***dftxbs3e__ is now known as dftxbs3e
<alextee[m]>how do you apply a patch after editing it so that it says co-authored? <alextee[m]>^ i hacked it together with git am and then git commit --amend and added the line manually ***catonano_ is now known as catonano
***nly is now known as nly-on-arm
***nly-on-arm is now known as xXx-nly-GoDsLaYe
***xXx-nly-GoDsLaYe is now known as nly-zweihander
***nly-zweihander is now known as vits
***vits is now known as vits-ushanka
<efraim>I thought that was the 'official' way of doing it <ZombieChicken>Hello. Has drive setup loosened up some? Is encrypted root possible atm? ***vits-ushanka is now known as nly
<efraim>xelxebar: my board locked up overnight but then recovered, it's now ~12 hours into building guile-static and should finish eventually. I think it's the last big thing to build before the image is ready <xelxebar>efraim: A million internet thanks, good sir. Any idea what caused the lock up? That's unsettling. <efraim>xelxebar: I got one of the really early pine64 boards, could be overheating, could be I'm using zram instead of regular swap, could be a hardware issue. It's definately the board and not the software though <str1ngs>I fixed jack2 build for aarch64 BTW was blocking quite a few things for my rockpro64 <xelxebar>efraim: Oh, cool. I keep my eye on the Pine64. Makes sense. ***apteryx is now known as Guest15218
***apteryx_ is now known as apteryx
<rekado>str1ngs: it just occurred to me that it’s unusual to have jack-2 in inputs at all. <rekado>we decided to build everything just with jack-1, because that’s sufficient, even when at runtime the user uses jack-2 <rekado>while it’s good to have a fix here, I think we should use jack-1 anyway <rekado>just tried building gst-plugins-good with jack-1, but the configure script rejects it. Too bad. <str1ngs>efraim: huh, strangely I can't get the jack2 for loop initialize error to replicate. I removed that phase and in a second in series. which is fine by me. <str1ngs>rekado: no problem. the jack2 would have cause quite a few rebuilds if jack1 is easy that cool. this atleast bumps and allows jack2 two to build on aarch64 <brendyyn>guix gc shows 150GiB freed when df -h shows only 85 GiB extra free space <str1ngs>rekado: I can look into gst-plugins-good its the blocking nomad and guix system vm for me. <efraim>brendyyn: some of that space was saved due to deduplication. <efraim>ie, 65 GiB weren't used before due to deduplication, and those same 65 GiB of duplicate data were removed with 'guix gc' <brendyyn>efraim: but it telling me fibs :(. the amount freed should be the real amount, not the amount+deduplication *civodul reconfigured berlin with the latest guix-daemon <jlicht>is there any way in which I can convince shepherd to give back the controlling terminal after a fork+exec? <civodul>janneke: well, we'll see! nothing broken so far :-) <civodul>jlicht: give back controlling terminal to the parent process? <civodul>janneke: the hurd-vm service still fails to start <jlicht>to the forked one! Trying to get a (non-guix) shepherd service for root-less Xorg going, and the easiest way is to have it run for the current VT and passing a -keeptty flag. It seems that it somehow has now 'tty' to '-keeptty' after the fork+exec call in shepherd though :/ <civodul>rekado: there's also a mumi-worker service that's stopped, but perhaps it's running elsewhere? <janneke>civodul: arg...but great that it also works for you <civodul>jlicht: maybe you'll need your own variant of fork+exec <janneke>civodul: i think someone should try to attach a vnc console, or run the childhurd qemu spell by hand <civodul>janneke: so qemu is started, but then the 'start' method gets ECONNREFUSED repeatedly, until it gives up <civodul>netstat says qemu is listening on localhost:5900, 0.0.0.0:11022, and 0.0.0.0:16900 <efraim>is there documentation somewhere about the correct path for gdb files? like that it should be %out/lib/auto-load/%out/lib or something? <civodul>see $(guix build guile)/lib/libguile*gdb* <efraim>efl has one at %out/lib/auto-load/usr/lib/*gdb, other meson ones have %out/lib/auto-load/%out/lib/*gdb <janneke>civodul: hmm, what is 0.0.0.0:16900? ... and we need 127.0.0.1:11004 for secret service, right? *janneke goes to read berlin configuration again <efraim>i guess %out/lib would be the easiest <civodul>guix offload: error: program `/gnu/store/lx7p76mgkq08ikr5gzl8wyra27lg90xc-guix-1.1.0-18.218a67d/bin/guix' failed with exit code 1 <civodul>where 1.1.0-18 must be the version running on the build machines <sneek>Welcome back nckx, you have 3 messages! <sneek>nckx, raghavgururajan says: The bot has been added, under the nick milkman[bot] <sneek>nckx, raghavgururajan says: Could temporarily add a line on ChanServ message about this new bot and to use `help` command to see options? Thanks! <nckx>raghavgururajan: I'm reconfiguring bayfront, but waiting for the big GC lock again. <nckx>ZombieChicken: Encrypted root has been supported forever (with the caveat that you get to enter your passphrase twice). What's not is encrypted /boot. <xelxebar>Can't you put a keyfile in the initrd to bypass the second passphrase check? <xelxebar>Anyone running a guix system, mind checking the perms on /proc/cpuinfo for me? <xelxebar>Probably 444, but just want to make sure. <rekado>civodul: I guess the worker just crashed <rekado>(it’s not entirely unexpected, unfortunately, because Guile Xapian has an undiagnosed crash) <rekado>I’ll stop mumi-mailer, though, and start it in a screen session <rekado>that one still needs fixing; it only works with some more env vars that are set in an interactive session <xelxebar>copy-recursively from (guix build utils) is like `cp -r foo/ bar/', but how do I accomplish the equivalent of `cp -r foo/* bar/'? <nckx>Something like (for-each copy-recursively (find-files "foo" ".") "bar") would do the trick. <nckx>Although there may be some gotcha with relative paths necessitating a (with-directory-excursion "foo" ...) instead. <xelxebar>for-each. That's what I was looking for. Chers <civodul>it's a lot of fun: i leave my laptop to have lunch, and when i get back, emacs has been OOM-killed <civodul>it's already at 141M resident, and counting <civodul>i thought it could be emacs-smart-mode-line or similar, which would explain the rapid leak <janneke>civodul: i have not used emacs as heavily as during hurd/bootstrap stints, but see nothing ooto *janneke has a pretty modest setup, though <civodul>janneke: yeah i think i have a modest setup too <janneke>civodul: except for EMMS, i use those too <nckx>Bonjour raghavgururajan. <nckx>Could you try logging in with your new key? <nckx>The reconfiguration has completed, at least. <janneke>civodul: i also have guix, paredit, exwm, debbugs, bbdb, desktop-environment <efraim>anyone have an idea how to setup something so that if a system is built with 'guix system container' it uses one option and for 'guix system docker-image' a different option? <civodul>efraim: i don't think you can distinguish between the two right now <efraim>that's what I figured. I guess I can just have one inherit the other and override the one difference *janneke is using /gnu/store/0rdyyb669wjnswv9awzbdw1qfz7nf9si-emacs-27.1/bin/emacs *efraim is using /gnu/store/036m4di972raga3p8nkp3id6mpwl1kkd-vim-8.2.1500/bin/vim *efraim doesn't have any memory leaks here! ;P <efraim>native tensorflow package causing problems ... <rekado>it was my impression that it’s not really useful for anything serious <efraim>tensorflow from guix or tensorflow with compiler optimizations? <milkman[bot]>guix-bioinformatics/guix-bioinformatics - gn/packages/machine-learning.scm at master - guix-bioinformatics - Genenetwork <rekado>oh, sorry, looks like we already removed the simple tensorflow package <efraim>I also have an ugly keras backport of AUC from 2.3.1 to 2.2.4 <raghavgururajan>milkman[bot] What is your excuse for not incrementing score for civodul then? <milkman[bot]>Your computer's union contract is set to expire at midnight. *raghavgururajan was saved from an ass-kick <nckx>Someone knows how not to get kicked. raghavgururajan: Does the new key work? <nckx>Excellence achieved. Just curious: why ed25519 → RSA? <raghavgururajan>Oh, I just created a key without any arguments. It just used default. *raghavgururajan did just `ssh-keygen` <civodul>it's terrible that we had a prototype years ago and never completed it :-/ ***leoprikler_ is now known as leoprikler
