IRC channel logs
2025-01-03.log
back to list of logs
<meatus>How do I get the mailing list working correctly? I sent patches, yet I was never forwarded people's replies to them. Now I want to reply to them but the mailto: link in the web interface doesn't work <dajole>I'm reading the manual, 22.1 Building from Git. It says "How do you ensure that you obtained a genuine copy of the repository? To do that, run guix git authenticate, passing it the commit and OpenPGP fingerprint of the channel introduction (see Invoking guix git authenticate)". However, neither from that section nor the one on channel introduction can I tell how to get the correct fingerprint. <dajole>Since the default channel is not one I set explicitly in my channels.scm <dajole>Any pointers would be greatly appreciated :) <dajole>The channel section (7.4) does say: "For the main channel, called guix, you automatically get that information from your Guix installation." <vagrantc>dajole: you would need to establish a trusted communication channel with the authors of the channel as to waht the right fingerprints and commits should be used <dajole>Well in 22.1 I'm just trying to verify guix itself <dajole>I'm guessing that for the author of that section it's obvious what needs to be done to get the fingerprint, but as someone who is still learning it's a bit confusing :) <vagrantc>bootstrapping a trust path properly is definitely a confusing subject! <vagrantc>many very experienced people struggle with getting it right ... for years guix had no real such trust path <vagrantc>i think the blog post in which guix authenticate was introduced might mention the fingerprint authenticated against .... <dajole>I think I may have figured it out. `guix describe -f channels` gives the fingerprint for the guix channel :) <dajole>Thank you for those links, I'll read up on them too! <dajole>Yup, both of those articles look helpful in providing more context. Thanks again for your help. <vagrantc>gotta head out now, but others might have some good suggestions <j8>Hi, I installed guix on trisquel 11 using ~sudo apt install guix~. When I run guix pull, i’m getting the error: guix pull: error: Git error: the SSL certificate is invalid: 0x01 - The certificate validity has expired <j8>I also tried guix install nss-certs; export SSL_CERT_DIR="$HOME/.guix-profile/etc/ssl/certs"; export SSL_CERT_FILE="$HOME/.guix-profile/etc/ssl/certs/ca-certificates.crt"; export GIT_SSL_CAINFO="$SSL_CERT_FILE"; <j8>guix --version: guix (GNU Guix) 1.3.0. any idea why it this is happening? <j8>Currently I'm not sure if this error is related to guix or trisquel, please ignore my previous text. <trixthethird>Hello everyone. I am new to guix, trying to migrate my NixOS setup here. I am hoping I can get some questions answered here as I go trough my journey. My first one being - when I inherit from a parent operating-system, and I want to add services in my deriving config, what's the best way to do it? This is what chatGPT came up with: <trixthethird>Also in general how would you recommend extending the installation-os? I am hitting some errors when trying to build my own image that inherits from installation-os. <homo>trixthethird paste.debian.net <homo>yes, and it doesn't require js, I use it via netsurf with disabled js <homo>alright, now wait for someone more skilled than me to help, since I am new here and have only 1 laptop :) <divya>Rutherther: I got the source from git, and I even checked if the source that guix fetches locally has the tests .toml files, but the build can't find it nontheless <sepeth>trixthethird: Not sure if it is something else but did you mean #f instead of #false? I think you might meant the former. <trixthethird>Yeah I did, though it is unrelated. The issue I am thinking now has something to do with channels. I installed guix via nix, and just now ran guix pull. I am hoping this fixes it. The stack trace indicated the whole thing failed when looking for some dir. <rekado>sepeth: #false is a valid alternative to #f. (I use it all the time.) <trixthethird>Ok after running guix pull and using the binary it produced, I am now able to build the image. <jlicht>isn't the packaged version of c-ares kind of old? What branch would an update to this package go to, per the latest-and-greatest branching policy? <efraim>jlicht: probably a feature branch of just itself and then whatever breaks <jlicht>I'll see if I can break some things, fix most of them and then push a branch + ML message :-) <jlicht>-P is the short form of build-dependents? If so, I was going to already, but thanks for the reminder :-) <yarl>I am reading "(guix)Packages with Multiple Outputs", in the penultimate paragraph, different dependency footprints are is mentioned, with WordNet as an example. It is said guix graph can show it. How? <Rutherther>divya: yeah, can happen, probably the build expects something else, and to figure that out you will