IRC channel logs


back to list of logs

<Guest42>I am with a error when using `guix pull` in a VM,
<Guest42>I don't understand why it is complaining about network issues when my network is working inside a VM
***form_fee- is now known as form_feed
<Guest42>in debian paste
<drakonis>ah, guile studio has become much better now
<Guest42>drakonis what is Guile studio?
<flatwhatson>raghavgururajan: FYI i prepared a new guix system vm as a dev environment to eliminate any strange behaviour with my alien host OS, rebuilt gtk4, enabled tests and see them failing
<flatwhatson>the tests all fail due to unable to access the display, there is some "xorg-for-server-for-tests" package which should provide this, next step is to research how that is supposed to work
<flatwhatson>baby steps :)
<Guest42>I got this error:
<Guest42>guix pull: error: some substitutes for the outputs of derivation `/gnu/store/mvf88n2v90jjxg9n8b315p22r6jrkbyb-libx11-1.6.A.drv' failed (usually happens due to networking issues); try `--fallback' to build derivation from source
<Guest42>guix/ui.scm:2117:12: In procedure run-guix-command:
<Guest42>Wrong type to apply: #f
<raghavgururajan>flatwhatson: I see. Thanks!
<daviid>flatwhatson: i think you should provide a similar env then forg-golf it self? ("xorg-server" ,xorg-server) ... then (add-before 'check 'start-xorg-server ...
<daviid>flatwhatson: in line 2044 - 2050
<flatwhatson>daviid: yep maybe, just focusing on the gtk4 tests first, if we're very lucky then once those are fixed everything else will "just work"
<daviid>flatwhatson: right, but gtk-4 tests likely test gdk-4, gsk-4, all need a server and a display ...
<daviid>flatwhatson: so you should start an xserver before to launch the gtk-4 tests, i think
<flatwhatson>oh, you think that testing server needs to be launched manually? that would explain a lot :)
<daviid>and make sure gdk-4/gtk-4 have access to a display,meaning gdk_display_get_default not NULL
<DynastyMic>Hi, I'm trying to package a software to contribute to the guix channel
<DynastyMic>However there are 2 versions of the sowftware
<DynastyMic>A version with Java included and another "platform independent" version
<daviid>flatwhatson: yes,just like for g-golf tests, ithink
<DynastyMic>which one should I use?
<daviid>which is why i pointed to theg-golf pkg def, so you couldjust copy tjuose lines and try again ...
<DynastyMic>or how should I handle such situation?
<flatwhatson>daviid: i misunderstood! yes, that's already present, i guess copied from existing gtk+ definition:
<flatwhatson>maybe it's failing to start though
<daviid>flatwhatson: how about commenting what you have (temporarily) - related to launching the xserverimean - and taking exactly those lines of g-golf package, which differs in a number of ways ... and do not add +extension GLX ? not sure
<daviid>flatwhatson: also, double check with guix gtk-3 with that respect, because very likely gtk-3 also needs a server ...
<daviid>very likely gdk-3/gtk-3 also need a server for their test-suite ...
<flatwhatson>good thinking! firing off a gtk3 build now as a "control" case
<emad-guix>Hi it seem version-control.scm has a bug
***lispmacs[work] is now known as forthmacs[work]
<bavier[m]>if I created a new signing key, do I need to do something with the `keyring` branch?
<ecbrown>yes, you need to put your ascii armored keys in there
<lfam>bavier[m]: I think that item 4 of Commit Access covers it
<lfam>You should also add it to your savannah account
<bavier[m]>lfam: thanks for the pointer.
<apteryx>emestee: looks like a bug indeed!
<apteryx>emestee: sorry, wrong person. I meant to address emad-guix
<apteryx>sneek: later tell emad-guix that indeed looks like a bug
<lfam>bavier[m]: And I think that `make authenticate` will tell you if you've done it right
<bavier[m]>lfam: is it alright to send a message to guix-devel and let a maintainer do the necessary bookkeeping?
<lfam>Yeah, definitely
<bavier[m]>is it recently that 'git commit' canonicalizes name and email from .mailmap?
<bricewge>Is there a way to get the default value of a record field?
<bricewge>I want to know if a field is set to the default value.