<nckx>raghavgururajan: Sure. It's just (IME) not common to see people switch ‘back’ from the smaller ed keys. Neither is meaningfully less secure. <stikonas>ed25519 works much better than rsa4096 with gnuk... <rekado>civodul: do you think this could be an Outreachy project? <rekado>with a good mentor we could probably get something useful soon <raghavgururajan>nckx: Ah I see. I thoughtopenssh used rsa as default, as must be widely suported/adopted. <raghavgururajan>milkman[bot] is undergoing diagnostics, to see what's wrong with scoring. <stikonas>well, these days everything should be getting widely supported, even dropbear now supports ed25519 keys <sneek>I last saw alezost in (here?) 12 days ago, saying: civodul: hello!. *civodul prepares for another berlin reconfig ***vlorentz1 is now known as vlorentz
<civodul>anyone here uses Emacs 27.1 with Helm? <reepca>civodul: did you get a chance to try that modified patch I debpaste'd yesterday? I think it should resolve that infinite loop. <civodul>reepca: no, i probably lost it in one of my Emacs crashes <civodul>ah ha, M-x profiler-report says it's helm-ff--cache-mode-refresh <milkman[bot]>Helm development is now stalled · Issue #2386 · emacs-helm/helm · GitHub <zimoun>I have tried “make release” and I get: guix-1.0.1.22205-a8360-dirty/gnu/packages/commencement.scm:// /gnu/store/cq0cmv35s9dhilx14zaghlc08gpc0hwr-tcc-boot0-0.9.26-6.c004e9a/lib/libc.a: error: 'sigprocmask' defined twice <zimoun>error: store file names embedded in the distribution <civodul>zimoun: so what's the deal with Helm? <mroh>civodul: try (setq helm-ff-keep-cached-candidates nil) for your "helm eating mem on every tick problem". <mroh>this cache is buggy like hell, IMHO. <mroh>my pleasure! hope it helps ;) <civodul>if i'm still here 3 hours from now, it may mean that it worked :-) <civodul>are you actually using Helm with 27.1? <zimoun>civodul: when trying “make release”, I fail with error: 'sigprocmask' defined twice. What does it mean? And what am I doing wrong? <civodul>zimoun: no idea, i'd need to see the log ***xez[m] is now known as xerz
<zimoun>civodul: the output is “guix-1.0.1.22205-a8360-dirty/gnu/packages/commencement.scm:// /gnu/store/cq0cmv35s9dhilx14zaghlc08gpc0hwr-tcc-boot0-0.9.26-6.c004e9a/lib/libc.a: error: 'sigprocmask' defined twice” and then I do not know what to do <civodul>zimoun: i'm missing context: what command spawned by "make release" produces this line? <civodul>perhaps that rings a bell to janneke though :-) <civodul>in other news: Cuirass is back up with the latest daemon <zimoun>I do not know which commands. I am reading relase.org and running step by step. <civodul>maybe you can open an issue where you paste as much as possible of the command and output <zimoun>ok, I will do. It was just in case I missed an “obvious” and “well-known” option :-) <zimoun>civodul: other news: SWH is move the sources.json from their staging to production. So it should be good this week :-) <janneke>zimoun, civodul...no real bells -- "something changed?"; is this on core-updates? <efraim>I think you mean to ask vagrantc <zimoun>janneke: I do not know. I could do wrong. From a fresh checkout, inside “guix environment guix”, I just run ./bootstrap, ./configure, make. Then “make doc-pot-update” and last “make release”, failing with the previous message. <janneke>i don't believe anything should have changed in the bootstrap, since 1.1.0..weird <zimoun>I could have something misconfigured. <janneke>civodul, zimoun: ah crap! yes, this rings a bell <janneke>the problem is the store file name in a comment in commencement.scm! <janneke>the comment should be changed to something like (litteraly!): <janneke>...-tcc-boot0-0.9.26-6.c004e9a/lib/libc.a: error: 'sigprocmask' defined twice <janneke>but unfortunately, changing the comment triggers a rebuild world <Tirifto>Hello! Has someone installed Guix System with a display manager other than GDM? I remember the graphical installer generating a system definition at the end, when trying it out in a VM, but I’m not sure it would be trivial to change. :) <zimoun>janneke: thanks for explaining. I am not sure to understand why/how the comment produces the error. Does it mean a bug should be opened? <janneke>zimoun: there is a check that no store file names are present, this comment makes the test fail <janneke>i'm sure civudul knows where that check lives <janneke>zimoun: yes, opening a bug would be good to discuss the way forward -- unless