptat>civodul, thanks
<civodul>it's great news anyway, thank you
<civodul>apteryx: that means something's referring to texlive-union before it's defined
<civodul>or referring to it via @@ or similar
<civodul>(and thus not getting it macro-expanded)
<roptat>civodul, I can't seem to find it, too many messages ^^'
<apteryx>civodul: I see! Thanks.
<civodul>roptat: found it!
<roptat>ah, thanks!
<roptat>I guess we could make a blog post for translations in general, because we have all the po files at weblate now: the website, guix, guix-packages, the manual and the cookbook
<roptat>I'll send patches later to change the manual and the Makefile
<roptat>then we can discuss further action
<civodul>roptat: awesome
<civodul>i guess a variant of Florian's post specifically about the web site would also be welcome
<civodul>interestingly we received translations even before publishing instructions :-)
<raghavgururajan>Folks! Do we have git-fetch option to download submodules as well?
<bavier[m]>raghavgururajan: it's the `recursive?` field.
<raghavgururajan>bavier[m]: Ah thanks!
<civodul>roptat: woohoo!
<roptat>although, don't use download-po too often, or you'll end up with "erreur 429 : Too Many Requests."
<roptat>another solution would be to get the files from the repository instead of the weblate
<civodul>does weblate send email notifications like the TP does?
<civodul>that's usually when i run "download-po"
<roptat>no, but it commits to the repository (currently at
<roptat>so if you subscribe to notifications there, you'll get them
<roptat>once we have the repo at savannah, we will be able to do the same there
<efraim>in magic news today, alt+tab fixed itself today for me on enlightenment/wayland. It used to be if I took too long the box would linger until I closed all the windows
<civodul>roptat: so a bot will have ssh access to our repos at Savannah?
<efraim>I spoke too soon, it's back to being stuck
*civodul uses C-x b instead :-)
<roptat>civodul, yes, that's why I'm requesting a new project
<roptat>and not just a new repo
***stikonas_ is now known as stikonas
<civodul>roptat: oh got it, pheww :-)
<apfel>hi there, i am playing around with with g expressions. I have the problem that gexp->script does not produce the script, even though the output tells me something else ... any ideas?
<mothacehe> apfel: hello, that's because you need to build the derivation, this way:
<apfel>mothacehe: this works. Thanks! Is there any documentation on this, besides the code itself? I only found - which left me stranded at that question you helped me solve.
***rekado_ is now known as rekado
<marusich>apfel, "built-derivations" is not documented. There are a variety of procedures that unfortunately are not documented. I don't think that is intentional, though. When it comes to figuring out how to do something, it is good to search the Guix source for ideas, or to ask here or on the guix-users/guix-devel email lists for ideas.
<marusich>If you just want to understand what a procedure does and it's not documented, the referring to the Guix source where it is defined is the only way to really undersatnd it.
<marusich>The "built-derivations" is a pretty common idiom for building derivations. It is the monadic analog to the "build-derivations" procedure, which is documented.
<apfel>marusich: ok, thanks
<luis-felipe>Hi, anyone using emacs-doom-modeline knows how to fix some icons not displaying? I just see the unicode rectangle with the corresponding code...
<luis-felipe>(Using Guix System here)
<luis-felipe>(I already tried running "fc-cache -rv" by the way)
<herlocksholmes>Do you have the emacs-all-the-icons package installed? or declared in your system config?
<luis-felipe>herlocksholmes: No, I didn't know about that, thanks.
<ngks>Apologies if people ask this all the time: is there a snippet available that shows how to make xmodmap-like keybinding changes in a Guix System config.scm? It looks like this could go in the set-xorg-configuration form that appears in the config.scm generated by the installer I guess?
<Ikosit>ngks: xmodmap-like?
<leoprikler>luis-felipe: if you don't like extra packages, you can also set up your emacs to use certain fonts for UTF-8 coverage, but that requires some init.el hackery
<ngks>Ikosit: if it uses xmodmap that's fine, if the same bindings apply outside of X that's even better. If possible I would like to have the keybindings managed just by scheme with no extra dotfiles
<Ikosit>I don't think that's possible
<Ikosit>The closest you can get to this is using kmonad
<jonsger>do others see errors like `guix system: error: the following groups appear more than once: lp` when reconfigure on current master?