<daoistmonk>is guix + systemd possible?
<ixmpp>Dont go there
<bricewge>Guix system with systemd as PID 1 no, see
<bricewge>Guix on a foreign OS with systemd yes
<daoistmonk>is it documented anywhere the design decisions that went into shepard as opposed to systemd?
<drakonis>might this satisfy you?
<drakonis>as far as i'm concerned, its design predates systemd's
<drakonis>hm, this was written in 2003
<daoistmonk>that's great, thanks!
<daoistmonk>in shepherd, is is possible to start services in parallel?
<daoistmonk>i.e. is init single threaded in shepherd and every service is started serially?
<drakonis>hmm, i think it still is single threaded right now
<drakonis>it needs work on that area
<bricewge>soheilkhanalipur: If it's PPPOE connection was working before, your friend could rollback to a previous working system "guix system list-generation" and "guix system switch-generation $generation-id"
<soheilkhanalipur>bricewge: Thank you for your good helps. I will tell him…
<soheilkhanalipur>bricewge: Oh, he deleted generations…
<leoprikler>So PPPoE still works, but LAN is now borked or how to understand this?
<efraim>has anyone gotten a download URI to work that has a space in the name and isn't from sourceforge?
<soheilkhanalipur>leoprikler: He first thought the problem was with LAN cable, so he bought a new cable. But the problem still persists
<leoprikler>it could still be, that the hardware handling the cable connection is borked
<leoprikler>what does ethtool say?
<florosGPL>hello, I installed nginx with guix. The binary is looking under /gnu/store/.../nginx/ for the log directory and the config files. That directory is read only and I can't find anywhere an option to point nginx to a different log directory and configuration file
<leoprikler>are you using Guix System? In that case, there's an nginx-service-type
<leoprikler>if not, you'll have to fake what it does on your foreign distro
<florosGPL>I am using Guix on some embedded board running Yocto. The problem is that I can't give nginx the config file
<leoprikler>why not? nginx has a "-c" switch as far as I can see
<florosGPL>I might've missed that. checking it right now
<florosGPL>actually I think that I found a problem with the configuration of nginx.
<florosGPL>this is the error that nginx fails with: 2021/06/15 08:14:38 [emerg] 1768#0: open() "/gnu/store/m3
<florosGPL>rn3mnxcmj5c0f8ywpjlcraq1m9yp5j-nginx-1.19.9/conf/nginx.conf" failed (2: No such file or directory)
<florosGPL>but, the guix installation has a config file under /gnu/store/m3rn3mnxcmj5c0f8ywpjlcraq1m9yp5j-nginx-1.19.9/share/nginx/conf/nginx.conf
<leoprikler>that might be an issue, but you're not supposed to use the nginx package as-is anyway
<muradm>emacs-next-pgtk updated with native comp seems look very nice, no problems so far, used to work on big java/typescript mixed project with lsp whole night, very fast
<muradm>for now only one issue, is that site-lisp packages from emacs-xyz don't get compiled
<muradm>for obvious reasons, output is readonly
<leoprikler>hm, could you replace the emacs for build with native-comp emacs?
<leoprikler>perhaps native-comp emacs-minimal
<muradm>hmm, just a sec
<muradm>leoprikler: basically you want emacs-minimal-next with gcc comp
<leoprikler>not necessarily, that'd be for future
<muradm>leoprikler: that is no possible i think, emacs-minimal is from 27.x which has no native comp
<leoprikler>right now it'd suffice if emacs-next-pgtk-native-comp could native-comp your xyz package
<muradm>emacs-next is from 28.x which has it and can be compiled
<leoprikler>i.e. you inherit the xyz package, but take native-comp as #:emacs
<muradm>ah, ok
<muradm>leoprikler: looks too complex, i use only 2 packages from guix, rest come with straight.el, which are mu (with mu4e) and pdftools, basically it is not critical to have them precompiled, and these come from gnu-build-system, not emacs-build-system
<leoprikler>fair enough
<muradm>as far as i see they do something like (emacs-generate-autoloads ...)
