IRC channel logs
2023-07-26.log
back to list of logs
<two[m]>i'm still one the old matrix channel and i see you as katco[m] <zamfofex>I’m on IRC and I see two[m]’s message to no‐one else. 😅 <zamfofex>Why was Matrix created? Why does it continue to exist? Human kind is still trying to find an answer. <bjc>because you can't get vc funding when you say you're going to build a jabber server <haugh><zamfofex> "Why was Matrix created? Why does..." <- Matrix is a federated realtime multimedia protocol created to provide general chat integration across software and infrastructure. <haugh>So far the worst thing about it is all the whining. <andreas-e>I think when restarting builds, cuirass does not sort them; so if it restarts Y before X when Y depends on X, then Y will fail because it depends on a cancelled (thus failed) X. <andreas-e>And as you guessed, deleting a branch or even a cuirass specification does not prevent the scheduled builds from going forward. You would need to cancel the builds first. <jpoiret>yay, GRUB 2.12 will soon be released, with improved LUKS2 support! <jpoiret>I think I looked at it but a second glance won't hurt anyone <cbaines>I still don't really understand where the problems are coming from, as if just introducing memoization broke things, it suggests that (%current-system) isn't the right value in some places? <andreas-e>cbaines: Can the files /tmp/nar-herder-source-nar-xxxxxx on bayfront be deleted? They end up taking 11GB. And make it more difficult to see what is being built using "ls". <cbaines>andreas-e, yep, I've noticed that too and I'm trying to get the nar-herder to clean up it's own mess <cbaines>jpoiret, also, I don't understand how packages which use (%final-inputs) are still working <jpoiret>I'd say they could be broken if you're using an explicit --system <cbaines>jpoiret, ah, I think I'm missing that both the base and commencement modules export %final-inputs but one is an alist and the other a memoized single argument procedure <andreas-e>cbaines: Thanks, I have deleted them for now. <jpoiret>ah yes, forgetting to cc someone and then wondering why they haven't answered <jpoiret>cbaines: wdym QA's data is confused? doesn't that rather indicate a mass-rebuild? <cbaines>jpoiret, the base revision for assessing that patch happened to be one which is missing most i686-linux and armhf-linux package derivations <cbaines>just because of the interaction of this bug and the random order in which the data service computed the derivations <jpoiret>what about the powerpc/x86_64/pc-gnu changes? <jpoiret>i think i586-pc-gnu on CI was canceled a long time ago when it couldn't build anymore <janneke>maybe "someone" could un-cancel it? ;-) <jpoiret>yeah, although I just noticed that CI won't work with the system-dependent %final-inputs <janneke>ACTION has updated build-aux/cuirass/hurd-manifest.scm yesterday with new packages/descriptions that should build now <janneke>ACTION built "hello" and then "hurd" in a childhurd yesterday and is now building "guix" <janneke>and would have hoped to get some backup/help from ci ;-) <jpoiret>i forgot to cc ludo on the gc-roots test change, so of course he hasn't replied :( <damo22>i dont know how we will get smp working properly on hurd, it boots but the locking in the kernel is borked <jpoiret>damo22: seems like #hurd spilled over in #guix :) <damo22>well anyway, i really would like smp so i can build packages with more than one core <janneke>sure, #:guix could use some of that too <damo22>is it a qemu image running inside guix with hurd on it? <jpoiret>it sets up everything almost automatically so that you can start native hurd builds from the host <damo22>what makes it unique from say, running debian hurd in qemu <jpoiret>that way you don't really need to interact with it, it's just there but you can do everything on the host <jpoiret>(by everything I mean developing and building hurd packages) <damo22>so if your laptop has 8 cores you can build hurd packages with -j8 <jpoiret>ah no, the native hurd builds do happen in the childhurd so it's still -j1 <jpoiret>although I guess you can start one vm per core :) <damo22>yeah i develop hurd on debian hurd in a vm <jpoiret>it won't parallelize a single package build but if you have a bunch of them to build that's going to be faster <damo22>but again i am stuck with -j1 for now <damo22>it is possible to build a smp enabled gnumach but its terribly slow <janneke>damo22: oh, that's less than great... <damo22>#hurd is very quiet, and the maintainer is away currently <damo22>i re-enabled some code paths that allow spinlocks to be timed and basic stats printed out for the most contended locks, and i figured out which one is taking most of the time, but it turns out its the vm_page_queue_lock