maybe civodul shares an a-ha moment about now ;-) <bavier[m]1>Tirifto: I believe a few people here run the SLiM manager <zimoun>janneke: thank you. I will report the issue and let discuss it overthere. <rekado>hmm, axoloti-patcher-next stopped working for me <rekado>specifically the cross-toolchain (arm-none-eabi-nano-toolchain-7-2018-q2-update) <rekado>I see this error when trying to compile something: <rekado>/gnu/store/gywf0ll6p5j4d9q3pbigsphg1cpmjwaw-libstdc++-arm-none-eabi-7-2018-q2-update-1.261907/arm-none-eabi/include/cmath:41:10: fatal error: bits/c++config.h: No such file or directory <rekado>does this look familiar to anyone? <rekado>CROSS_CPATH, CROSS_CPLUS_INCLUDE_PATH, and CROSS_LIBRARY_PATH are set in the wrapper of the executable. <rekado>looks like the problem is that the armv7e-m directory no longer exists in /gnu/store/yzkb3vnb25w848qv62vviypxk5lsz3yc-arm-none-eabi-nano-toolchain-7-2018-q2-update-1.261907/arm-none-eabi/include/arm-none-eabi/ <apteryx>by no result, I meant not the result I'd expect (43423) <raghavgururajan>apteryx: Thanks for the advice on SSD. I have placed an order for WDS100T2B0A. <rekado>probably the worker not … working <rekado>probably the same issue as for the mailer <nckx>raghavgururajan: We have two memtests, one for during Linux, one for before. <nckx>‘guix search memtest’ finds both. <nckx>That's what memtester does. <nckx>It can't test kernel memory, but that's a small percentage. Boot memtest86+ if you need to test 100%. <rekado>apteryx: the worker (responsible for indexing) is now running in a screen session. <nckx>A fine idea but I'd like to see it optional. The fewer ‘but I don't care about X!’ that can block a reconfigure when they fail to build, the better. <nckx>By which I mean: ‘Check my memory’ isn't something people need to do in a pinch. They can add memtest-boot-service or whatever & reconfigure & reboot. If their RAM is so full of holes that ever that fails, well... <nckx>Options for off-line fsck (--repair) for relevant file systems would be cool. <nckx>Although what it's really working around is the lack of a usable (non-bournish) rescue mode... <raghavgururajan>nckx: What argument should I pass to memtester for full/complete test? <nckx>I assume you discovered ‘man memtester’, or that --help still prints all options 🙂 <nckx>It needs to allocate memory to test it so the less you're running (no HTML5000 browsers plz) the better. ***dftxbs3e_ is now known as dftxbs3e
<zimoun>why sometimes guix build foo –with-source=path/to/local/archive.tar.gz says: with-source had no effect on foo? <nckx>File name not containing the package name? <zimoun>nckx: ah yes, now I remember; you already told me. :-) <aitzkora_>when using --container, is it normal to have problem with network interfaces ? <zimoun>aitzkora_: you mean “guix environment –container”, right? If yes network is down and you need the option –network <zimoun>nckx: this link between file-name and package-name, why? <nckx>How would Guix know which package source was being substituted otherwise? <nckx>If you omit PACKAGE in --with-source=PACKAGE=SOURCE, Guix has to guess. It does so based on the package name. <nckx>We could add hueristics such as ‘if the file name matches nothing, it applies to the first package name on the command line’ but this is both complex and fragile. Just specify PACKAGE. 🙂 <nckx>(guix)Package Transformation Options basically restates what I said above, but let me know if anything's unclear. <nckx>(‘Complex’ in the sense that every little exception/special case adds complexity to a large system; of course it would be trivial to implement & document on its own.) <goldenshimmer>Do I need to back up anything from the home directory before running guix package --delete-generations, or is everything interesting included in /var/guix/profiles? <nckx>zimoun: That said, it was suprising to me that ‘guix build --with-source=ncurses=... knot’ didn't do what I expected, so we're not making full use of the flexibility we claim. <nckx>goldenshimmer: What is it you want to back up? <nckx>Or should I ask: what is it you fear losing? <goldenshimmer>I'd like to have a copy of my old profiles for reference, so I backed up /gnu and /var/guix; I intend to delete old profiles and then garbage collect the store to free up disk space.