IRC channel logs

2018-09-24.log

back to list of logs

*nckx finishes up a day staring at 'registering closures' just when the fixes land... :-) Thanks!
<civodul>nckx: heh, down from 90mn to 3mn :-)
*civodul -> zZz
<civodul>g'night!
<nckx>Is anyone here https://github.com/guix-mirror?
<EternalZenith>How hard is it to use nixpkgs and their options on GuixSD?
***eric_hill is now known as Guest80326
<civodul>Hello Guix!
<janneke>hello civodul!
<g_bor>hello guix!
<janneke>hi g_bor!
<g_bor>janneke: hello
<g_bor>I'm really glad to see the progress on rust bootstrap.
<g_bor>I remember that on the last r-b summit it was determined to be one of the hardest to bootstrap language.
<janneke>g_bor: yes, very nice
*g_bor wondering if I should try package t-rex
<snape>!!! HN's #1 and #2 are anti-Chrome article :)
<jonsger>I don't see any reason to use Chromium/Chrome instead of Firefox. Even if you "ungoogling" it
<Sleep_Walker>jonsger: some sites quickly stop responding (typically huge pages or JS page with dynamic content), that's why I use Chromium a lot even though I always prefered Firefox/Icecat and Conkeror
<jonsger>oke. Can't really confirm this, maybe because I use uBlock and NoScript to get rid of this bloat...
<rekado>I need to reboot berlin.guixsd.org today. Hardware upgrades.
<rekado>I’ll also use this chance to try relocating /gnu/store.
<buenouanq>is there any roadmap for GuixSD and Hurd?
<civodul>rekado: great, thanks for the heads-up!
<g_bor>rekado: thanks
<civodul>buenouanq: for Hurd specifically the road map is unwritten, but you can get the ideas by browsing the list archives
<g_bor>anyone willing to co-mentor the outreachy project to create video documentation?
<g_bor>mentor link is here: https://www.outreachy.org/communities/cfp/gnu-guix/project/create-user-video-documentation-for-gnu-guix/mentor/submit/
<civodul>g_bor: could you ping the list as well? also say what the hard deadline is :-)
<g_bor>civodul: all right, will do
<g_bor>just for the record, the deadline is Oct. 16, 2018.
*g_bor just tested roptat's first postgresq patch, it is ok
<roptat>g_bor: :)
<roptat>g_bor: there are only two patches in the series, the last I sent is the second one
<roptat>anything in between was sent too early
<roptat>g_bor: also you'll be happy to read http://issues.guix.info/issue/32813 :)
<g_bor>roptat: oh, nice
<g_bor>now running reconfgigure with the second patch applied.
<g_bor>I guess it would be nice to mention in the documentation, that contrib extensions should not be listed.
<roptat>contrib extensions?
<g_bor>yes, like hstore, dblink, intarray...
<g_bor>they work out of the box, as these are developed inside the postgresql tree.
<g_bor>ok, I have no luck so far...
<g_bor>roptat: i've just done a filesystem indexing, and searched for postgis, got nothing
<g_bor>roptat: bingo
<g_bor>it works like a charm
<g_bor>i had a missing use module
<jlicht>'ello guix
<g_bor>hi!
<roptat>g_bor: normally the extensions field lists packages
<roptat>your postgis variable must have been defined elsewhere
<roptat>could you send your config that didn't work?
<g_bor>roptat: yes, I can, but it's just I added geo to use-package-modules. Then it worked.
<civodul>hey jlicht!
<jlicht>o/
<g_bor>roptat: https://paste.debian.net/1043908/ is the working config
<g_bor>not working was the same, except use-package-modules missed databases and geo
<ng0>is it correct that #:select ((package-name->name+version . hyphen-separated-name->name+version)) selects package-name->name+version and uses it as hypen-separated-name->name+version in the module loading it? some parts of the guile documentation are hard to search
<roptat>g_bor: do you mean that it didn't fail with "unbound variable postgis"?