have to read through the source code. (or just disable the troublesome tests if you don't want to bother and are fine with not having tests) <muradm>could some one look at 75270 and 75311, thanks in advance <sneek>orahcio, you have 1 message! <orahcio>homo: I could not see your screenshots, sorry. Good year for you! <homo>orahcio thanks and sorry, those screenshots were made many days ago and that site doesn't store them for long <homo>btw, I am on sway now, and I use it as gnome - mouse and touchpad instead of keyboard <orahcio>I use sway with my digital tablet instead the mouse, but I like to use keyboard to switch between workspaces <homo>3 finger swipe with touchpad is a good default <homo>unless you have programs specifically designed to react to 3 and more fingers <orahcio>I am in the fight to delete the gdm yet, and make autologin without broke the home packages <orahcio>homo: yes, I agree with 3 finger swipe <homo>orahcio it's strange to have broken home packages because of autologin, sounds to me like something doesn't invoke bash as login shell <orahcio>homo: I think the sway starts without my profile, the packages are therem but not in the PATH, if I login with another tty, the home packages are there <homo>orahcio there is big differences between "sway" and "bash --login -c sway" <homo>s/differences/difference/ <Rutherther>orahcio: so it's no wonder you don't get path and other env vars. If you are using a shell that is not posix compliant, you will have to make your own profile scripts that will source the profiles, as /etc/profile is not used <homo>the goal of fish is to be interactive, not interpreter of bash scripts, and your profile is a bash script <homo>if from other tty you get your profile, it is because it was like "bash --login -c fish" <homo>also on gnu system I wouldn't expect any shell other than bash to work as login shell <orahcio>Rutherther: thanks, but just the tty1 with sway autostart has this problem, another tty my profile runs <orahcio>homo: yes, maybe I need to use bash as login shell and configure fish to use just with my terminal emulator <homo>orahcio that is what other tty basically does, it invokes bash only for login, and then it automatically gives you fish <homo>orahcio that wiki that you linked assumes that you have written your login script for fish yourself, so if you really want fish as your login shell, you need to teach "guix home reconfigure" to generate fish scripts and symlink them to your $HOME <yarl>I am reading "(guix)Packages with Multiple Outputs", in the penultimate paragraph, different dependency footprints are is mentioned, with WordNet as an example. It is said guix graph can show it. How? <homo>omar_b the answer is if it doesn't get attention, it means this community of volunteers is small, and volunteers with commit access 1. have their private lives 2. are focused on other parts of guix 3. don't owe us anything <Rutherther>yarl: I am not sure what the best/intended way is, but one would be to use --type=references on the store path you are interested in <Rutherther>also I doubt there is better way since to know dependency footprint you first need to build it to actually know what inputs are used <yarl>Rutherther: Does that mean that, let's take a concrete example with qemu-minimal, even if I don't need "doc", python-sphinx will be built? <yarl>Well, that's kind of bad. <Rutherther>yarl: there is no other way. There are just one "inputs". All these are used. Then after the build is done daemon scans what paths are actually used and others can become inactive (if nothing else holds them) <orahcio>homo: After this autologin adventure I will write the README for my guix-config, I hope it will be helpful for somebody <yarl>Rutherther: it's not possible to make the input field depends on the output demanded? <homo>orahcio it will be easier to notice if it gets into official guix documentation <Rutherther>yarl: no, you would have to make different package <yarl>Rutherther: Well, I am quite surpised. <Rutherther>yarl: note that due to how outputs work, you cannot build just one output, you always build all of them <yarl>Multiple outputs is not a such useful feature then, I think. <runejuhl>I'd like to try out guix from the go-team git branch (re: bug #69617) -- what's the easiest way to do that? Can I run `guix shell` and have it install guix using an arbitrary git branch? <Rutherther>yarl: it is, it greatly improves the space necessary if you don't want particular outputs. What is utilized during build is also more, but it gets removed afterwards. It's the same with outputs. If you don't need them, you remove them -- or if you are just installing and the path can be substituted, then you of course can substitute only some of the paths <runejuhl>Ah, right, I remember reading about that :) I'll see if I can make it work -- thanks! <yarl>Rutherther: I see but it feels only half-way has been