<nckx>Yes, but it's extremely simple. <nckx>For something that would have blown my mind 2 decades ago. <nckx>mehlon: Native pulling, you mean? <mehlon>no, the qemu thingy, I think that works now <pkill9>is it possible to have guix offloading also build the dependencies on the remote? <pkill9>the manual says it copies them over <nckx>pkill9: When they are already available locally? I don't think so. Why? Which insane performane/bandwidth ratio would motivate that? <nckx>g_bor[m]: Oh, thank you! <pkill9>i'm thinking when using a remote over the internet <nckx>I only ever offload over the Internet. <mehlon>guix seems to complain that "I am a x86-64-linux machine" <pkill9>nckx: i was under the impression from reading the manual that guix will send the dependencies required to build from the local to the remote <pkill9>but i guess not if you only ever offload over the internet <nckx>pkill9: No, your reading is correct. To keep things concrete: if my laptop has gcc, and for some reason my build server doesn't, my laptop will send the gcc binary along with the request (derivation) to build hello. It won't ask the build server to build gcc from scratch. <pkill9>it's just that my internet upload speed isn't great <pkill9>i would have thought it would instead download those binaries from it's own substitutes <leoprikler>sneek later ask raghav-gururajan whether they can depropagate clutter from the packages they've contributed. It appears to break gst-play. <nckx>pkill9: Good point, I don't use substitutes (and if I would I'd prefer to use them as little as possible) but I can see your point. <yrk>kmicu: what wasn't accurate, in particular? <pkill9>even if substitutes aren't available, it would possibly be faster just to build on the remote too <nckx>That sounds like a very costly feature to support because of the way offloading vs. substitution are ‘layered’, though. <nckx>It becomes a much more complicated negotiation. <sameerynho>what is the emacs-guix exactly ? is a emacs bundle or just a elisp library ? <nckx>Server: ‘Please wait… nope, couldn't find GCC and x and y, please send those now.’ <nckx>sameerynho: Just an emacs library package. <nckx>pkill9: That might sound more negative than intended; I just mean than offering the choice isn't trivial. <nckx>sameerynho: Thank you! What's the source page? *nckx never uses the Web manual so is a lost puppy here. <sameerynho>nckx: well it's easy to read them on an ebook reader <apteryx>could someone explain to me the relationship between .pc files and the need to propagate the dependencies it lists? <sameerynho>nckx: yupe after a refresh It's gone, do we do caching on these docs in some sort ? <nckx>I'm as puzzled as you are. Not that aggressively. OTOH I've heard that mainstream browsers often aggressively cache (despite any headers to the contrary) to ‘improve speed’. That wouldn't suprise me, but I'm not a Web person. <nckx>The dead emacs-guix link is still there though, I just don't know where to point it. <nckx>.gitfoo.io pages always have an air of impermanence about them but I'm biased. *sameerynho looking up "impermanence" <sameerynho>quick question, I'm far from finishing the docs. Is it possible to define the state of a system in scheme and ask guix to deliver such system with that description ? ( when guix is installed in a different distro ) *nckx finally found out about ‘doc/htmlxref.cnf’, that was a wild grep ride. <nckx>sameerynho: Not ‘system’ in the ‘Guix System’ sense because on distroes Guix only manages profiles (roughly: your installed Guix packages), nothing else. You can fully capture the state of a profile with a manifest (documented). <nckx>mehlon: Yeah, it's not just at the bottom but it's very obvious there. Thanks. <mehlon>so.... how do I setup an environment with gcc and automake and all that stuff? <nckx>mehlon: ‘guix environment --ad-hoc gcc-toolchain automake the-stuff’. <nckx>automake usually requires ‘autoconf’ too. <leoprikler>sneek: later tell raghav-gururajan perhaps my issue can also be resolved by updating cogl. To detail it, `gst-play-1.0 --videosink=coglsink` (with coglsink being the default apparently) fails for any video, at least for me. ***daviid is now known as Guest91543