<g_bor>roptat: I will take another look, but it seems that is correct
<jlicht>the (github hosted?) elogind project lists #guix as the place to chat about it. It that still correct?
<g_bor>roptat: checked, it does not fail for me with unbound variable.
<roptat>weird
<roptat>does it have postgis defined from somewhere else?
***kensington_ is now known as kensington
<civodul>jlicht: i suppose we're one of the main users and it's where it originated, thanks to wingo, so... :-)
<wingo>since a long time, elogind has another maintainer
<wingo>who does all of the work!
<wingo>anyway he's very active/responsive on the github so filing issues is a good forum for discussion as well as resolving issues
<jlicht>It seems to be actively developed afaics indeed
<jlicht>is there any presssure for Guix to have a more up-to-date elogind version packaged?
<wingo>i guess we are a year out of date at this point
<civodul>i suppose we should consider upgrading
<jlicht>it seems that elogind moved to the meson build system
<jlicht>also, wingo, you are still listed as the receiver of all blame for elogind in the README ;)
<wingo>probably have to look at https://github.com/elogind/elogind/blob/master/NEWS to see what is likely to break
<wingo>haha sure, i am good at receiving blame without doing anything about it :)
<civodul>:-)
<wingo>ah sorry, that NEWS file is for systemd
<wingo>and very out of date
<g_bor>roptat: I don't think so...
<ng0>ah. i searched in the wrong place. I was right.
<wingo>(or is it? looks like NEWS is just systemd's NEWS)
<g_bor>at least guix package -s does not return anything else...
<lfam>Good morning Guix!
<g_bor>roptat: oops, no, it's not that...
<wingo>hello lfam :)
<g_bor>it actually return an error.
<g_bor>hello lfam
<civodul>moin lfam!
<jlicht>wingo: I think they actually track (part of?) systemd upstream, so that would make sense
<g_bor>at least return value is 1, but gives no output...
<roptat>g_bor: I see that
<roptat>after a second run I get unbound variable
<roptat>weird, I get unbound variable now
<lfam>I'm on Debian, with both root and my user at commit 88600acc939365b4f. The daemon is 0.15.0-3.3d43017 (from root's Guix package). When I try to update with `guix pull` to any later commit, it crashes like this: https://paste.debian.net/1043915/
<lfam>The change in the next commit after 88600acc939365b4f is c122a2e509df98c4 (serialization: Remove redundancy in 'write-file'.), which looks like a really simple change to me
<civodul>lfam: same here
<civodul>it built fine on berlin though
<civodul>hmm /gnu/store/njdhsdv216hyk23nfrc7wh9hbq17ymxk-compute-guix-derivation has no references
<g_bor>civodul: we observed something strange with roptat.
<g_bor>I've create a config, which misses a use-package-module, but on the first system reconfigure I get no output, and exit code 1.
<g_bor>are error messages hidden?
<civodul>g_bor: no idea; could you send the details to bug-guix?
<g_bor>ok, will do
<g_bor>also, I will try to reproduce this on current master
<civodul>thanks!
<civodul>lfam: indeed c122a2e509df98c4391750a066fecba37465eab3 seems to be the culprit
*rekado stopped cuirass and guix-daemon on berlin for a final sync of /gnu to the external storage before reboot
<dustyweb>hello!
<dustyweb>I'm running the new firefox, horray
<wingo>wooo!!
*wingo should get that going
<roptat>dustyweb: firefox, not icecat?
<dustyweb>sorry, icecat :P
<dustyweb>also... and yeah I know I shouldn't be running proprietary javascript in theory, but I have to for my job. something that's happening for me though:
<dustyweb> https://github.com/ looks completely unstyled
<roptat>same here :/
<dustyweb>like, no css whatsoever level unstyled
<jlicht>dustyweb: how did you fix that?
<dustyweb>jlicht: fix what?