<ngks>Ikosit: thanks
<luis-felipe>leoprikler: I don't mind installing extra packages, but I'm interested in that option too if you have a link :)
<jonsger>on commit ddb51f3e903ed53ce3c3a88e2af107360f2c51ac its still working, that is 4 days ago
<lfam>That group has to do with printing, right jonsger?
<jonsger>lfam: I suspect the cups changes of nckx :P
<luis-felipe>Damn it! All I see is still tofu...
*luis-felipe goes through all-the-icons troubleshooting steps
*jonsger got remembered that I wanted to create an icon mockup ^^
***amiloradovsky1 is now known as amiloradovsky
<ngks>Ikosit: Guix System actually is able to rebind keys in X using just scheme in config.scm, no packages or dotfile needed. I used the manual section System Configuration > Keyboard Layout. I should have looked there first, it was much easier than I expected...
<jonsger>it's commit a3002104a84c60556b6616d100cb98019e48759d of leoprikler which breaks it
<leoprikler>oops, what did I break?
<leoprikler>Hmm, if that commit breaks something, it means you had a broken configuration before and simply didn't know.
<leoprikler>Should we perhaps demote that to a warning?
<leoprikler>luis-felipe: Not sure where I got it from, but you'd use set-fontset-font, like so:
<luis-felipe>leoprikler: Thanks. I'll try that because I couldn't get all-the-icons to work...
<PotentialUser-52>Why are "Old Persian cuneiform" and "Avestan" characters not supported in GuixSystem?
<lfam>PotentialUser-52: Can you offer any more detail for your question?
<lfam>Like, what makes you thing they are not? In what context are you trying it?
<lfam>What did you try?
<Ikosit>jonsger: I had that exact same issue
<PotentialUser-52>Those characters are not displayed on web pages
<Ikosit>ngks: Do you mean xkb options?
<rekado>PotentialUser-52: did you install a font that has glyphs for these characters?
<PotentialUser-52>rekado, No, fonts must be installed to display characters?
<Ikosit>PotentialUser-52: Yes those characters are most likely not contained in the default fonts
<rekado>PotentialUser-52: yes, that’s the primary purpose of fonts. To provide glyphs for code points.
<rekado>I just noticed that reads like it was written by a Nix evangelist.
<Ikosit>rekado: :D
<rekado>It’s terribly annoying to read.
<lfam>rekado: It's always seemed like the owners of that page dislike Guix
<Ikosit>rekado: What did you earlier this day mean with the manifest, i don't quite get it
<rekado>Ikosit: you wanted to rewrite the package graph for all installed packages, right?
<rekado>Ikosit: a manifest is not merely a list of packages to install; it can also use the full Guix API to rewrite the package graph.
<rekado>Ikosit: so you’d have your list of packages and then apply package transformation functions on them to replace any instance of an original Guix package with your variant.
<Ikosit>sadly manifests are only marginally documented in the manual
<rekado>you’d have to recompile the whole graph up from the point where you replaced inputs, but the result would be the package graph as specified by Guix with the modifications you requested.
<rekado>Ikosit: take a look at 8.3 Defining Package Variants
<rekado>there are a few examples of package input rewriting
<rekado>“libressl-instead-of-openssl” is a simple one that demonstrates graph-wide replacement of one node for another, recursively
<rekado>lfam: yeah, I get the same impression
<Ikosit>Yes, but i can't rewrite packages used by services and packages the system configuration uses like that, can i?
<rekado>lfam: but does every second sentence really need to have the interjection “like Nix”? Even if you dislike Guix: make it at least compelling prose.
<rekado>Ikosit: most services allow for packages to be replaced
<lfam>Some people really hold a grudge against Guix, and always call us out when we don't credit Nix
*Ikosit wants to overwrite xkeyboard-config
<lfam>Who knows what they are feelings
<Ikosit>rekado: I know, the problem is that i don't want to manually rewrite setxkbmap, xkbcomp, etc.
