<buenouanq>someone mentioned the other day and I think it's a great idea, having a bunch of example os configs <buenouanq>I'm trying to figure out how to write a service right now and that would be really helpful. <civodul>cbaines: there's a Special Hack for gnu/packages, see %package-module-path in (gnu packages) <buenouanq>docs are good, real life examples make them meaningful <quiliro>i think i have not read the whole manual...i will <civodul>cbaines: there's currently no syntax to specify that in GUIX_PACKAGE_PATH <adfeno>quiliro: If it was after an upgrade using `guix package`, then use double-dash counterparts of what I told for roll-back and switch-generations (note the S in the end now). <quiliro>please make a hack to install guixsd offline <quiliro>thank you. have to disconnect now. it is late and they are closing up <adfeno>quiliro:Take notes of what I just suggested :) <cbaines>civodul, so if I am reading that correctly, you don't have to use GUIX_PACKAGE_PATH, but just the GUILE_LOAD_PATH, and magic will happen... <adfeno>quiliro: Hm... I think that debbugs.gnu.org provides a bug search page. <civodul>cbaines: you do have to set GUIX_PACKAGE_PATH if you want to be able to refer to packages by name from the user interfaces <adfeno>Yep, debbugs.gnu.org does provide it. <adfeno>quiliro: I'm sorry but I really must go now. I 'll install the file manager from Guix, and I plan to return with the results tomorrow. <quiliro>bye people and thank you for such a great system <cbaines>quiliro, there are answers, but in the next period <quiliro>the first link did not arrive to my email <ng0>finsihed the 3rd run through the rust crates, sending them now <ng0>if we get this merged, I can start adding this item I talked about a year ago, the reverse-engineering <kyamashita>Is the GHC (Glasgow Haskell Compiler) functional for anyone else? <kyamashita>I should say "not functional." GHC gives an error about a missing ncurses library. <mange>I've got a program which I'm trying to package which uses confstr(_CS_PATH, ...), which on GuixSD gives "/bin:/usr/bin". Inside the build environment there isn't a /bin/sh, though, so the tests all fail (it works fine when installed). We have a bug about this (#23112), but it seems to have been closed as wontfix. Any thoughts on how I should proceed? (Re-open the bug? Skip the tests? Patch the source?) <roelj>Err, `guix environment: error: clone: 2080505873: Invalid argument'. Any ideas? (guix-daemon is running as root). <civodul>roelj: you're using 'guix environment --container'? <civodul>presumably your kernel does not support user namespaces, then <roelj>guix environment --container --pure guix --ad-hoc ... <roelj>Is that needed, even when guix-daemon runs as root? <civodul>in tests/guix-environment-container.sh, there's an example of how you can test that <civodul>roelj: 'guix environment' does not go through the daemon to create its container <roelj>I'll just enable user namespaces then <jmd>I was wondering if there is a way to use "guix environment --container" (or other command) to create a temporary enivonment where /bin, /usr/bin was populated the way that certain scripts like them. Is that feasible? <civodul>jmd: yes, it's been proposed a few times :-) <civodul>having a --fhs flag or something like it <roelj>This is really nice civodul: guile -c '((@@ (guix scripts environment) assert-container-features))' <jmd>civodul: Something like that, yes. <civodul>roelj: i just realized that 'guix environment --container' already calls this procedure anyway <roelj>More friendly error messages probably. It could've outputted "Error: User namespaces are disabled. Please see <link to the manual>.". <civodul>yes, that's what it normally does, with the above procedure <civodul>but in your case it thinks user namespaces are enabled <roelj>Oh, that's strange.. I'll try to debug it further then <jorgesumle>But in step 3 I get this error: qemu-system-x86_64: -net default: Invalid parameter 'default' ***retro is now known as Guest32852
***retro is now known as Guest22699
<ng0>Do we have a test I could run to see if curl is fixed now after the changes I applied, or should I try and rebase darcs and see if it works? <civodul>adding guile-bash to entertain myself a bit <ng0>the fact that new tests are failing for cURL after my change shows at least that something is changed <rekado>there are some files that seem only to be for mingw but cmake aborts the build because they don't exist <rekado>I started writing a blog post about my Haskell bootstrapping work <rekado>I'd like to mention bootstrappable.org at the end. <rekado>haven't yet taken the time to fix up all the text on that site <rekado>should do this soon and officially announce the project <ng0>the guix builder has no knowledge of /etc/, right <ng0>that's why this one new tests fails then <ng0>returns curl code 77 instead of 51 <ng0>which means no access to file or file not present <ng0>but in theory it looks good.. I would just skip the test <rekado>civodul: what's guile-bash? Where can I find it? <civodul>rekado: it allows you to extend Bash in Scheme <bavier`>inkscape is moving to a cmake-based build system <bavier`>it'd be nice to know what features all these projects are looking for, and see how the autotools could be made to fit <rekado>bavier`: I'm currently building the latest version of Inkscape. <bavier`>I know with cmake projects get "better cross-platform build support", but there's a lot of autotools features that then get left on the table <bavier`>inkscape: "cmake build system, which builds faster and is easier for developers to work with" <rekado>they have a dedicated page for CMake on the Inkscape wiki <rekado>but it's not very helpful in determining why they ditched autotools <bavier`>in my experience, cmake can produce slower makefiles. maybe they're referring to ninja support from cmake <rekado>I see that this could be useful. Inkscape is *still* building on my laptop... <bavier`>but is that autotools fault? or something else? <bavier`>e.g. missing parallelism in the makefiles, recursive make, c++ code, etc <civodul>bavier`: using cmake is about one of (1) supporting