<jlicht>I also have that issue with several websites (also the AMO website for example)
<dustyweb>roptat: oh, interesting
<jlicht>the no-css-at-all thing
<dustyweb>ah I haven't fixed it jlicht
<dustyweb>I did disable all extensions except for ublock origin and tree style tab
<dustyweb>still gettin git
<dustyweb>even with ublock origin disabled, i'm getting the problem
<jlicht>same story in private browsing mode
<dustyweb>jlicht: ok, if 3 of us are getting tthis
<dustyweb>then maybe that warrants filing a bug
<dustyweb>jlicht: roptat: got other website examples you can link to?
<jlicht>dustyweb: https://addons.mozilla.org/en-US/firefox/ :-)
<civodul>lfam: looks like the reference scanner wrecked havoc or something, really weird
<civodul>if i rebuild compute-guix-derivation.drv locally, it gets a proper list of referenecs
<civodul>*references
<lfam>Hm
<civodul>for now i've rebuilt it on berlin as well and got the narinfo regenerated
<civodul>unsatisfactory, but it solves the immediate issue
<lfam>civodul: How do you figure out which drv it is?
<civodul>"guix gc --derivers ...-compute-guix-derivation"
<lfam>Ah, I've never used that before :)
<civodul>so i don't know if one of the build machines is in a bad state
<civodul>super weird!
<lfam>That's really weird indeed
<civodul>the reference scanner hasn't changed since Guix exists, so it'd be really surprising to find a bug there
<lfam>I guess the database write must have failed?
<civodul>but then the whole transaction would fail and the item would not be in the ValidPaths table, AIUI
<davexunit>'guix pull' isn't working for me because it insists that the savannah ssl cert is invalid. setting GIT_SSL_NO_VERIFY doesn't unblock me. :(
<davexunit>presumably since it's using libgit2 which doesn't honor that env var
<wingo>maybe set your clock to the past :)
<dustyweb>jlicht: ah right! I did see that too
<wingo>relive the best parts of the worst timeline
<davexunit>:(
<lfam>davexunit: Are you sure the TLS-related environment variables are set properly? <https://www.gnu.org/software/guix/manual/en/html_node/Application-Setup.html#X_002e509-Certificates-1>
<davexunit>lfam: no I am not
<davexunit>I'm using ubuntu where everything in /etc/ssl is usually just fine. first time I've seen such an issue when using 'guix pull'
<davexunit>let's try using what guix has...
<lfam>If I understand correctly, OpenSSL is used here. So you could try with Guix's OpenSSL: `echo | openssl s_client -connect git.savannah.gnu.org:443 | openssl x509 -noout`
<davexunit>somehow installing nss-certs pulls in ghostscript, pango, python, lvvm, etc.
<lfam>Hm... that's annoying
*civodul discovers unsent messages in the draft folder that really should have been sent
<civodul>davexunit: did you try: export SSL_CERT_DIR=/etc/ssl/certs ?
<lfam>Unfortunately nss-certs depends on openssl, perl, and python 2, so the dependency graph is pretty large, although they are only needed when building from source
<davexunit>civodul: that worked! I tried it earlier but used SSL_CERT_DIR=/etc/ssl :(
<davexunit>so thanks!
<civodul>yw!
<civodul>FWIW this bug has been fixed: now 'guix pull' uses /etc/ssl/certs if it exists when SSL_CERT_DIR is unset
<davexunit>oh cool. my current guix is a couple of weeks old
<davexunit>whoa... 'guix pull' just used substitutes for everything... I've never seen that happen.
<lfam>:)
<civodul>heheh
<civodul>it's not so rare actually :-)
<ng0>does anyone know how third_party/rust/miniz_oxide_c_api/tests/miniz_tester.cpp is used in the build of icecat? If we do not patch out __DATE__ __TIME__ in the else block ( printf("miniz.c x86 Command Line Test App - Compiled %s %s\n", __DATE__, __TIME__); ) this means we are still getting our chroot date+time right.