***Guest91543 is now known as daviid
***jonsger1 is now known as jonsger
<jackhill>nckx: I'm seeing a hash mismatch for font-iosevka <jackhill>nckx: you're welcome, and thanks for keeping the package updates flowing ☺ <nckx>It is my purpose in this life. <nckx>jackhill: The mismatch was for the -slab variant, which inherits from the updated font-iosevka. I've updated the hash, and added a note to -iosevka for the future. I'll push soon, unless you need it immediately. <jackhill>nckx: cool, thanks. No rush, I just hoard all the fonts :) <lafrenierejm>I'm working on adding a package, but during build I get an error "File $filename is read-only; trying to patch anyway". Any ideas how to work around this? <apteryx>pkg-config question: Requires.private can go to inputs rather than propagated-inputs, right? <nckx>lafrenierejm: You need to make them writable yourself, see the java-xom package for an example phase. <nckx>apteryx: I think so. It's certainly a good idea to avoid propagation whenever possible. <sneek>Welcome back raghav-gururajan, you have 2 messages. <sneek>raghav-gururajan, leoprikler says: whether they can depropagate clutter from the packages they've contributed. It appears to break gst-play. <sneek>raghav-gururajan, leoprikler says: perhaps my issue can also be resolved by updating cogl. To detail it, `gst-play-1.0 --videosink=coglsink` (with coglsink being the default apparently) fails for any video, at least for me. <apteryx>nckx: thanks for the input (no pun intended). <oriansj>rndd: you need (set-current-output-port (open-output-file "test.txt")) then display and write will automatically go to file test.txt <oriansj>Then you can just do (main) and main will just output to the file <civodul>i'd recommend: (call-with-output-file "file.txt" (lambda (port) (display "hello!\n" port))) <civodul>that ensures the file descriptor gets closed when the lambda exits <leoprikler>Sorry for my weird patch series, where each patch gets its own bug ID. How do I configure git send-email to not do this? <lxsameer>hey folks, is it possible to install guix stand alone with no root access and use it too install a profile for dependencies of an specific software ? <leoprikler>there are some attempts at rootless Guix and it seems possible, though I am not really sure about usability <leoprikler>depending on the context, it might be easier to `guix pack` the desired software on an existing Guix system and send it over to the one you desire <lxsameer>leoprikler: i want to use it as a dependency manager for emacs <lxsameer>leoprikler: specially for external dependencies such as platinium searcher or things that are no on melpa <nckx>Good morning Guif! Gueff whof back from the dentift. <leoprikler>lxsameer: Hmm, that will be hard to achieve with `guix pack` unless you can send over a never-changing for years, known to work configuration <lxsameer>leoprikler: yeah, i think guix itself can be a good solution if I can make it work without root and be able to change the store path to someplace else <nckx>raghav-gururajan: That (almost) parses as (nonsense) Dutch. <leoprikler>Perhaps you don't even have to change the store path, which is great for using substitutes. <lxsameer>leoprikler: that's true but i don't want to conflict with the system wide guix installation <nckx>raghav-gururajan: I know 😉 <leoprikler>why would you want to install rootless guix on top of that? <lxsameer>leoprikler: i can't pressume my users to have guix installed <leoprikler>Okay, I take everything back, that's really what `guix pack` is for. <leoprikler>In case your users do have guix: Provide the package. In case they don't: provide the pack as well. <lxsameer>leoprikler: ah guix pack works for the bundle itself, but i want to manage the dependencies of my bundle via guix because <lxsameer>i want to be able to lock versions rollback and use other goodthings about guix <leoprikler>your dependencies are implicitly managed by guix already, though? <leoprikler>assuming your bundle works fine with upstream guix alone, you can just jump to a random guix commit, evaluate the bundle and have a release <leoprikler>if it requires custom sources from a channel, there's a bit more overhead, but in principle it's the same <leoprikler>For reference, you might want to look at guile-studio (which packs everything together in its own package) or gnome/xfce/other meta packages, which simply propagate dependencies. <lxsameer>leoprikler: hmm but you're assuming that the system wide guix is available right ? <g_bor[m]>I was wondering how could I get the getter of a field of a define-record-type*. <leoprikler>The only assumption I make, is that you can use Guix for deployment, which should be sound. <leoprikler>g_bor[m]: There's probably an internal procedure generated, that you can access through @@ <leoprikler>You'll have to understand the syntax transformer to get to its name, though. <g_bor[m]>I was wondering how to make one of my procedures more generic, and I would not like to pass in the getter if that can be avoided. <leoprikler>lxsameer: If you package your bundle using guix, you already have the package description for Guix users. <lxsameer>leoprikler: thanks I'll have to try it then <leoprikler>All that's left is to `guix pack` the very package you just created for non-Guix users. <leoprikler>As doubly-relocatable tarball, as squashfs image, you name it. <g_bor[m]>leoprikler: I will show what I have in mind. <g_bor[m]>I would like to have a (record-remove-field record-type field pred record) instead, so that I can define these helpers in terms of that. <leoprikler>i.e. you type out (record-remove-field shepherd-service-requirement pred record) <leoprikler>alternatively you could define a syntax transformer like (define-record-field-remover name record field) <g_bor[m]>leoprikler: I still need the field, in th start of the row after inherit. <g_bor[m]>So this contains both the field and the accessor as far as I can see... <g_bor[m]>(record-type (inherit record) (field-name (operation (get-field record))) <g_bor[m]>so I am looking for something like (getter-of record field)... <leoprikler>fwiw it appears that Guix records are implemented on top of SRFI 9 <leoprikler>Try having a look at (srfi srfi-9 gnu) set-field and set-fields. <leoprikler>(Either using them directly or reimplementing them for your purpose). <g_bor[m]>these do not provide the earlier value I believe. <g_bor[m]>I did a display on a shepherd-service in a repl. <g_bor[m]>and for example <shepherd-service>, the associated record type is not exported. <g_bor[m]>I also displayed that, but it does not help.. <g_bor[m]>it seems that these ar only attached on the syntax level. <leoprikler>Writing your definition as part of (guix records) should work though, should it not? <g_bor[m]>leoprikler: I believe that the records could be extended to provide this info, but as of now I think I will just pass in the accessor... <g_bor[m]>But I think it would be beneficial on the long run to add the info needed for reflection to the record-type. Wdyt? <g_bor[m]>in this current form <shepherd-service> does not seem useful... <g_bor[m]>however this might have to be done on the srfi-9 level instead... <leoprikler>I don't think we should have to export each and every accessor. <leoprikler>Try set-fields on a <shepherd-service> inside the module where its defined. <g_bor[m]>But it would be useful to be able to get to them through the defined record type. <g_bor[m]>leoprikler: that works fine. But does not acheive what I am looking for. I can set the field to whatever I want, but can't get the old value. <g_bor[m]>leoprikler: how do you think so? could you elaborate? <leoprikler>(set-fields record ((field-name) (my-transformation (field-name record)))? <leoprikler>field-name is the full name in both cases of course <g_bor[m]>ok, I will tray that and see what happens <g_bor[m]>Not the way I expected, I thought that the field name is the importent, but actually the accessor is... <leoprikler>It's clearly documented in "info guile set-fields" though 😐️ <civodul>i'll almost surely have to go for a walk in Paris to switch train stations :-) <jonsger>I hope the date of my oral examination will not cross my plans to attend... <g_bor[m]>I observe a strange behaviour when I run guix system build from a login-program <g_bor[m]>on the first run it fails with a strange error message, but on the second run it proceeds as expected. <leoprikler>what do you mean by login program? immediately after logging in on a VT? <g_bor[m]>leoprikler: I added a gexp as a login-program filed to a mingetty service, like it is done in the installer. <g_bor[m]>actually it does not matter that this is run as a login program, i believe. <leoprikler>if that login program runs before your shell is initialized, your environment will be borked <civodul>g_bor[m]: what's the strange error message? <civodul>i guess few people use "guix system build" as their login program :-) <g_bor[m]>it's unexpected path from substituter, I will send the exact message soon. <g_bor[m]>civodul: a very similar thing is done in the installer. <g_bor[m]>I was just trying to come up with a minimal reproducer. <g_bor[m]>It just does and (invoke guix system build ...) essentially. <g_bor[m]>I am building a version where I don't start it automatically, so that I can test running it from a vt. <civodul>g_bor[m]: it'd be great if you could paste the error or post a screenshot <g_bor[m]>it's just that I can't do this guix system vm as it fails on the read-only store. <g_bor[m]>civodul: what do you think about the outreachy blogpost draft? <civodul>g_bor[m]: ah sorry, i haven't checked yet <civodul>would be nice