and i dont know how to optimise the memory management code, thats a bit over my head <jpoiret>ah yes, the classic "building kernels is actually hard and could be a full-time job" <jpoiret>esp. since SMP is probably very different from where it was back when it first got implemented in kernels <jpoiret>the implementations had time to mature along with hardware <jpoiret>now we have totally unreasonable expectations <damo22>although, mach used to function with multicore back in the VAX machine <damo22>its just reenabling code paths that have bitrotted for 20 or so years <cnx>hi, i need help with packaging ibus-table, none of the tables in ibus-table-others show up <cnx>so far i've overridden ibus with IBUS_TABLE_* search paths; on nixpkgs they seem to also patch ibus-table replacing them with paths to its $out, what's the guix way? <next4th>cnx: hello, i'm iyzsong's another nick :) overriden mean patch ibus, and ibus-tables is another package? <cnx>yes, ibus-table is an engine, and table-others is a collection of data for some `layouts' <cnx>how do i test an ibus engine independent of ibus? <next4th>um, maybe it doesn't have builtin one.. and search tables in share/ibus-table/tables, which you had patch to IBUS_TABLE_* search path? <cnx>what's the guix way for $out? <next4th>it's #$output or (assoc-ref outputs "out") <next4th>yw, ibus-table hardcode '/usr/share/ibus-table', better to modify it to use a search-path variable (ideally support multiple directories like PATH, XDG_DATA_DIRS), and contrib to upstream <cnx>tks, i'll see what can be done and report back later <razlix77[m]>hello does anyone have insights on how to use "maybe-xxx" options inside "define-configuration" ? I've been looking for a few days at various service definition in the guix tree but all I manage is to get wrong value to apply <razlix77[m]> * (define-configuration my-config (name maybe-string "some optional input")) <next4th>razlix77[m]: um, in (gnu services networking.scm), it has: (define-maybe/no-serialization string), maybe you need that too <razlix77[m]>If I read the documentation properly if no value is provided no serialization will be performed which works <razlix77[m]>trouble is if I do provide a string value it complains with "Wrong type to apply" <mirai>(define (serialize-interface-custom-name field value) (if (maybe-value-set? value) #~(string-append " " #$value) "")) <mirai>#~(string-append " " #$(maybe-value value "") "")) <razlix77[m]>```scheme (define myconf (firehol-configuration (version "6")(interfaces (list(firehol-interface (name "rth0")(custom-name "dssd")(ip "1234")))))) <razlix77[m]>```ice-9/eval.scm:173:55: Wrong type to apply: "1234" <next4th>i think mirai means in that procedure, the value is a <maybe> type, not a string, so you got wrong value to apply, you can fix that by unbox the maybe value. <razlix77[m]>umm uh hmm ... but then the documentation is a mess 😅 🥳 <next4th>um, i'm wrong, the value is a just symbol '%unset-marker% or a plain value. not sure where the wrong apply came from... <razlix77[m]>But by reading other services it appears "maybe-xxx" is used in quite a few different ways... My head hurts 🤕 <next4th>razlix77[m]: debug a bit, i think what's wrong is in serialize-ip-add, it should be #~(if (...)) not #~((if ...)) <next4th>i got a similiar failed log, then look at its backtrace for primitive-load "/gnu/store/xxx-builder", and in that builder it's a bit easier to look syntax errors... <next4th>guix build -f xxx.scm, then zless its log <pjals_fox>"but then the documentation is a mess" sounds like guix :p <razlix77[m]>pjals_fox: nah it's all me getting tangled in parentheses <next4th>razlix77[m]: your paste had "View build log at '/var/log/guix/drvs/p9/2jw5rj99v4k7v4p37s9q7r4bsswh7q-firehol.conf.drv.gz'" too <next4th>yeah, next is in (primitive-load "/gnu/store/9qrz7bik627n2zk...?") <next4th>cat that builder file or open it in editor, you'll find (("1234")) is there <next4th>((if (equal? "ip" "ip") (string-append "1234")), but yeah.. <razlix77[m]>ok thanks for the tips I know I should emacs with all the bells and whistles but I don't know emacs 🤫 more of a helix-editor person 🙈 <migap_>Hello, I am a new guix user unable to successfully running: sudo <migap_>guix system reconfigure /etc/config.scm. I have a message telling <migap_>me to restart the servie NetworkManager, which after doing-it I <migap_>have a similar message telling me to restart the networking <migap_>service which herd is unable to do. Any tip will be appreciated, <mirai>razlix77[m]: yeah, the doc for that isn't the best <migap_>I have a lot of "Soliciting pool server" from ntpd, <migap_>and "successfully established" from avahi-daemon. There is <mirai>seeg45: if its your first mail to the gnu mailing lists you might need to wait a few hours for your address to be (manually) approved <seeg45>just asking because I didn't see a signup page <seeg45>btw, what is the guix way to perform a regexp replace on a file? <seeg45>i guess i could `(invoke "sed" ...)