<muradm>these tasks will need to have additional step for ahead of time compilation
<flatwhatson>i'm fairly sure that it will native-compile your site packages, lazily
<flatwhatson>the .eln are written to ~/.emacs.d/eln-cache by default
<flatwhatson>regardless of where the source .el comes from
<flatwhatson>but yes precompiling as part of emacs-build-system would be nice!!
<muradm>flatwatson: yes it tries, and gets error in the end that output could not be written, and eln-cache does not contain these
<muradm>and tries to do that on every startup of emacs
<leoprikler>that's strange and probably a bug in the native-comp logic then?
<muradm>no, it is fair, if package source path under site-lisp, it should be eln'ed there as well
<flatwhatson>but that's not how it works, it writes them into ~/.emacs.d/eln-cache
<flatwhatson>well, it writes them to the first writable entry of native-comp-eln-load-path
<muradm>("/home/muradm/.config/emacs/eln-cache/" "/gnu/store/yf67yydcc5kwldq3z7pgdgwgprxhndbn-emacs-gcc-pgtk-next-28.0.50-1.794ec93/lib/emacs/28.0.50/native-lisp/")
<flatwhatson>ok, so it should be ~/.config/emacs/eln-cache in your case
<muradm>as far as i read, user packages go to first, site-lisp to second
<muradm>amd it can't write ther
<flatwhatson>the source of the packages doesn't matter, eln paths are not relative to the el path
<muradm>/home/muradm/.cache/guix-extra-profiles/desktop/share/emacs/site-lisp/mu4e.el: Error: File error Opening output file
<muradm>if they would be precompiled to somewhere ../lib/emacs/.../native-lisp/ that would be ok
<muradm>on the other hand, it would be unsafe to have source unnder site-lisp, and binary under user eld-cache
<muradm>that is security flaw
<muradm>if user have right to manipulate loaded binary
<leoprikler>those rights are normally given to the user in an emacs setup
<muradm>user can't edit site-lisp, but can edit native cache? )
<leoprikler>user can't edit site-lisp, but user site-lisp supersedes system site-lisp
<leoprikler>so yeah
<leoprikler>there is no security in emacs
<muradm>lets put it this way, I'm admin, and in my company i want site-lisp have common code, which user can replace in it eln-cache?
<leoprikler>you can put common code to site-lisp, but you can't guard against your employees saying "screw that code, I'm using this"
<leoprikler>this has nothing to do with eln-cache
<muradm>of course, i'm just trying to point out the logic of why precompiled binaries of site-lisp should no end up in user's eln-cache imho
<leoprikler>users can already set up their emacs to not read your common code
<flatwhatson>having someone else in control of the code your system runs is completely antithetical to the whole idea of emacs...
<muradm>by this logic, right now if I want/have to edit mu4e.el under site-lisp, i have to run sudo emacs
<leoprikler>you can just copypasta mu4e.el to your ~/.config/emacs, chown it and edit it there
<flatwhatson>(assuming ~/.config/emacs is on load-path)
<muradm>yes, and it will be another mu4e.el, of course, whose precompiled binary will go to .emacs.d/eln-cache
<flatwhatson>anyway, how you imagine it works is not how it actually works
<muradm>then one under site-lisp/mu4e.el should go to lib/emacs/native-lisp whatever
<leoprikler>maybe if precompiled, but not if autocompiled
<muradm>hmm, may be because of NATIVE_FULL_AOT
<ngz>Ewwww. Rust killed me (again). Or Cargo, I cannot even recognize my assassin.
<muradm>lisp with emacs gets precommpiled
<leoprikler>killed how?
<ngz>I get an undecipherable compilation error
<ngz>I mean, the error itself is clear, but why it happens in the building environment is not.
<muradm>New compiled files are deposed always in the first entry of comp-eln-load-path. from:
<muradm>you're right
<muradm>should end up as per description under first eln-cache dir
<flatwhatson>first writable entry on comp-eln-load-path, unless native-compile-target-directory is set
<flatwhatson>(with different handling for byte+native-compile, which is the bootstrap phase during emacs build)
<flatwhatson>so... if it's not putting them into your ~/.emacs.d/eln-cache or ~/.config/emacs/eln-cache, something is wrong!