to not send two posts on the same day <civodul>what's the deadline in your opinion? <civodul>g_bor[m]: so the deadline to post the article? *raghav-gururajan is deparately looking for mentor(s) <g_bor[m]>It does not include any concrete info beside the proposed topics. <g_bor[m]>maybe the timeline and the eligibility info page should be linked.... <g_bor[m]>raghav-gururajan: also, if you find something interesing among the proposed project, and a mentor does not show up for your proposal, then we would be happy to work on that with you. Does that make sense? <g_bor[m]>But I hope that someone will eventually turn up :) <g_bor[m]>civodul: guix system: error: got unexpected path <store-path-here> from substituter <g_bor[m]>civodul: it also fails if run from a shell <g_bor[m]>here is the relevant part of the code, and the systm config is a plain-file with a really minimal os config. <raghav-gururajan>sneek: later tell sirgazil: I have sent patches #39068 and #39069 to fix gnome-contacts and gnome-calendar, respectively. :-) ***ng0_ is now known as ng0
<gnutec>upgrade GNU icecat to 68.4.1. Maybe the number issues is solve. <sneek>Welcome back gnutec, you have 1 message. <sneek>gnutec, nckx says: Thanks for that interesting link! It took me a while to find a ‘GPL’ font, and I ended up just refreshing the page until I found one (https://www.1001freefonts.com/montserrat.font, which is actually good enough to package). Do you know if there's a way to filter by licence? ***janneke_ is now known as janneke
<gnutec>nckx: I try but is complicate. But I only download GPL fonts that I found. [About filter in 1001freefonts] <gnutec>Nop! Still can't see numbers in GNU Icecat 68.4.1. I write to bug of gnu and gnuzilla. <mjw>I noticed that too. Sometimes a sentence on a page in icecat looks funny because all numbers are skipped. Any idea what causes that? <nckx>Not installing font-gnu-freefont-ttf (I think). <nckx>Otherwise, another font with a similar name. <nckx>If you already have it or installing it doesn't fix a thing, run fc-cache -r and make sure you've completely restarted IceCat. <gnutec>nckx: I'm install the font-gnu-freefont-ttf now and work fine. <gnutec>nckx: I lost this tutorial. rsrsrsrsrs <nckx>Understandable. This should be better documented or solved (in a way that doesn't involve propagation). <gnutec>nckx: That is why I don't like report a bug. I think that is will be unnecessary in the end. <nckx>gnutec: I assume your bug was the Portugese one. All right if I close it? <gnutec>I install font-gnu-unifont too. Just for fun. <nckx>Hehe. I think that's a bitmap font, though, which means it only looks good at a single font size. But it is a very complete font, and GRUB uses it to provide ridiculously good Unicode support for a bootloader. <leoprikler>I can understand lacking Unicode to some extent, given that most nonlatin scripts need special fonts (or at least noto) to be usable, but numbers? <bandali>or does it fail *with* fonts installed, too? <nckx>Yeah, it's a bizarre way to fail. <nckx>(Mozilla, that tagline's free.) <bandali>are there *any* fonts installed by default? <nckx>bandali: What's ‘default’? <mjw>gnutec, thanks! That was an easy fix. I got all my numbers now :) <nckx>I don't know if fonts in the system profile work. <leoprikler>If you use gnome-desktop-service, you'll at least have Cantarell, if nothing else. <bandali>sure. doesn’t matter though, since afaiu guix applications are by configured to not use system fonts on a foreign distro <mjw>BTW. Mine is a system install using the "gui" installer. I did select Gnome. ***Server sets mode: +cnt
<nckx>bandali: My point is that there's no ‘default’ user profile in Guix, it's empty. So while we could add it to %desktop-packages or whatever it's called (still just ‘a’, not ‘the’ default) it won't help non-System users. But I am drifting. <nckx>So it matters very much. <nckx>gnutec, mjw: Happy to help 🙂 I decided not to close the bug, but to merge it with 37345 and keep it open because it's not really ‘fixed’ if we have to answer this question every week. <leoprikler>could we use a post-install warning like the one with "source $GUIX_PROFILE/etc/profile" <leoprikler>Like "I see you've installed IceCat, but your profile contains no fonts. Unfortunately, IceCat does not like it if you have no fonts. Please install some to get numbers rendered." <nckx>We don't have a generic post-install warning system. <nckx>Also post-install messages are a mistake and a warning that your PM sucks. <leoprikler>And when fonts are installed/updated. "I see your fonts got changed. It would be a shame if an old font-config cache