<ng0>(I meant to end this with a questionmark)
<ng0>my guess right now is that …%s\n", __EPOCH_TIME__); or whatever we called it in reproducible-builds would be better
<ng0>unless we do something to DATE+TIME in the chroot I have forgotten right now?
<jlicht>Seeing as guix already includes openrct2, would a openrct2 service also be accepted if I write one?
<jlicht>*an
<lfam>jlicht: Absolutely! What would the service do beyond running the game?
<jlicht>lfam: not much, just allow configuration + package management to be done with Lispy-goodness :-)
<rekado>berlin.guixsd.org is down now; waiting for hardware updates to be applied
<rekado>is there a way to express a dependency of one channel on another channel?
<nckx>o/ Guix!
<nckx>rekado: Is that downtime expected to be measured in hours or days?
<rekado>hours
<rekado>I’m having some problems with the file systems declaration, unfortunately.
<rekado>the /gnu/store still gets mounted from /dev/sda1 (where the root file system is), not from the drive that I asked it to mount to /gnu.
<`ani`>hello.
<`ani`>is this the support channel for GuixSD?
<nckx>`ani`: The one & only.
<`ani`>I would like to know what to do if a free software package is missing.
<nckx>Ideally: package it. Which package?
<`ani`>nckx: freedroidrpg and openarena are two examples.
<pkill9>rekado: are you trying to put /gnu/store on a different partition?
<ng0>openarena would be tough for the beginning.
<rekado>pkill9: a different disk.
*nckx takes a look at Freedroid.
<pkill9>rekado: do you have a oslution for starting the daemon *after* mounting the disk to /gnu/store?
<pkill9>solution*
<rekado>pkill9: needed-for-boot? #t
<pkill9>ah nice, so you put that in the filesystems declaration?
<rekado>yes.
<rekado>unfortunately, whatever I did here didn’t quite work.
<rekado>(rebooting servers takes forever…)
<rekado>hmm, only the local disk is visible in GRUB.
<rekado>that means I can’t mount /gnu from the external disk and boot from there.
<rekado>bleh.
*nckx packages it.
<ng0>why would an git-download throw errors? Initialized empty Git repository in /gnu/store/pjlph3zi871fsgngp7grjnccasr54pj3-emacs-nim-mode-0.4.2-2.d56b77a-checkout/.git/ .... fatal: cannot run ssh: No such file or directory .... fatal: unable to fork .... fatal: cannot run ssh: No such file or directory .... fatal: unable to fork ..... (.... is new line) looking familiar to anyone?
<ng0>i have run git source builds in the last 60 minutes which did not produce this
<jlicht>ng0: maybe something interesting shows up in the `dmesg` output?
*rekado gives up and boots an older system on berlin.guixsd.org
<jlicht>maybe you hit your process limit?
<ng0>hm
<ng0>i'm building chromium offloaded, and this is local
<`ani`>uh.
<`ani`>did anybody answer me? I can't see in the scrollback.
<ng0>jlicht: i'll look at the log as soon as this other process is done
<nckx>`ani`: I'm already playing freedroidrpg on Guix. :-)
<nckx>It wasn't too hard to package, but there's a test failure that I haven't looked at yet.
<`ani`>ok.
<`ani`>I'm migrating from Parabola.
<`ani`>I see you already have tintin++.
<nckx>Are you running Guix yet? This would be a reasonably easy first package apart from an include issue with our split SDL packages.
<nckx>If you're not interested in giving it a try yourself, I'll submit it myself after fixing the tests.
<nckx>s/ [^ ]*self//g
<`ani`>nckx: do you mean I could first try it by installing the package in Parabola?
<nckx>`ani`: No, try writing your (I presumptively presume) first Guix package...
<`ani`>aha.
<`ani`>not interested at the moment to package for Guix.
<nckx>OK.