<muradm>for example: ls -la /gnu/store/j602ji3axnpzkf780wahmxq9arbcl5xc-mu-1.4.15/share/emacs/site-lisp/
<muradm>includes both el and elc
<muradm>so it is fair to expect eln to be in the same location
<muradm>i would not say that it is a bug, but a feature )
<muradm>that probably can be fixed by setting native-compile-target-directory
<muradm>i will restart to see if it works
<florosGPL>I fixed the error with nginx, I copied everything to a different folder and set -p pointing to the new "share" folder. nginx runs flawlessly, but it can't run with guix under any circumstances due to the read-only filesystem
<muradm>faltwatson: native-compile-target-directory didn't work, still same, looking closer to its explanation it is not for deffered compilation
<muradm>so basically now as far as i understand it just tries to write .eln next to .elc
<muradm>regardless if it is bug or feature )
<flatwhatson>wait, are you talking about runtime or build time?
<flatwhatson>what does (comp-el-to-eln-filename "/home/muradm/.config/emacs/init.el") eval to?
<ngz>Ah! ngz : 1 -- Rust : 0.
<leoprikler>are you winning?
<ngz>I think so.
<ngz>It was a bogus entry in the inputs provided by the importer.
<ngz>So I guess the bogus data was upstream.
<ngz>Now, I let the CI arbitrate.
<muradm>flatwatson: /home/muradm/.config/emacs/eln-cache/28.0.50-ffa276e2/init....eln
<clone>Is there a way to download additional files in a package definition? I'm trying to package a program with a set of patches but some of the patches rely on additional source files. I can add patches to the definition's source, but i don't see anyway to add the source files
<flatwhatson>muradm: right, that is where async native compilation will try write files when doing "deferred" compilation (regardless of source location)
<muradm>flatwatson: this error: /home/muradm/.cache/guix-extra-profiles/desktop/share/emacs/site-lisp/mu4e.el: Error: File error Opening output file
<muradm>might be something else to do with mu4e it self
<muradm>i checked other things from site-lisp getting loaded, some of them are compiled correctly
<flatwhatson>yes, i think so, try triggering it after M-x toggle-debug-on-error
<ngz>clone: package inputs can include full (origin ...) sexp.
<muradm>flatwatson: toggle-debug-on-error does nothing for it
<muradm>full async compilation output on emacs start
<muradm>may be it is something to do with all those unknown symbols preventing compilation?
<muradm>but error message misleading then )
<flatwhatson>muradm: do you have a read-only /tmp?
<muradm>no, touch /tmp/test works
<flatwhatson>native compilation actually writes an elisp script into /tmp and executes it
<muradm>yes, i see them now
<muradm>there are few remained
<muradm>giant oneliners )
<civodul>did anyone else notice that Git bug: ?
<flatwhatson>muradm: strange, it could very well be something specific to mu4e
<flatwhatson>anyway i think you've demonstrated that deferred compilation is actually working, just not for that one
<muradm>flatwatson: yeah, it works fine except for mu4e and few others which are not critical any way
<muradm>i'm happy with ultra fast lsp now
<muradm>it just got so more responsive than before, i tried to use your emacs with gcccomp few times, but never had time to wait for libgccjit compilation )
<muradm>flatwatson: maybe you can put this to your channel also :)
<muradm>will make life so much easier )
<muradm>for now
<muradm>is it so important to have libgccjit-10 instead of 9?
<flatwhatson>well, it's slower due to some buggy handling of string constants requiring a workaround from native-comp
<flatwhatson>i don't really want to add more package definitions, but you could make a personal channel which wraps emacs-pgtk-native-comp with default libgccjit
<muradm>flatwhatson: faster anyway after no-comp at all )) i keep packages locally for now and use "guix system/package/etc -L ~/.config/guix ...." which is cheaper for now, no issue for me
<muradm>flatwatson: do you know how to change default user "eln-cache" directory? will it be enough to change first item in native-comp-eln-load-path in early-init.el?
<muradm>trying it now
<muradm>worked fine
***dekenevs is now known as mitzman
***mitzman is now known as kitzman
<flatwhatson>yep, that's the way to do it
<abcdw>hi guix!
<soheilkhanalipur>leoprikler: I do not know how to work with ethtool. Please help me. What command should I enter?