Microsoft's proprietary compiler/IDE, (2) making things easier for developers at the expense of users <bavier`>I'd need some examples of (2) though; in my experience things are just harder for everyone in cmake <rekado>I have significantly more problems with cmake than with autotools <rekado>build succeeded after 4875.2 seconds <rekado>no rule to make target 'test'. Stop. <roelj>rekado: lol. `guix build --continue-previous inkscape' would be nice here.. <civodul>do you have a slow computer or did they have fun with template meta-programming? :-) <rekado>I guess this counts as a slow computer <civodul>you should offload to your workplace's cluster ;-) <roelj>Especially in comparison to civodul's new beast ;) <roelj>civodul: GuixSD with EFI? :O that is awesome! <civodul>roelj: yes, though it was a bit involved <civodul>and i'm cheating by disabled fsck on the vfat partition <roelj>civodul: Oh, tell me.. I still haven't moved to GuixSD yet.. <rekado>I should offload to my workstation. 16 cores and 32G RAM. But I'm fresh out of Ethernet ports in my office... <roelj>rekado: Can't you then just start the build on that machine instead of offloading it..? :) <roelj>Especially in comparison to the tiny x200 :) <rekado>and I don't have a GPG subkey for it, so I can't sign commits <rekado>I thought Inkscape would build a little faster, actually <rekado>wanted to do a real quick update and push it <dvc`>civodul: about the cheating, I still have the btrfs patches. Have we decided yet what it should look like the fschk stuff? Or shall I make another proposal when I have some time? <sneek>dvc`, ng0 says: I thought when you wrote about the toolchain, that there was a reference I did not see. <sneek>dvc`, lfam says: Good morning! ;) <roelj>rekado: Your socat solution is working fine here too. :) <ZombieChicken>I just want to say now nice it is to be able to file bugs via email and not need Yet Another Account to do so. <civodul>dvc`: i like the proposal i made :-) <civodul>so i would suggest going in that direction <dvc`>I don't recall. I'll have to look again <dvc`>civodul: Ah yes, I started on that <dvc`>that was what I had in mind first too <dvc`>civodul: Did you see the email I cc'd you in? I hope I'm not overstepping with making such recommendations... <suitsmeveryfine>Hi! Is the ghc package broken or is it me who's forgotten to configure something? <dvc`>civodul: I don't have a filter I just say whatever I think =P <jmi2k_>Anyone knows why /dev/tty1 is owned by me, while the others are owned by root? I'm logged into all ttys. <jmi2k_>The problem I had with xorg is caused by this, because it disappears if I try to open it in another tty. <jmd>Is there a way to have "guix environment --container" start with a uid other than 0 ? <jmd>ok. Second question: Is there a practical way to change uid after it's started? <bavier`>always be learnin' the standard unix stuff <jmd>But the problem is, guix environment --container gives me just a single user in /etc/passwd <davexunit>so you can create additional users if need be <jmi2k_>alezost: I finally solved the problem :D I removed the suid from X, used the command line parameters you said and added myself to the input group. It was something silly, but I come from a systemd distro where I didn't have to do anything of this. Really, thanks for helping me yesterday! <civodul>davexunit: we could add an option to map to the UID of 'nobody' instead of UID 0 maybe <davexunit>is nobody guaranteed to be a user on the host/ <davexunit>I'd like to create a UID 1000 with the user's current username, but then the same uid on the host would be mapped to 2 uids in the container and AFAIK that isn't possible currentlyh. <civodul>davexunit: it's not guaranteed to exist, but it's likely to; we could error out when it's not, or use UID 0 <paroneayea><civodul> rekado: it allows you to extend Bash in Scheme <civodul>dunno what to think of this Rust trend <civodul>it's great that everyone takes memory safety seriously <adfeno>Do you know, by chance, why our recipe for GNUnet is still refering to an older version of it? <civodul><adfeno> Do you know, by chance, why our recipe for GNUnet is still refering to an older version of it? <ng0>Was it Mark Weaver who is responsible for most of the icecat/mozilla packaging? <ng0>I have a question posted where I need someone with experience in mozilla packaging... Subject: Mozilla help needed - 0ad, or: how to pass the version to mozjs-38? <ng0>adfeno: there's a short answer, and there's a long answer. The long answer will follow in some months or weeks probably, I'm discussing this with some people at the moment. Short answer: there are real issues which must be fixed first. I can post something about it this or next month <adfeno>Are the subject of these discussions too complex or private? <adfeno>Can you give me a summary (for a non-developer)? <ng0>I work on many things at the same time, and so do other people, which ends up in lagged conversations. <ng0>so far I don't want to give a summary, I need to get some answers first. I don't want to share things I only have half an answer to <ng0>well I can look into the thread and give a very rough summary <adfeno>... I asked that first question regarding the version because I'm having a bug with that version. <adfeno>I made a bug report of it in the GNUnet bug tracking system, and I'm also currently discussing the bug with grothoff-mobile at #gnunet <adfeno>And he just pointed out that my version is about two years old. Don't worry, I'm not here to make fun of it or criticize/be angry, I'm just trying to understand the possible reason for the delay so I can try to help <ng0>Okay, I think I should make a post about it later this month <adfeno>ng0: Out of curiosity, where would you make such post? <ng0>either my blog or post in post to the mailing list <ng0>but I think the mailing list is what it should be, as the question arose before here <ng0>but Whonix is also interested in gnunet (gnunet-fs to be specific) now, and at the congress the gathering of the eV had some promising outcomes/reports (regarding work with pretty easy privacy /pEp foundation, etc)