<ng0>does anyone know where melpa keeps the version numbers it builds from? I need to map what looks like a build date on melpa + revision to a git commit. by date and numbers it is not occuring in the git
<ng0>oops
<ng0>got it :)
<`ani`>there has to be a better way to burn an image to a USB stick than disk destroyer.
<ng0>choose whatever works for you
<jlicht>`ani`: I sometimes use https://etcher.io/, but I don't think it's packaged for Guix yet
<`ani`>I saw that software in PureOS.
<`ani`>(the webpage).
<`ani`>I would want a better alternative to choqok as well.
<`ani`>it has no mastodon support and no images are loaded.
<`ani`>all in all it's pretty lousy.
<ng0>jlicht: nothing in dmesg
<jlicht>ng0: did you hit some weird user process count limit?
<ng0>idk
<jlicht>and of course, did you try restarting it?
<jlicht>;)
<ng0>i was just fixing up my packages, and would have to take a break now.. exam tomorrow, so I guess I'll do this tomorrow night
<jlicht>good luck!
<ng0>thanks
<nckx>ng0: +1 on the good luck.
***paul__ is now known as kristianpaul
<`ani`>does PureOS have an IRC channel here?
<nlyy>hi
<nlyy>why do i get this? fatal error: linux/limits.h: No such file or directory
<bavier`>nlyy: depends on what you're trying to do
<nlyy>I am following the guile reference manual 2.3, i get this on the gcc part
<nlyy>I am assuming its to do with how I have installed packages
<roptat>I think we still need more context to understand. If you are trying to follow a guide, can you give us the link?
<roptat>"the reference manual" is a bit vague ;)
<nlyy>C-h i m Guile Ref <RET> n n ] ] ]
<nlyy>lol
<nlyy>"Linking Guile into Programs" gcc compilation part
<nlyy>*C-h i m Guile Ref <RET> n n n ] ] ]
<roptat>ah, then do you actually have a linux/limits.h file on your system? (are you on GuixSD?)
<nlyy>I am sure everyone uses emacs
<roptat>I don't :p
<nlyy>nah, I am on nixos
<nlyy>I am dying to switch to GuixSd anytime
<nlyy>:)
<roptat>mh.. you may have to install linux-headers or equivalent in your profile
<roptat>then you may need to set C_INCLUDE_PATH or something
<roptat>I can't remember the correct one
<nlyy>I see package linux-libre-headers in guix
<roptat>that's the one that provides linux/limits.h
*civodul reads about what looks like a full Rust bootstrap, woooohoo!
<nlyy>guix provides useful info about the paths :) yay! Trying...
<civodul>nlyy: perhaps you need to install 'gcc-toolchain', not just 'gcc': https://www.gnu.org/software/guix/manual/en/html_node/Application-Setup.html#The-GCC-toolchain
<nlyy>ok, ill try civodul's method first
<nlyy>it (gcc-toolchain) worked! Thanks :)
<civodul>cool, yw!
<ng0>nckx: the good luck i actually more of a "hope my prof in C is a decent human being and I don't have to sue him for past business relationship (which, due to their company basically being an we '"like" OSS' sweatshop, didn't end like they wanted it to) reflecting in any regard on the result of the exam"
<ng0>so it will be luck :) because often people are not capable to be neutral and separate workplace stuff.
<nckx>...wow. Now that's a complicating factor. o_o
<ng0>ye :D
<ng0>i mean, I acted and closed down the contract in a civil and polite way, it was just them acting not so .. let's say "good".
*nckx adds that to the 'maybe not' column of the 'should I consider looking for a free software dev job' table.
<nckx>So many horror stories.
<divansantana>is qutebrowser package currently broken? I get QtWebEngine is not installed msg.
<ng0>it depends. i had to take this because i needed the money
<ng0>due to my students loan and blabla legalities I have to keep my rates far below what my work is worth, but my current contract is really nice :)
<bavier`>divansantana: yes, it's currently broken for some reason
<divansantana>bavier`:ok thanks.