done. <homo>it is just too much work to create separate package to build only docs <Rutherther>yarl: I don't think so, having individual buildable outputs would mean every package that has them would be a lot more complex to write, and would take a lot more time to build, as you might have to essentially build the same thing multiple times and just take what you want <homo>v4l-utils is the worst offender here, you can't easily build only libv4l package and then make another v4l-utils package that links to libv4l package <Rutherther>and if you wanted to merge these (building multiple outputs at once) together afterwards, it would be even more complex <homo>basically blame build systems of those projects <yarl>Rutherther: I don't get it. Why would it be a lot more complex to write? <homo>this is my failed attempt <homo>because v4l-utils doesn't install <homo>this example of my failure is good enough evidence why it is more complex to write <homo>I will be happy if you properly split libv4l and v4l-utils, because it is too much work for me to fix its build system <homo>it is more complex to write because existing projects make it more complex to write <fnat>Anyone using "guix shell" in eshell? I seem to get conflicts in the way the prompt is handled by eshell versus Guix's shell. <homo>fnat guix shell your_package -- bash -i <fnat>It might be sensible to just use "guix shell" in a terminal, but I thought of asking in case the glitches I see were due to my config. <homo>this is how I make guix shell working in plan9 userspace <homo>so might work for emacs as well <fnat>homo: Hm, what does that do? I see '-i' is for 'interactive'? Not sure what that means here though and how it differs from the default behaviour. <yarl>homo, Rutherther: I think I see your point. I was only thinking about my particular example. It's usually easy for the docs, no? some --withorwithout-docs or no detection of dependencies (sphinx). <homo>fnat emacs is not a terminal emulator, so when you run bash outside of terminal emulator, it doesn't know that you interact with it, this is why you need to pass -i <homo>gnu's not unix, so programs are not designed to work together :P <fnat>homo: I see, but it doesn't seem to solve my particular problem (probably due to how I construct the Bash prompt vs my eshell prompt). <fnat>So, to answer my question, I *think* this might have to do with my configs (Bash vs eshell prompt). <fnat>My Bash prompt is clever and it shows I'm in a Guix Shell. But then eshell doesn't know how to parse that. <fnat>It's probably wise to add 'guix shell' to the list of commands that should be handled by a terminal. <david95>Hello everyone and Happy New Year. I can't overcome the problem with locales, what haven't I done )) Here are my settings: <yarl>Rutherther, homo: what I mean is that (I think) {program, lib, gui, whatever} and {docs} are 'orthogonal', you don't need any input from the first set to build the second and vice-versa. I am not sure if it is worth to ALWAYS separated those. Bu as I am fighting with cross compilation these days I am wondering how can things be the most reduced. <runejuhl>homo: seems to work perfectly, ended up with this invocation: `guix time-machine --commit=288abf3fbf -- shell --network --container --nesting bash coreutils findutils grep nss-certs -- bash -i` <yarl>But thank you anyway Rutherther and homo I understood more how guix works. <runejuhl>david95: which locale do you use? I seem to remember that I saw that message a whole lot due to having `LC_TIME=en_DK.UTF-8`, or something in that vein <yarl>david95: In your settings paste, "echo $GUIX_LOCPATH" is wrong <david95>Hi, thanks for the answer. I'll show you now. <yarl>david95: In your settings paste, "echo $GUIX_LOCPATH" is wrong <yarl>Try manually exporting then trying again guix search <runejuhl>yeah, C locale shouldn't be a problem, so it's your GUIX_PROFILE environment variable <david95>I don't know what to do)) write the correct command what I need to do, I will do it. I have already tried many different options, so I wrote here <runejuhl>you can try doing `source ~/.bash_profile` first, that should set your GUIX_LOCPATH to the correct value, if HOME is set. From the echo statement in your paste it looks like `$HOME` isn't expanded correctly, so GUIX_LOCPATH simply becomes `/.guix-profile/lib/locale` instead of e.g. `/home/username/.guix-profile/lib/locale` <Avron>Hello all, I just installed guix in a vm controlled with virt-manager, using the guix net installer with config.scm as made by the installer unmodified (GNOME as desktop environment), which I did several times before, but after the guix pull, sudo guix system reconfigure /etc/config.scm eventually fails with "guix system: error: kexec-load-file: Invalid arguments". If I restart it, the outcome is the same. This looks like bug#75211 reported on Monday, except <Avron>that it may affect all new Guix installations made