<rekado>Ikosit: it’s not something you can easily do for the whole OS, I think. But each service that uses a package should allow you to replace them via their configuration.
<Ikosit>All those programms, that use xkeyboard-config in some way
<rekado>Ikosit: I see.
<rekado>I’m not familiar with these tools and how they relate to xkeyboard-config; but maybe this doesn’t matter. Perhaps there should be a rewrite function that can be applied to an operating-system record.
<rekado>Perhaps you could bring this up on guix-devel
<rekado>ugh, this article … it’s really poorly written.
<rekado>really bothers me
<Ikosit>rekado: The problem, that one may want to install setxkbmap with guix package
<lfam>rekado: I think it's best to let it go. Other Guix hackers have tried to contribute to the article but it's not possible. The page owners won't allow it
<Ikosit>rekado: I would really like to do this in a clean way, so that it may be possible to implement
<lfam>It's a recipe for frustration
<rekado>Ikosit: it seems that input rewriting would be overkill for this use case.
*rekado tries to forget about the article
<Ikosit>rekado: What would be an alternative?
*rekado looks at the German version
<rekado>Ikosit: a change to an existing system service for keyboard layouts
<rekado>Ikosit: I don’t know what that would look like exactly, but I’m sure we can come up with something after a short discussion on guix-devel
<rekado>oh goodness me…! The German article is just three paragraphs long and mentions Nix three times.
<rekado>what the hell is this?
<lfam>You might have better luck with that article, rekado. I doubt the same people are editing it
<rekado>kudos to the German editors for actually writing “GNU-System”
<lfam>Not a lot of American English speakers are multilingual
<rekado>I’m sure this violates some Wikipedia consensus
<lfam>Ah... consensus
<lfam>Defined as "leadership by the longwinded"
<rekado>ha :)
<Ikosit>rekado: What would this system service do, i'm not experienced enough in that topic (yet)?
<lfam>It looks like anything related to common lisp / sbcl fails to build on berlin for aarch64
<lfam>Is that expected? Should we even be attempting these builds?
<lfam>That is the name of the build farm
<lfam>It's named that way because it's located in Berlin
<rekado>Ikosit: I don’t know what the system service would do, but 10.6 Keyboard Layout describes the keyboard-layout procedure. I suspect that it could be changed to accept an optional list of packages that together should provide the contents at /share/X11/xkb
<rekado>it would default to the xkeyboard-config, but when given another package its /share/X11/xkb directory would be used instead (or together with the original xkeyboard-config’s directory)
<luis-felipe>I finally got the icons working properly in Emacs doom modeline.
<luis-felipe>I thought installing emacs-doom-modeline would grab all the necessary things for it to work, but no.
<rekado>luis-felipe: looks like you need to execute all-the-icons-install-fonts
<luis-felipe>Yeah, that's what I did a few minutes ago.
<rekado>I was very disappointed when I discovered that
<rekado>I would have preferred to have those fonts installed for me when installing the emacs-doom-modeline (or emacs-all-the-icons) package.
<luis-felipe>I just thought installing emacs-all-the-icons would grab the right fonts.
<rekado>perhaps this should be changed
<rekado>it doesn’t feel right to push this extra installation step to users
<luis-felipe>rekado: Yes, exactly. That's what I was expecting.
<luis-felipe>Install emacs-doom-modeline and you're done.
<rekado>luis-felipe: I think it’s fine to treat this as a bug.
<lfam>Thanks for updating the staging branch, efraim
<luis-felipe>rekado: I'll report it then.
<rekado>luis-felipe: excellent, thank you!
<Ikosit>rekado: setxkbmap,etc. has to know the path of xkeyboard-config, so i don't think one can just replace the xkeyboard-config package
<User-73399>I have trouble reconfiguring my system
<User-73399>it complaints about:
<User-73399>guix system: error: the following groups appear more than once: lp
<User-73399>even when i remove both lp and lpadmin from my suplementary groups
<rekado>Ikosit: setxkbmap (et al) could be modified to accept an environment variable specifying the location of layouts.