<ng0>ah, when i wrote "OSS sweatshop": to call it open source was borderline. technically what I did was closed source + open source in the liberal sense.
<divansantana>I find the new icecat can't render github.com for me properly. I've even disabled all the extensions. Is this the case for others too? Surely must be. I've started a fresh icecat instance and this occurs as well.
<divansantana>Finding a decent browser is really hard nowadays. :( Looking forward to next browser being packaged.
<divansantana>Used to use vimperator+icecat which worked great.
<bavier`>divansantana: fixing qutebrowser is on my todo list; but I of course would not object to someone else fixing it in the meantime
<lfam_>divansantana: Could it be <https://bugs.gnu.org/32820>?
<ng0>do we have a bug for Palemoon/Newmoon/New Moon? I maintain it outside of guix and will finish the package, but it is not worth getting properly supported as the browser is now in maintainenance mode, soon to be obsoleted by a new engine+browser
<ng0>I don't really trust in the work the team does around this browser and I have doubts about communicating with them after what I've experienced with openbsd.
<ng0>my one email with upstream was okay'ish because i asked the right thing only
<divansantana>bavier`: thanks for the effort. It will be a great addition to get that in guix.
<ng0>could someone close the corresponding ticket if it exists? I'm busy with pen and paper right now
<ng0>unless you really want New Moon at some point, but for me it's just something I want to finish because I started it.
<divansantana>lfam_: That looks exactly it. I need to get debbugs going in emacs and start checking the bugs.
<lfam_>divansantana: In the meantime I can recommend the Epiphany browser
<divansantana>lfam_: Cool. I'll Use that but will miss all the vimperator like look+bindings. Will see if there is anything similar for it, but suspect not.
<divansantana>Maybe chromium with an extension. Not ideal. But at least the guix chromium is well packaged in terms of privacy/security.
<lfam_>Yeah, that's another good option. But it may take a while to build (unless there is a substitute server) while Epiphany should be available now
<pkill9>i currently use nix's chromium with a wrapper, until it's in guix
<nckx>ng0: 'free-as-in-gimme-free-stuff'. Got it. ;-)
<nckx>lfam_: Wait. Are you implying I... over-think things? I think so! ...on the other hand, maybe you're not? ... Although maybe...
<lfam_>Lol
<rekado>re browsers: Eolie and Epiphany are decent browsers.
<nckx>re browsers: Should I refrain from updating Icecat to 60 on GuixSD for now?
*rekado looks forward to merging the wip-gnome-updates branch
<divansantana>Turns out what I'm after is a good privacy/security browser+keyboard driven+customisable. Seems icecat with an extension may be best once that bug is fixed. Qutebrowser is good for me too. Will maybe run in via my arch vm. Later next browser perhaps.
<divansantana>rekado: thanks.
<rekado>Eolie supports Firefox plugins (AFAIU); it also has easy controls for disabling JavaScript and others.
<divansantana>nckx: I think so, but it's now "insecure". It's quite a bit of work to upgrade if you used to customise the old one a lot and those extensions are broken now.
<rekado>it feels like a more flexible Epiphany.
<divansantana>rekado: Hmm. Cool, I'll look into it and report back sometime. Thanks.
<nckx>divansantana: As in: you think I should wait?
<nckx>I've used (and like) new FF, extensions aren't the problem, but I've heard rumblings about missing page elements etc.
<nckx>Bah. I'll spend the 24h to build it. I can always roll back.
<divansantana>nckx: Then just that css bug may be a pain/blocker for you. You could always upgrade and then roll back if it's an issue for you. It is for me.
<divansantana>lol
<divansantana>cool
<nckx>AMD APU FTW
*divansantana looks forward to building a guix build farm at work.
<rekado>I just made some progress on the CSS bug.
<rekado>editing the CSS with the devtools makes Icecat render it just fine.