with the installer, which looks highly problematic. <Lumie>david95: hold on, do you use home config? <Lumie>You can declare all environment variables there <Lumie>Without "polluting" your profile <david95>I installed the guix package manager <david95>please write what is the installation algorithm, I will do everything from the beginning <homo>I was sure it is pretty much documented how to setup profile on foreign system, as I somehow used guix on foreign system before... <Lumie>But I've only done it one on a foreign distro <homo>guix on foreign distro is like gnu on windows <Rutherther>yarl: if you meant that specifically docs should be separate, then yeah, and easiest would be to just make it different package. I thought you were arguing that all outputs should be separated <david95>maybe i found the reason. i used fish , now i disabled fish and entered again the command echo $GUIX_LOCPATH /home/david/.guix-profile/lib/locale Is it correct ? <Avron>homo: I don't use windows so I don't know, I use guix on trisquel and it works at least with the default trisquel environment, but setting environment depends on the distro and on the DE, and different things may be needed. On trisquel, installing the guix package puts a file in /etc/profile.d which does the job. <yarl>Rutherther: I guess that's what I mean now that I understand more how this works. At the beginning I didn't knew how and why the multiple outputs. <yarl>Rutherther: The docs are misleading, the example taken is glib with the glib-with-documentation hack. <yarl>I kind of get it but I'm not confident enough to improve it. <david95>great, thank you very much. Now I need to adjust another parameter a little, because I used Fish and got confused because of it... when I enter the command: $ echo $GUIX_PROFILE I get this output: /home/david/.guix-profile it seems to be correct, what do I need to do next so that I don't see the message about locales? <david95>'$ guix install glibc-locales ' '$ export GUIX_LOCPATH=$HOME/.guix-profile/lib/locale' <runejuhl>david95: installing `glibc-locales` should be all you need, but what's the issue you see with the docs? <janneke>after 2h of trying to set cflags for ghc-7.10.3 i'm giving up and build it with gcc-13 <david95>yes you are right)) to solve one issue with one command you need to spend hours on it and the problem is that following the wiki what should work does NOT work. I downloaded glibc-locales but still get a message that I need to download. <david95>thank you very much for the link Fish but now I would like to figure out Bash)) so that everything works correctly in Bash, then I will do it myself Fish <david96>how can i remove the message about locales, i did everything as in the wiki, it doesn't work <david96>and for example, this professional blogger recommends running the same command as root <david96>I watched his video and even this pro every time he launched commands he was very worried whether it would work or not)) what can I say about ordinary users... <graywolf>I have stupid question. I have a script that basically boils do (program-file "foo" #~(system* #$curl "https://example.com")). I run it from a shepherd service, and it cannot find the certificates. <ieure>graywolf, Is the nss-certs package available in whatever context that's running in? That's what provides the certs. <ieure>The SSL_CERT_DIR environment variable should point to /etc/ssl/certs in the nss-certs package. <graywolf>Would that be set for a shepherd service? <graywolf>I guess I can just dump (environ), will try <Rutherther>graywolf: it is not present in shepherd system process. It could be in user shepherd <graywolf>Soo... Should I just hard-code it to SSL_CERT_FILE=/etc/ssl/certs/ca-certificates.crt? <graywolf>Or is there some way to compute the value? <ieure>graywolf, Honestly hardcoding it probably isn't the worst idea. <ieure>It'll break on any system without nss-certs, but that's in %base-packages now. <Rutherther>graywolf: yes, there is, you could build a profile and source its etc/profile file using a shell <graywolf>Hm, I guess that would work, but I somehow find that worse then hard coding it. I do not really need shell at all here, so using it just for this seems overkill <graywolf>Will go with the hard-coding. I guess there are good reasons why curl does not have the path hardcoded, but it is a complication :/ <Rutherther>graywolf: I don't know if it's worse than hardcoding. But hardcoding /etc/ssl/certs definitely sounds worse rather than using a gexp with nss-certs package itself <janneke>anyone know how to set cflags for perls amazing makemaker? <janneke>(CFLAGS=..) works sort of, but creates a broken Makefile <graywolf>Rutherther: the ca-certificates.crt seems to come from ca-certificate-bundle package, not nss-certs. <janneke>okay CC=.. works, aparently the CFLAGS concept is too hard for perlies <graywolf>Uff, so I guess I can convert nss-certs into a