<thorwil>does ardour as packaged in guix support VST3? i see nothing in the packaged definition that would explicitly enable or disable that. Preferences have a VST page, but so far no luck getting a VST plugin to work.
<muradm>soheilkhanalipur: what do you want ethtool for?
<muradm>soheilkhanalipur: better to see your guix configuration
<soheilkhanalipur>muradm: I'm not a Guix professional. Please tell me step by step what I should do!
<leoprikler>soheilkhanalipur: they want to see your guix.scm
<leoprikler>ethtool can be used to test your interface, e.g. ethtool -t eno1
<soheilkhanalipur>leoprikler, muradm: I am surprised that this problem occurred by itself!
<soheilkhanalipur>So, as I said, maybe this problem is due to Ungoogled-Chrumium?
<leoprikler>I really can't tell you more from where I am
<leoprikler>Well, even if I had physical access, I'm not that much of a hardware person
<soheilkhanalipur>leoprikler: OK, thank you
<thorwil>the answer to my question is in About -> Config, which includes "VST3 support: True"
<soheilkhanalipur>Please help 😭
<soheilkhanalipur>Do I need to edit config.scm?
<soheilkhanalipur>Any help 😕
<apteryx>civodul: hello! Is it possible to use the returned value of gexp->derivation as the input of another gexp?
<apteryx>As part of the debian-archive generator, I'd like to simply call self-contained-tarball, then gexp-unquote its result into the builder, but that currently gives a gexp-input error.
<civodul>apteryx: no, gexp->derivation is a monadic procedure, so you can't do that
<civodul>however, you can use computed-file, which is the non-monadic equivalent
<apteryx>In the G-Expression section of the manual, it says: "When a high-level object such as a package or derivation is unquoted inside a gexp, the result is as if its output file name had been introduced." Is the output of gexp->derivation not the same kind of derivation mentioned there?
<apteryx>could I rewrite self-contained-tarball in terms of computed-file, so that it can be composed into other g-exps (e.g., the builder of debian-archive)? Would that entail to adjust something else in the (guix scripts pack) code?
<apteryx>hmm, probably easier to factorize out the builder from self-contained-tarball and reuse just that bit
<apteryx>do we still have tar versions that do not support --sort in our bootstrap binaries?
<efraim>yes. it spits out a warning and continues on
<apteryx>OK, thanks for confirming.
<ngz`>efraim: Hello. I'm contemplating your rust-ndarray-remove-blas-src-dep.patch. I'm probably facing the same issue, i.e., blas-src depends on a non-free crate.
<ngz`>However, I'm wondering if the solution would be to remove the crate from blas-src instead of working at a higher level.
<ngz`>(oh, it looks like I was banned from #guile oO;)
***ngz` is now known as ngz
***terpri is now known as robin
<maximed>Hi Guix
<ixmpp>Can even nest the home directories
<ixmpp>My system is essentially single user
<ixmpp>So why not make use of "users" as environments
<dstolfa>ixmpp: how would that differ from profiles, or would it just be the same idea as profiles, but for your home data?
<dstolfa>(if so, maybe it should be a home profile?)
<ixmpp>Basically yeah. And you then can actually make use of group permissions for data too
<dstolfa>yeah, that makes sense and would be cool imo
<ixmpp>Neat. I might try it
<leoprikler>it does kinda feel like a weaker sandbox environment tho
<abcdw>debbugs doesn't join patches into to thread by in-reply-to header?
<efraim>ngz: I suppose if you want to create a fake blas-src crate that would also work
<abcdw>What is the best way to join bug reports into one? merge all consequent tickets to first one, or resend patches to the first thread and close all other tickets?
<ngz>efraim: But more importantly, could you explain me how you got this patch off? I'm a bit puzzled.
<ngz>IIUC, you're patching Guix auto-generated Cargo.toml, so it wasn't generated from a cloned repository.
<ngz>(I'm trying to generate the same for ndarray-0.13)
<ngz>abcdw: the former is usually what we do.
<abcdw>ngz: ok, will try to merge them. `merge 49045 49046 49047 49048` should work.