was still around and ruined your experience." <nckx>I just see two bugs that need to be fixed properly. <nckx>Not saying it's easy, mind you, at least the first one. <nckx>But papering over them with a ‘hey you, fix my broke’ message isn't nice and leads to things like Gentoo. <leoprikler>In that case, what about a wrapper for icecat, that appends freefont or some other unicode-complete font to some path variable that makes sense? <nckx>leoprikler: That would be my naive suggestion but I don't know if it makes sense. <nckx>Hence my coyly waiting around while ICECAT UPSTREAM is in da house and waiting for them to pipe up. <bandali>i don’t think we’d wanna push some specific font choice onto users <nckx>(substitute* (("Freefont") (string-append font "/share/fonts/Freefont"))) but then in a way that actually work with FC. <bandali>but for fonts, an automatic fc-cache -fv would be nice upon install/removal of fonts <nckx>bandali: I agree, hence my opposition to setting a precedent by propagating fonts/icons. <nckx>bandali: But there shouldn't *be* a font cache outside of the profile to begin with, just like we don't update some global infodir/mandb/…. I've never understood why fc is special but I do not know many things. <leoprikler>But how do we deal with multi-profile setups, then? <gnutec>I think Youtube is better after I installed the font-gnu-freefont-ttf to fix the missing numbers. Better then when I used GNU Icecat 67.x. <leoprikler>Does fontconfig have the capabilities to load multiple caches? (I assume yes) <nckx>Apparently font-dejavu (and hence, probably a whole bunch of other ones too) also fixes this. So IceCat just wants ‘a font’. *nckx pictures 3 people reading the fontconfig sources rn. <nckx>Only configurable through XML AFAICS, which won't work for (combined) Guix environments. So someone add FONTCONFIG_CACHE_PATH and we're golden. *nckx delegating am i doin it rite <jackhill>where is temporary stored when invoking guix environment --container --user ? <jackhill>nckx: ah, so I should be careful not to write too much stuff there :) <nckx>jackhill: I guess you could bind-mount /tmp in the container to [a subdirectory of] the initial working directory. (‘For containers, the default behavior is to share the current working directory with the isolated container and immediately change to that directory within the container.’) <nckx>Also, on many distros /tmp has been a tmpfs by default for years, so it's not a place you can safely write gigs anyway. <jackhill>yeah, I don't think it should be a problem, but after writing a buch of stuff, I started to wonder, and then couldn't find it <jackhill>I'd like to feel confortable answering this question by reading the source, but don't think I'm quite there yet <jackhill>oh, it's because the whole container / in the container is a tempfs except for what's explicitly mounted. I was confused becasue I didn't see the creation of the tmpfs. <nckx>jackhill: Have you tried --expose=/home/jackhill/foo=/tmp? <jackhill>I wonder if it's worth mentioned something in the manual (I'm thinking about something breif in the --user option) <jackhill>I had not, but I think what I want is actually --share=/big/scratch/space=/home/containeruser <nckx>Hokay. I did not get that from ‘where is temporary stored’. <nckx>Yes, that'll definitely work (or you've hit a real bug). <jackhill>I was probably being unclear, but you managed to answer my question anyway, so hurray! <nckx>Fill all the file systems! <jackhill>and I don't have to mount my scratch space on home, but some software just really likes to write there >.< <nckx>TBH it's less of a container and more of a skip now. <nckx>I wouldn't recommend it if security is at all a goal. <nckx>Folks, don't forget to update your patchelf'd Tor Browser. <jackhill>nckx: guix environment -C? No, I'm just wanted to try running maven without it leaving trash everywhere on my system. <jackhill>Or course it is downloading random code from the internet and compiling it, but uh, I have to make a compromise somewhere ***apteryx_ is now known as apteryx
<Formbi>I have xfe4-genmon-plugin, but I can't add it to the panel <Formbi>does someone know how to fix it? <nckx>No, hence ‘your patchelf'd’. <sneek>Welcome back sirgazil, you have 1 message. <sneek>sirgazil, raghav-gururajan says: I have sent patches #39068 and #39069 to fix gnome-contacts and gnome-calendar, respectively. :-) <oriansj>raghav-gururajan: anything can be put on guix if you are willing to do the work; The only things straight out of the box are the things that someone here feels is worth the effort... <vup>Gooberpatrol66: josm for example