single package manifest, part it to (@ (guix profiles) ca-certificate-bundle) and use the result in gexp to get the path. <graywolf>But ca-certificate-bundle returns a derivation, not something I can directly use in file-append or in #$. Welp, hard-coding /etc/... it is. <graywolf>Rutherther: I am still on board with the not-hard-coding it, but no idea how to actually write it :/ <ieure>If I have Cuirass on an amd64-linux system, and add aarch64-linux to `systems', does it automatically run stuff in QEMU? Or do I need to do more setup for that? <cluelessguixer>When a service fails to start, how do I get the crash message so I can investigate? <freakingpenguin>cluelessguixer: /var/log/messages contains system shepherd log messages <freakingpenguin>Anyone know why this occurs? PATH contains /home/richard/.local/bin at the start. <ieure>freakingpenguin, Are you executing them from a shell? <ieure>freakingpenguin, Are you sure $PATH has that directory? If you `echo $PATH' in the shell you're running the commands from, do you see it? <ieure>freakingpenguin, Have you run `hash -r' or logged out/back in since adding this? <ieure>Execute bits set on the files in .local/bin? <ieure>freakingpenguin, Do the shebangs in the scripts point to an interpreter bash can locate? <freakingpenguin>Yep. The only thing that's slightly concerning is root:root ownership, but o+x is set and I can execute them correctly so I don't think that matters. <graywolf>freakingpenguin: What error do you get? Just `command not found'? <ieure>freakingpenguin, Yeah, not an issue, anything in the store has root:root ownership. <freakingpenguin>graywolf: Yes, command not found. And the shebang is /bin/sh, which does exist. <ieure>freakingpenguin, Does `which' find your scripts? `which foo' should print `/home/richard/local/.bin/foo' (for some value of foo that exists in that dir) <graywolf>Does guile -c '(pk (search-path (string-split (getenv "PATH") #\:) "some-script-name"))' find the script (after you adjust the name)? <Rutherther>freakingpenguin: can you send output of "ls -al ~/.local/bin" and "printenv PATH"? <graywolf>Interesting. Maybe a typo somewhere. Since you verified $PATH contains the directory, if you copy paste together the directory from echo $PATH and /name-of-the-script, will it work? <freakingpenguin>Seems to be the same for .guix-home/profile though so that may be expected. <Rutherther>yeah, I wanted the one with /, but forgot it's a symlink <Rutherther>interesting, really no typo or anything as can be observed from the outputs you sent <freakingpenguin>graywolf: Yes, $ /home/richard/.local/bin/git-change-id executes the script. <graywolf>You could try to strace bash -c git-change-id to see what paths it tries to access? <ieure>freakingpenguin, What if you change $PATH to have the trailing slash? ex /home/richard/.local/bin/:$PATH <ieure>The different `ls' output for trailing-slash vs. not is normal for symlinks. <freakingpenguin>It's trying to access a file a "PATH=/home/richard/.local/bin/git-change-id", i.e. my PATH variable contains PATH=. <ieure>Many "interesting" problems boil down to something dumb like this, in my experience. <freakingpenguin>Rutherther ieure graywolf: Thanks for the help! :) As you can tell I have a very special set of skills. <weary-traveler>ieure: every time someone who's active on IRC gets a commit bit, i'm elated. especially if they have a passing interest in emacs ;) <ieure>weary-traveler, I might have more than a passing interest, heh. <cluelessguixer>The java version of i2p seems to fail installation when it tries to chmod something, but I have no idea how to fix it. Any ideas? <ieure>cluelessguixer, Can you paste the build log on a pastebin? <cluelessguixer>ieure: I mean, it's more me doing "java -jar i2pinstall.jar". When it fails it says "Error executing /bin/chmod a+x /home/guix/.i2p/postinstall.sh". Here's the crash log: https://bpa.st/Q6OA <ieure>cluelessguixer, Do you have the package definition somewhere? <ieure>cluelessguixer, Is this happening during the package build? Or does the package build, and this is failing when you run its output? <cluelessguixer>For the java version of i2p I just get the jar from the website. There's only i2pd when I use guix search. <ieure>cluelessguixer, Okay, so this isn't a Guix package at all? <ieure>Hummm, this Guix System VM I have running is in kind of a wack state. It's not mounting /boot/efi, so `guix system reconfigure' fails to install the bootloader. If I try to mount it manually, I get `mount: /boot/efi: unknown filesystem type 'vfat'.' <cluelessguixer>ieure: Yeah, not a guix package. I just tried building from source and got an error "Unable to find a javac compiler". I've got openjdk installed. <ieure>cluelessguixer, It probably needs JAVA_HOME set. <ieure>Yeah, something is very wack about this VM. lsmod / insmod / rmmod aren't in the profile, neither hostname, find, a bunch of other things I'd expect in a base install. <cluelessguixer>ieure: It's set to /gnu/store/...