<bricewge>abcdw: Debbugs joins patch with in-reply-to headers but you need to wait for the referenced header to be processed first
<abcdw>bricewge: got it, thank you, will wait next time)
<solene>how could I list the available packages from the gnu/packages/*.scm files? I would like to work on
<abcdw>Seems merge completed sucessfully. Am I understand correctly: They will be closed automatically once one of them is closed?
<dstolfa>solene: guix search <package-file-name>
<dstolfa>so if it was text-editors.scm, you would type in `guix search text-editors`
<dstolfa>that's how i do it, anyway
<dstolfa>oh you mean in scheme
<tschilptschilp23>Hi! I'm rather new to writing package definition, and do have problems debugging. The package I'm just trying to define is a C-library for number theory ( that I see my collegues using on Debian and CentOS. If I enter an environment with gcc-toolchain, gmp and mpfr it compiles just fine. But once I go with writing definitions things fail at configuration phase. The program just answers with the configure --help
<tschilptschilp23>output. I think it can't really take the default configure-flags passed by gnu-build-system, because out of the ones passed there's certainly not a single one available in the program's configure-script. I already understood, how to ~add~ flags to the default one, but I would't find anything to remove the default ones (and kind of badly feel, that this is not the masterplan). Am I missing something, or does this tell me to go with
<tschilptschilp23>trivial-build-system and (hand-in-hand with that) go deeper into guile?
<dstolfa>not sure, i guess you'd have to see what the module provides?
<solene>dstolfa: yes I would need a command to list all packages available, in scheme
<solene>I'll search how "search" works
<mbakke>oh, the latest Poppler depends on Boost, adding a hefty 149 MiB to its closure size :/
<dstolfa>solene: i'm not sure... i'll let someone who's messed with it before answer. ultimately i think they're all exported in the module
<solene>dstolfa: what module?
<dstolfa>solene: (define-module (gnu packages networking) ...) for example
<dstolfa>so all the things in there that are marked as (define-public ...) should already be accessible
<dstolfa>it's just that i'm not 100% sure how to write the guile code to actually list everything out
<dstolfa>tschilptschilp23: might be worth sharing your package definition so others can see it maybe?
<mbakke>tschilptschilp23: that problem is common with "hand-written" configure scripts that bail on unknown flags ... you have to override the 'configure' phase to only pass your #:configure-flags, i.e. (replace 'configure (lambda* configure-flags #:allow-other-keys) (apply invoke "./configure" configure-flags)))
<mbakke>I forgot a #:key in that lambda
<tschilptschilp23><dstolfa "tschilptschilp23: might be worth">
<tschilptschilp23>mbakke: thanks for the expression! I will try that -- it sounds pretty much what I am missing on 'scheme-words' yet :)
<efraim>ngz: sorry, I wandered away for a bit. It was a while ago but I image I patched the result of $(guix build rust-ndarray@0.12)
*mbakke has to go, good luck!
<tschilptschilp23>dstolfa: this code is from yesterday or so, I just understood by today, that for my needs ```use-modules``` (instead of defining the whole flintlib module) and ```package flintlib``` (without defining it public) would be smarter and less complex. Also quite some of the ```#:use-module``` parts are superflous yet!
<dstolfa>tschilptschilp23: outside of what mbakke said, one minor nitpick would be that it doesn't need the "maintained by..." in the description :). looking forward to it building & working!
<ngz>efraim: Yes, I grabbed the file from guix build -K (regular build fails, obviously). This is were I saw the Cargo.toml to patch. But apparently, my patch-fu is weak, because the patch I generated isn't accepted. If you have a few minutes to share, I'd like to know what you remember about that process.
<apteryx>oh, progress: /gnu/store/qq8kvlmbi7v2r9m3fdzd5sigmc4maign-deb-pack.deb: Debian binary package (format 2.0), with control.tar.gz, data compression gz/
<apteryx>installation failed because: package architecture () does not match system (i386). Probably the magic value in the ar archive mentioned in 'man deb'
<apteryx>not sure how to bake this in
<pkill9>does anyone run guix system on a pinebook pro?
<dstolfa>i think vagrantc does
<pkill9>im thinking of getting a slim laptop for basic use
<iskarian>Hello guix :)
<tricon>iskarian: Hello.