<Ikosit>User-73399: That's probably a problem with cups
<rekado>it doesn’t really need to be the xkeyboard-config’s output directory
<User-73399>seems similar to jonsger's problem
<Ikosit>I had the same error till i commented the cups configuration out
<User-73399>I do use cups, yes
<Ikosit>Maybe i guix pull will fixed, idk
<User-73399>how will I use my printer without cups?
<rekado>Ikosit: we can make things up as we need them to be. The lack of global state that can be accessed by different applications sometimes requires a few minor changes.
<lfam>User-73399: Can you send a bug report to <>?
<User-73399>ye, let us hope :)
<Ikosit>User-73399: Not :P
<rekado>User-73399: please include your operating system configuration in the bug report, so that we can attempt to reproduce the problem.
<Ikosit>rekado: Ok, i didn't thought about that
<User-73399>I will try after diner
<User-73399>is it just me, or is there no link to the issue tracker from
<rekado>User-73399: it is linked to here:
<rekado>see “Test and Bug Reports”
<User-73399>I ghave also stumbled across a lot of these errors latley:
<User-73399>guix install: error: got unexpected path `Backtrace:' from substituter
<rekado>User-73399: this doesn’t look normal. Could you please also report this with a short recipe to reproduce this?
<User-73399>when installing packages or upgrading
<lfam>Is there more to it User-73399: Is that all it prints?
<User-73399>$ guix install evolution geary
<User-73399>The following packages will be installed:
<User-73399> evolution 3.34.2
<User-73399> geary 3.34.1
<User-73399>guix install: error: got unexpected path `Backtrace:' from substituter
<User-73399>haveing same issues with the integrated pastebin ;(
<lfam>You can use <>
<lfam>Anyways, it's probably a network failure
<lfam>I often get flaky name resolution on Guix System, and it crashes the substituter
<lfam>There's no more output?
<lfam>I'm asking, isn't there any more?
<User-73399>if I retry same times it comes
<User-73399>or works
<rekado>this is not great as it suggests that “guix install” processes the textual output of the substituter, even though it failed
<rekado>so even though it is probably a network failure the appearance of the bug is a bug in itself.
<User-73399>it's a verry resent problem
<jonsger>rekado: I'll revert that commit for the time being, it brakes reconfiguring on my both production systems...
<doctor_rodik>Hi! I have some odd error message while trying to do guix pull on GuixSD: guix pull
<doctor_rodik>Updating channel 'guix' from Git repository at ''...
<doctor_rodik>Building from this channel:
<doctor_rodik> guix 373f808
<doctor_rodik>Computing Guix derivation for 'x86_64-linux'... /
<doctor_rodik>guix pull: error: got unexpected path `Backtrace:' from substituter
<doctor_rodik>Maybe somebody has encountered the same problem?
<Ikosit>I just the same error ("substitute: guix install: error: got unexpected path `Backtrace:' from substituter")
<lfam>doctor_rodik: It was reported by another person only a few minutes ago
<Ikosit>while installing audacity
<Ikosit>Though trying a second time fixed for me
<lfam>Can you send a bug report to <>?
<cbaines>I'm guessing 'Backtrace:' is the first part of a backtrace, but I'm not sure how to get the useful bits of it...
<lfam>As rekado said, "this is not great as it suggests that “guix install” processes the textual output of the substituter, even though it failed"
<lfam>It could be a transient network error
<lfam>But the problem is that it crashes Guix
<lfam>Indeed, it's happening to me as well
<lfam>This is with a Guix and guix-daemon from commit b4384e61165623b16b77b8cab16c81423c6853ed
<lfam>I'll write the bug report
<jonsger>me now as well :(
<lfam>How strange
<lfam>It is bug #45828
<rekado>FWIW I cannot reproduce this with 6be271460b9f8110f94dc7c363205153744ac9c5
*rekado pulls
<lfam>It will happen... a dozen times in a row... and then things will work
<lfam>I marked the bug as "grave"
<cbaines>I can't even get the backtrace with strace, I think the process trying to print the Backtrace is killed before it finishes
<cbaines>it prints "Backtrace:\n", which causes an error elsewhere, which prevents the rest of the backtrace being printed
<rekado>I can reproduce it with “guix pull”
<rekado>cbaines: with “strace -f”?