` <apteryx>hm, at least this was the original idea, and I think it was commited to (gnu build activation) in the end to avoid a world rebuild <andreas-e>seeg45: Also issues.guix.gnu.org has a bit of delay compared to debbugs itself. <mirai>hmm, the original idea was to subsume it into (guix build utils) <podiki[m]>is CI stuck? not seeing evaluations or builds again <mirai>apteryx: could you ping that issue to see what else is left? <apteryx>I think the Guix-side is done (it could be moved to (guix build utils) on core-updates)) <apteryx>some things on Guile side may still be pending, I don't know <mirai>not really any different for whether the procedure lives under (gnu build …) or (guix build utils) <andreas-e>podiki[m]: I think cbaines's conclusion was that it was stuck again, yes. <andreas-e>podiki[m]: Did you see my remarks above on cancelling builds? <podiki[m]>andreas-e: sorry did not, just got back to computer now and didn't scroll too far back <apteryx>mirai: yeah. especially changing permissions is useless for building packages <apteryx>the issue is old and forgotten, I think it can be closed <mirai>idk, (guix build utils) seems to contain things that are used here and there… <mirai>but I think your rationale for the build side of things is sound <apteryx>it'd be logical to group mkdir-p and mkdir-p/perms though, from a discovery standpoint <apteryx>it's the only non-activation specific API exported by the activation module <mirai>mayhaps grouping it within (gnu services configuration) would make more sense? <podiki[m]>andreas-e: ah I see in the log. (as perhaps expected, the plumbing to matrix room was not working for a while and i didn't have those messages locally) <apteryx>mirai: I prefer (guix build utils). Also moving something *out* of (guix build utils) is a major undertaking with little returns <mirai>I'm not too particular about the location for tho :) <apteryx>major because it's probably used a hundreds of places <mirai>oh I was talking about the /perm variant <mirai>apteryx: you could extract the supplementary groups from the user record <andreas-e>podiki[m]: I considered talking to sneek, but since you were marked as present, I think it would not have worked. <mirai>that is supplied in the 'user' field <mirai>I read your issue for requesting posix CAP support in shepherd <mirai>so that one will have to happen first <apteryx>then perhaps the default group for mpd-user should be "audio" instead of "mpd" in the meantime <apteryx>lilyp: could you remind me of what happens when a user specifies two objects for the same group as part of their OS configuration? <dlowe>python-django package isn't passing tests :( <apteryx>janneke: the oldest commit I've managed to travel with 'guix time-machine' to so far is 49bb544b77c7dcd7d2a1bade00fb0844a2e8f7eefrom Fri Sep 3 12:54:55 2021 -0400 <apteryx>49bb544b77c7dcd7d2a1bade00fb0844a2e8f7ee from* <nckx>sneek: later tell seeg45 Mail approved! <apteryx>most older commits in that era hangs computing derivations... not sure if I can make guix time-machine verbose to see if it's actually doing something or truly hanging <apteryx>looks like idling based on CPU/memory load <janneke>have you tried using a stepping stone commit? <janneke>yes, travel back from a commit a while back <apteryx>I initially tried going to around the end of 2018 <apteryx>but it failed similarly I think. I'll try a couple more commits <cbaines>apteryx, data.guix.gnu.org has 897f303d2fa61497a931cf5fcb43349eb5f44c14 (from the start of 2019) so I'd expect you'd be able to travel at least that far back <podiki[m]>for CI, are we just waiting for it to do something or does it need some action? <apteryx>probably requires to be restarted, though I think cbaines had done so a couple days back <cbaines>yeah, I think it's getting stuck, potentially on some networking stuff that doesn't timeout <cbaines>we could restart it again, but really it requires some work to get Cuirass to timeout and better report errors <cbaines>yeah, and rekado has reported Git issues at the MDC as well I think. Though it's the bits where it's not logging that it's getting stuck I think <nckx>Just to point out that restarting isn't the answer: I've done so 3 times during the past week. <cbaines>one practical thing we could do is remove/disable some of the specifications <cbaines>they're processed one after the other, so removing specifications will reduce the changes of it getting stuck <apteryx>so is the suspicion that some git repo cuirass needs to fetch as