<tricon>pkill9: Ooh! Hadn't seen these. I'd like to switch to ARM for my computing.
<apteryx>if you're ever curious to know what your current machine GNU triplet is: $(guix build config)/bin/config.guess
<DynastyMic>Hi, I'm trying to depoloy the protégé package to the Guix channel, however I want to know if I should upoload the platform independent version or the Java included one
<DynastyMic>or should I make two separete packages for each
<DynastyMic>how could the Guix community benefit more from
<nckx>Three billion devices run Java. How's the ‘platform-independent’ version ‘platform-independent’? Naively, I'd say that one, but I'm awaiting a gotcha like it being in Node.js.
<nckx>(Eventides, Guix.)
<dstolfa>nckx: how is life
<dstolfa>better now that i have finished merging around 50k commits on a force pushed history dating back to 1980s with all hashes recomputed
<dstolfa>(to be clear, the force push was kind of necessary, but it still made my life miserable for 2 days)
*nckx probably doesn't want to know…
<dstolfa>well it's not all that terrible (it's a result of freebsd moving from svn to git), but i'll spare you the details :D
<iskarian>ah, efraim: would you be interested in reviewing my gccgo patches, since you were involved in this before?
<tricon>Are there any known libre Bluetooth 5.0 adapters?
<nckx>dstolfa: Orly? It's been too long since I've used FreeBSD.
<nckx>This wasn't something many users were expected to do, surely?
<dstolfa>nckx: the user transition was relatively straightforward, the tools just switched to git from svn and continued working as usual. however, we've forked the original github `master` branch that was a mirror back in 2015 or so and continued working on it quite extensively. after the migration to git happened, `master` was frozen and a new branch `main` was created which had completely different hashes
<dstolfa>so in order to merge the new code, i had to deal with reconciling old hashes on a fork from `master` and new hashes from `main`. the reason a force push was needed IIRC is because something in the mirroring didn't preserve a couple of important things on the freebsd repo, so a fully new conversion was needed
<nckx>Hah. Heh. Ho.
<jackhill>efraim: thanks for reviewing/pusihng the goffice update
<apteryx>there's something apparently very computationally intensive (derivation wise) in doing: './pre-inst-env guix pack --target=i686-linux-gnu hello'
<solene>when I look at "easy" issues on the issue tracker, I wonder what the hard one look like :/
<efraim>iskarian: very much so. hopefully tomorrow or the day after
<apteryx>perhaps it was just downloading stuff in the background although I used -v3. Hmm. Now it's flying.
<raghavgururajan>Hello Guix!
<ngz>efraim: actually, it looks like you used git diff. I'm probably missing something obvious.
<apteryx>eh, dpkg has humor: unable to clean up mess surrounding './gnu/store/01b4w3m6mp55y531kyi1g8shh722kwqm-gcc-7.5.0-lib' before installing another version: Read-only file system
<nckx>So… not that file, but something near it?
<nckx>That's Guile-tier error reporting.
<efraim>ngz: it was probably `tar xf $(guix build -S rust-ndarray); cd rust-ndarray<tab>; git init, git add .; git commit -m "start"; $EDITOR Cargo.toml; git diff -p > ~/path/to/repo/gnu/packages/patches/rust-ndarray-no-blas-src.patch'
<ngz>OK, so you re-created a git repo in order to generate the patch.
<efraim>it looks like the raw git diff is
<ngz>Yes, this is what I was looking at. I just generated a new one based on your technique above.
<dstolfa>nckx: btw, i lost all my logs... what changes did you want me to make to the last diff i submitted for strongswan again?
<efraim>I've also learned that if I want to keep something in my git stash then I can use `git stash show stash{0} -p | git apply`
*dstolfa should start logging irc so this doesn't happen
<nckx>dstolfa: Uh oh. Ehm. Are they not on
<nckx>(I know, the search is broked, that aside.)
<dstolfa>oh, they might be
<nckx>Middle-click & C-f have served me well enough for now.
<dstolfa>oh, no changes
<dstolfa>that's why i don't remember it
*dstolfa does other things then
<iskarian>efraim, can't you just do `git stash apply stash{0}`?
<nckx>(Phew, I'm not as senile as I thought.)