<rekado>cbaines: trace the daemon instead
<cbaines>I'm tracing from the guix-daemon down
<cbaines>I can see the process starting to print the backtrace, and I can see it getting SIGKILL'd in the process
<cbaines>I tried again, still no luck, someone else might get lucky though
<raghavgururajan>Guix has been throwing "guix build: error: got unexpected path `Backtrace:' from substituter"
<raghavgururajan>Oh already reported?
<lfam>It's bug #45828 raghavgururajan
<rekado>hmm, I cannot reproduce it on commit 6be271460b9f8110f94dc7c363205153744ac9c5
<lfam>Everyone is experiencing it, but we haven't figured out what the problem is yet
<raghavgururajan>lfam: Ah okay.
<lfam>Was rebooted recently rekado?
<rekado>but I do see it on commit b54b3f5419849286611279231d8a2c61d3e935b4
<lfam>I notice that there is a huge amount of unused memory. In the past it is usually full of cached pages
<rekado>almost a month ago
<lfam>I meant in the last day or so
<lfam>It's weird to have unused memory on that system, right?
<cbaines>I wonder if this is representative of what's going wrong
<lfam>I wonder if the bug has been exposed by some mass cache invalidation on that server
<cbaines>I copied the command and the query from the strace output
<rekado>should we restart nginx and/or guix publish?
<lfam>I have no idea
*lfam attempts similar strace to as sbaines
<lfam>cbaines, I mean
<roptat> returns a 500
<rekado>I just restarted guix-publish
<roptat>now it's a 404, which is probably more correct
<cbaines>being triggered by a broken response from makes sense, as that explains why people hit the problem at the same time
<rekado>we also only got this for recent substitutes, not for those that were available a week ago (which is why I didn’t hit this with my older Guix)
<roptat>the Bad Response-Line is probably because we fetch multiple narinfos at once (I got that when trying to fix there's a port that contains a series of HTTP responses; if we get a 500, that's an error and we're supposed to eat the body of the response, which was empty
<cbaines>interestingly, I think the caching masked the issue locally
<lfam>Which caching would that be, cbaines?
<cbaines>narinfo response caching, the stuff in /var/guix/substitutes/cache or ~/.cache/guix/substitutes
<cbaines>I got crashes more consistently when I deleted the cache (this was testing with manually running guix substitute --query though)
<cbaines>anyway, there's probably still some code bug, but it looks to me at least like rekado restarting guix-publish has made guix pull work again
<rekado>cbaines, roptat could you please add your findings to lfam’s bug report?
<lfam>Hm, a version of "1.7.0" is not considered an upgrade from "1.7b1"
<roptat>I think I understand: the 500 error did not have a Content-Length header, so we executed the code that reads until the end of the port; for the following substitute, the port was already at eof, which probably explains the error
<cbaines>I've sent some stuff to the bug report now
<roptat>ah no, that was not it: if the port was read until eof, guix substitute would have complained about #<eof>, not ""
<roptat>unless a network port behaves differently from a file port
<kozo[m]>This is one reason I love using Guix. You guys are very fast and assessing and fixing bugs.
<kozo[m]> * This is one reason I love using Guix. You guys are very fast and assessing at fixing bugs.
<kozo[m]> * This is one reason I love using Guix. You guys are very fast at assessing and fixing bugs.
<lfam>We aim to please! Everyone is welcome to take part
<jonsger>User-73399: lkosit: I'll revert commit a3002104a84c60556b6616d100cb98019e48759d, this fixes the failure of `guix system reconfigure` for me :)
<lfam>Please make sure to notify the commit author :)
<lfam>It's possible that what we had before was also broken but in a silent way