--openjdk-21.0.2. I'm guessing that's what it gets set to after doing "guix install openjdk", but maybe it's wrong? <ieure>cluelessguixer, Is javac on your $PATH? <Rutherther>cluelessguixer: javac is in openjdk:jdk output. Do you have that output? <lfam>ieure: it could be that the VM machinery assumes a traditional BIOS rather than EFI <lfam>If that's the case, it would be nice to switch it over since EFI seems to have taken over <ieure>lfam, QEMU uses EFI, my image declares a vfat /boot/efi partition, efi bootloader, and successfully boots; I don't think it's that. <lfam>Right, but *something* is wrong <lfam>Like, the image creation tool doesn't set up the initrd correctly or something like that <lfam>I say this because EFI does work on bare metal <ieure>lfam, I'd expect the VM not to boot if that were the case. I can boot, log in, do Guixy things. Just the operating-system inside is horked in multiple ways. <cluelessguixer>Rutherther: Not sure where I find that. It doesn't appear like I have "javac" command. <ieure>Like not supporting vfat filesystems. <lfam>ieure: Right, something about the OS declaration only works on boot, but it isn't written correctly by the VM image creation tools to support reconfiguring <Rutherther>cluelessguixer: If you don't have that command, then you don't have it <ieure>cluelessguixer, Like Rutherther said, it's the openjdk:jdk output you need. <luca>Hi, does anyone have any examples of how to run (or package?) a python script with some given dependencies? It's not a full on pip package, but I guess if I have to I can make it into one <rekado>luca: the Guix manual has an example of using "guix shell" in the shebang line to get the dependencies of a Python script <cluelessguixer>ieue: Oh, I just realized "openjdk:jdk" added to guix install works. I'm afraid my name is not just for show, lol. <ieure>cluelessguixer, IMO it's pretty unusual for something to dump a shell script in your $HOME and execute it when it runs. Strongly recommend that you audit the script contents before executing it. <ieure>cluelessguixer, This is 99% paranoia, but if you're using I2P, maybe you share that paranoia. <cluelessguixer>ieure: I'm playing around in my Guix VM, so I should probably be fine. It's also meant to be containerized once I figure out how to install it. <cluelessguixer>Got it to build a litte after fixing gettext dependency, but this error isn't simple to decipher. "build.xml:643: java.lang.ExceptionInInitializerError". Further down I see "Unable to make field protected volatile java.util.Properties java.util.Properties.defaults accessible: module java.base does not "opens java.util" to unnamed module". <ieure>cluelessguixer, Most likely a Java major version mismatch. <cluelessguixer>"See the file "build.properties" if this step fails." was above the build failed part. A bunch of build properties in there that "end users" shouldn't touch. Guess I got my work cut out for me. <cluelessguixer>I suspect you were right about the java major version mismatch, ieure. It seems Java 9 or higher should be fine? <ieure>cluelessguixer, 9 is pretty old, that's actually the oldest version packaged in Guix. <ieure>21 is common for new projects. <cluelessguixer>"Strictly speaking, you must either use the compiler for the minimum Java version you are targeting (default 1.8, see above), or specify a bootclasspath, which means you need the JRE for the target version installed as well." <cluelessguixer>With 21 I interpret it as I should be good, so maybe something else is stopping the build. <ieure>cluelessguixer, Java versions are incredible, the "Java 8" marketing version is 1.8 in reality. <ieure>cluelessguixer, Emacs is the same way, "Emacs 29" is 1.29. <cluelessguixer>Looks like I got a little further by uncommenting "NoEXE"... "build.xml:1903: java.lang.NoClassDefFoundError: java/util/jar/Pack200". Below I see "izpack" mentioned. Shouldn't be an outside dependency... <cluelessguixer>Yeah, it's izpack that's the issue. Don't see an easy solution for this one. <Hamled>it's not possible to use wrap-program to create a wrapper for an executable in a different package right <lilyp>it's quite a weird use case, but you can copy-file that executable <Hamled>hmmm, I'll give that a shot. I'm trying to write a simple wrapper around icecat without changing the actual icecat package and thus have to rebuild it <meatus>dear lord guile's error reporting is bad <cluelessguixer>What the... I installed the lowest version of openjdk I found, but forgot to try building again. Was about to clean up shop and... buil successful. <cluelessguixer>I just discovered the whole building thing was unnecessary. You can install with the dowloaded jar from the official i2p website, have it fail, do the chmod command yoruself, then run postinstall.sh. Finally the router is run with runplain.sh. 