<dstolfa>nckx: it's fair to assume that anything in the past two days has been a bit of a blur for me, i spent way too long staring at git logs and ediff
<dstolfa>(so if anyone is misremembering things, it's probably me)
<efraim>iskarian: I've never tried that
<efraim>looks like you can. I'll try to remember that for the next time
<raghavgururajan>How to test a package that uses polkit authentication and provides `.policy` file?
<raghavgururajan>Either in pure or unpure env, the already running polkitd will not be aware of that policy file provided by that package right?
*apteryx got the first guix pack produced .deb archive installed on a Debian 10 VM :-)
<apteryx>nckx: yeah, not sure exactly :-)
<nckx>That's really the feeling you want your users to have after printing something scary 👍
<civodul>apteryx: woohoo, fun!
<civodul>maybe it's the kind of .deb that would make vagrantc scream ;-)
<solene>I fixed a 2 years old easy bug \o/
<apteryx>civodul: haha, probably!
<apteryx>I installed hello so far
<apteryx>I'll try to install perl on top of it
<civodul>solene: nice, thank you!
<apteryx>one thing I noticed is that the #:target argument of the guix pack generator only takes --target. If I provide --system the generator has no means to know that the target != host.
<apteryx>solene: wooh :-) thank you!
<civodul>apteryx: the generator doesn't need to "know", normally
<civodul>you'd use #+ (ungexp-native) or #$ (ungexp) as needed, depending on the use case
<civodul>or did i misunderstand your comment?
<apteryx>OK. The deb control file has an Architecture: field that specifies the Debian machine type. Dpkg barfs when it's missing.
<civodul>oh i see
<civodul>(guix docker) does something similar
<apteryx>so I'm extracting this from #:target, else the current machine. But for QEMU user emulation that doesn't work (it uses uname, which doesn't change).
<civodul>see the build-docker-image call in (guix scripts pack)
<apteryx>Yeah this was enlightning, the #:system (or #$target (utsname:machine (uname))) part
<civodul>(utsname:machine (uname)) does change under emulation, i believe
<apteryx>I do not thing so, based on my tests.
<civodul>i just checked (running the result of "guix build -s aarch64-linux guile" on my x86 laptop), and i get "aarch64"
<civodul>should be fine!
<civodul>then i think Debian has different names, like "arm64", etc.
<apteryx>right, I made some translator for that based on the data/cputable file in dpkg sources
<apteryx>perhaps my observation was valid only for i686-linux
<civodul>ah yes, that one's not emulated
<civodul>but yeah, it can be a problem
<civodul>so the docker code is wrong in that case
<apteryx>kind of a corner case really :-)
<apteryx>Unpacking deb-pack (0.0.0) over (0.0.0) ...; seems I'll have to give the pack some less generic names, else they just overwrite each other
<civodul>we should use %host-type instead of uname
<civodul>(it's a built-in Guile variable)
<apteryx>OK! Neat.
<civodul>that one's known good
<civodul>re names, that looks fun :-)
<civodul>does dpkg check the file list before unpacking or does it just go ahead?
<apteryx>you mean if it's optimizing to unpack only things that are not already installed?
<apteryx>Based on a previous error when I had Guix running on the VM (the store was ro, dpkg couldn't touch it), I'd say it first deletes an existing file then install the new copy.
<apteryx>if you scroll back a bit you'll see the error (Read-only file system)
<solene>good night guix
<rekado_>I just upgraded my old thinkpad after a few months offline; previously it would boot straight into xfce, but since the upgrade I get an opaque gdm error.
<rekado_>I already erased /var/lib/gdm and ~/.cache/gdm, but to no avail
<rekado_>bit puzzling, because I deleted gdm-service-type ¯\_(ツ)_/¯
<rekado_>any ideas why gnome-session-binary and gdm are running on this system even though it is reconfigured with gdm-service-type deleted from %desktop-services?
<rekado_>oh, it’s from set-xorg-configuration!
<civodul>rekado_: oh yes, that's kind of a trap
<civodul>apteryx: i was wondering whether it detects "collisions" like when the .deb it's about to unpack provides files already there
<civodul>from what you say it just goes ahead and overwrites things
<civodul>brute force!
<civodul>that's reasonable in this case