part of processing the job specifications are sometimes failing and this causes it to fail? <apteryx>it's weird because the main loop seems to be going still <cbaines>that's not particularly wierd, they used to be separate threads I think, and now they're separate fibers <apteryx>it'd be cool to be able to attach a debugger and see what it was doing, a la GDB <cbaines>if you look through the cuirass.log file, you can normally see where it outputs "fetching channels for spec" but then nothing after <cbaines>apteryx, you can do that, although it would probably be easier if you have the non-blocking procedure set a identifiable thread name on the thread it creates <cbaines>then you'd be able to pick it out in gdb and look at the stack <apteryx>but GDB not supporting Guile specially (or does it?), introspection would be very difficult, right? <podiki[m]>trying to get something with expired certificate? or down? <Guest18>Does an Intel Arc GPU have better open driver support than AMD or Nvidia? <cbaines>apteryx, I really don't know too much about gdb, but I have found it useful <apteryx>Guest18: Intel Arc also requires nonfree firmware blobs... <apteryx>I'd be curious to see someone using it in a live debugging session some day :-) <Guest18>apteryx: Okay, so there is no current gen GPU without blobs? <cbaines>often GDB is useful when you're not actually debugging guile code, but guile interacting something like hardware or calling out in to native code <apteryx>Guest18: I think there aren't at least from the maintsream vendors (nvidia, intel, amd) <apteryx>I'm not sure about intel integrated GPU <podiki[m]>(ignore me, i'm missing most messages; thanks matrix) <apteryx>podiki[m]: have you considered upgrading to 'podiki'? ;-) <cbaines>if anyone is up for making some Cuirass code changes, you could probably make it more reliable by having a fiber per specification, so pulling the for-each out of proecess-specs <cbaines>that way one specification getting stuck wouldn't block all the others <apteryx>cbaines: that sounds like a good suggestion <podiki[m]>apteryx: yeah at this point either i run my own bridge locally (to help debug/see what i can do) or just use irc directly which is probably the better idea <Guest18>Someone knows what would be the latest gen without firmware blobs before I dig into this myself? <apteryx>Guest18: that 2020 nvidia you found on h-node.org was surprisingly recent to me. If it really works that's probably among the better options. <Guest18>apteryx: Does not support HW accel for codecs which I need to watch media. I did some digging and all new hardware from the big manufactures require a firmware nowadays. Best bet is apparently a Nvidia 770 TI. Does not require any special firmware and has the most compute power. compared to other options. This is a 10 year old card. Oh <apteryx>you won't find anything supporting video acceleration/deceleration that doesn't require proprietary nonfree frimware, I'm afraid. <apteryx>Guest18: so does that mean the report about that nvidia from 2020 used on parabola is wrong? It'd be indeed surprising that a 2020 model does not require proprietary firmware <apteryx>I thought the critical year was more around 2013 <Guest18>EXTFW means it requires firmware blobs. Therefore wouldn't work with Linux Libre. Is this correct? <nckx>That is how I interpret that page. <devmsv_web>hello everyone, has anyone manage to mount cifs with sec=krb5? (or NFSv4) <devmsv_web>I have done some tests and I have been unable to make it work, not sure if it is because something is missing in Guix or from my end <bjc>my nfs mounts are all v4.2, but i don't use kerberos <RavenJoad>I am trying to diagnose a misbehaving application. From what I am reading online, the program may require an xdg-desktop-portal. Does Guix have one? <d00s>guix search xdg-portal should bring up a few. I'm using one as we speak <lilyp>apteryx: when it's eq? objects, they are be merged, otherwise a warning (which should really be a hard error) is issued <Guest28>acceleration. Therefore if I understand it correctly, the mentioned page should be correct <apteryx>Guest28: where do you see on that page that it requires firmware (for nv160) <apteryx>nv40 looks like a fine choice from that table <apteryx>it has video decoding acceleration without blobs <apteryx>but maybe the codecs it supported are near-useless because it's so old ^^' <WorldsEndless[m]>I've tried hunting things down, which eventually led me to mktexslr, but that also fails because it can't edit a dir. I'm wondering if I did something SUDO when I shouldn't have, or did the opposite by leaving off sudo when I should have used it <WorldsEndless[m]>After doing guix pull and guix fetch -u My libreoffice won't start. I get that error <apteryx>lilyp: OK. Does guile support equal? on record objects? <WorldsEndless[m]>sorry, yeah. guix package -u this was before a long break, so I had to review my bash history <apteryx>WorldsEndless[m]: I'm guessing you have a mismatching glibc-locale requirements from the system vs your user profile <apteryx>time to upgrade your Guix System as well and reboot <WorldsEndless[m]>WorldsEndless[m]: I've tried removing and reinstalling it, and making sure glibc was updated, but nothing works <apteryx>VP2 seems a bit more useful (has H.264) <WorldsEndless[m]>apteryx: is that `guix system reconfigure ~/.config/guix/system.scm` etc? <apteryx>VPE1 or VPE2 doesn't, but they don't support anything useful <apteryx>if that's where you keep your system config <lilyp>apteryx: I think there's a chance that using equal? will get you bogus results, what with source locations being embedded etc. <lilyp>also I think it'd be a weird api with more potential for surprise <apteryx>we'd need something tailored to compare for equality <apteryx>that only compares the normative fields <apteryx>could be added to (guix records) if it was typed <lilyp>I think this should be simple enough to code up as part of the current syntax, but it'd require one more parameter :) <WorldsEndless[m]>in the off chance that someone has a quick answer, has anyone had success getting xscreensaver to lock screen on guix? <lilyp>perhaps match-record can also be used <apteryx>ACTION has spent 1 hour trying to figure out how to update expiry on their PGP key and subkeys at once <andreas-e>WorldsEndless[m]: In the past, the menu option in the bar of my xfce4 desktop worked, but it stopped recently (around the core-updates merge in April?). Since then I have been using "xlock -mode blank". <andreas-e>xlock is a setuid binary in /run/setuid-programs <apteryx>apparently something like "gpg --quick-set-expire YOUR-KEY-ID 2026-06-06 '*'" works if the subkeys haven't expired yet <andreas-e>I always use the key edit menu. And every year I need to look up again how it works. These subkeys are confusing. <andreas-e>And usually I update my key once someone tells me it has expired... <WorldsEndless[m]>andreas-e: ah --- yeah, #setuid is what I couldn't make work with xscreensaver. I remember reading about a solution like xlock; but no images, right? <andreas-e>Well, xlock comes with lots of different screen savers. I juste use "-mode blank" for a black screen. <WorldsEndless[m]>andreas-e: wow! I'm gonna look that one up. It might be a replacement for xscreensaver <apteryx>cbaines: I got "build of `/gnu/store/cz4sr5whzf3qhp8c47bzlfzcxxwl485i-python2-2.7.15.drv' failed" for 'guix time-machine --commit=897f303d2fa61497a931cf5fcb43349eb5f44c14 -- environment --ad-hoc hello -- hello' <apteryx>does someone know why running M-x texinfo-all-menus-update from Emacs on the cookbook breaks the guix build? (make as-derivation). It prints errors such as: "guix-cookbook.de.texi:570: @menu reference to nonexistent node `A ``Hello World'' package'" <apteryx>maybe I need to apply the same treatment to all of them? <apteryx>maybe someone involved in the translations will know <apteryx>the xref_command macro we have in doc/local.mk seems potentially related <apteryx>rekado: my IRC logs are failing me. do you remember the command you recommend me to disable OAUTH code generation? <apteryx>rekado: found it! 'ykman config usb --disable OTP' <Guest52> https://logs.guix.gnu.org/guix the search feature is kinda useless. I searched for rekado since I thought I may could help apteryx. But it just shows stuff from this year january and oauth doesn't even find the mentioned message 5 min. ago <apteryx>rekado: sent to guix-patches, you are in CC <apteryx>Guest52: I've added it to the cookbook to avoid loosing it again <Guest52>No, I mean how. You searched throug the irc bouncer logs? <apteryx>my desktop is nearly always on and acts as an IRC relay/bouncer <Guest52>I guess basically everyone here is using a bouncer. Which one is recommended? <jpoiret>janneke: i think the arch is disabled in the job config directly on ci.guix.gnu.org <rekado>Guest52: there’s a bug in the deployment of the log viewer; it’s as if the indexer writes to a different database than the interface has access to. It last worked before it was moved to a container. <rekado>just checked: there are recent results in the database and they can be accessed from the search interface. <apteryx>is there some better provider than duckdns.org out there for a free dynamic DNS service? <apteryx>duckdns.org works well, but requires google/github/twitter/reddit for the initial setup <rekado>the order of results is weird, though, and there’s a limit, so perhaps that’s where things go wrong <rekado>do you know of freedns.afraid.org? 