IRC channel logs


back to list of logs

<retard>what is the relationship between "source" and "target" under mapped devices for LUKS encryption?
<retard>I don't quite understand the "mapping", but I have two partitions: root and swap. I would like root encrypted
<retard>well, really, I want full disk encryption so I suppose swap should be encrypted too.
<retard>nvm, found solution:
<retard>I think it's solution was more helpful for noobies than the official guide, fwiw
<civodul>retard: which solution is more helpful?
<quiliro>civodul: the one on the link
<retard>in regards to the encryption and mapped devices poriton
<quiliro>i have disconnected keyboard and monitor from the machine which i
<quiliro>am installing
<civodul>retard: ok
<quiliro>is there a way i can connect from another machine on the lan to see the state of the installation
<civodul>retard: please consider submitting the doc change to
<retard>civodul: ok@
<civodul>preferably as a patch against doc/guix.texi, which is the source
<civodul>ACTION -> zZz
<civodul>good night/day!
<ng0>I keep running into issues with guix system vm.. did something change with the vm?
<ng0>do I need to run a clean of .go files again? I ususally just 'make'
<quiliro>is there a way to connect remotely to a running live install guixsd?
<quiliro>i tryed ssh root@
<quiliro>ssh: connect to host port 22: Connection refused
<quiliro>but ping works
<quiliro>perhaps ssh-server is not installed on the live usb or root is denied remote access
<ng0>I expected that you generated the disk-image from a recent checkout and not the 0.12 one.. newer commits have ssh enabled in the disk-image afaik
<quiliro>i downloaded the 0.12 from the website
<quiliro>1 week or so
<ng0>the 0.12 doesn't change. it's a snapshot
<quiliro>so what would you suggest?
<ng0>you can make your own disk-image
<quiliro>but what should i do with the running installation?
<ng0>idk.. you could also try telnet or whatever. if there's no way to get in then there's no way to get in
<quiliro>$ telnet 22
<quiliro>telnet: Unable to connect to remote host: Connection refused
<quiliro>ng0: i will connect a usb keyboard and a monitor and connect locally
<ng0>it would be port 23, but I forgot that this needs a service as well
<quiliro>i cannot do it today because i already disconnected the ps2 keyboard
<quiliro>it will not work if i reconnected
<quiliro>i think
<ng0>in any case, disk-image is what you want when you want ssh now, before the 0.13 snapshot release
<quiliro>will it damage the port or the keyboard if i connect the keyboard while the computer is on?
<ng0>the ps2? no
<quiliro>ng0: how long +/- do you think it will take to get published?
<ng0>no idea.. 'soon'
<quiliro>so you suggest to connect?
<ng0>do what you think works for you
<quiliro>soon is less than 6 months?
<quiliro>less than 1 month?
<ng0>more like this weekend or start of the week if no major issues appear
<ng0>I'm not in the loop
<quiliro>is it possible to continue an installation later if i suspend it and shutdown now?
<quiliro>i get guix system : error: build failed: build of /gnu/store/...-guix-0.12.0-11.....drv failed
<quiliro>phase check failed after 1146 seconds
<quiliro>gotta go
<ng0>cleaning out the .go files and bootstrap again did not fix the vm issue...
<DoublePlusGood23>can anyone assit me with a setting up a offload server?
<retard>How much room should I allow my root partition to be?
<rain1>it must be bigger than 4GB
<retard>so 10GB is enough?
<retard>you don't think that'd harm me down the road?
<rain1>I used 10GB it worked
<retard>I just ask because I ran out of space on disk, probably forgot to format and clear my harddrive first
<retard>ok thx
<brendyn>My /gnu/store is 18.8 GiB at the moment and I'm not even on GuixSD, just some stuff installed on parabola
<sturm>Hmm, I can't seem to "sudo guix pull" at the moment. Getting some warnings about "loading ....go failed ... bad header in object file", then continues and halts with "error: sed-hurd-path-max.patch: patch not found"
<sturm>(this is just normal "sudo guix pull" on GuixSD, not from a checkout)
<retard_>after "herd start cow-store /mnt I get "herd: exception caught while executing 'start' on service 'cow-store': invalid arguement", followed by "ERROR: In procedure mount: mount "/.rw-store" on "/gnu/store": Invalid arguement"
<retard_>Now I'm getting a different ERROR: in procedure rmdir: Device or resource busy.. what might this mean?
***amuck_ is now known as amuck
<alezost>sturm: it may be that your "guix" still uses Guile 2.0, but modules in ~/.config/guix/latest/ are compiled with Guile 2.2. Look at the first line of the guix script ("which guix"), and check what guile version it is
<catonano>git config returns my real name and yet the patches prodced by git format-patch still use the nickname I used previously
<catonano>git config --list | grep humanities
<catonano>this is thhe only occurrence of "humanitiesNerd" I can ind in te git config
<janneke> /join #lsh
<janneke>yay, /me got lshd pubkey login setup...RTFM
<janneke>somehow i thought lshd would use ~/.ssh/authorized_keys ...
<wingo>doing a guix package -u, it seems there is a cached copy of texlive-texmf-2016 again
<wingo>3.whatever gigabytes
<wingo>installed at least
<wingo>i thought we didn't want to cache that one?
<janneke>hmm, system reconfigure gives
<janneke>janneke/dundal.scm:49:5: janneke/dundal.scm:49:5: In procedure module-lookup: Unbound variable: menu-entry
<rekado_>I hope that in two weeks I’ll have a replacement for the annoying texlive mega-package.
<rekado_>everyone: if you contributed something since 0.12.0 and think it should be mentioned in the NEWS for 0.13.0 please let me know.
<ng0>isn't the listing of packages in NEWS generated?
<ng0>ah, something in general.
<sturm>thanks alezost: I'll give that a Guile 2.X thing a look
<rekado_>ng0: yes, the list of new packages and updated packages is generated from the log.
<rekado_>ng0: I meant noteworthy changes
<ng0>I would have a second new window manager to add, but guix system vm is currently broken for me and I have to wait until a long conversion process is done here.
<OriansJ>lightdm package definition is currently showing 404 "Not Found"
<ng0>not here..
<ng0>you mean the homepage?
<OriansJ>sorry lightdm-gtk-greeter
<ng0>same, description is not empty
<OriansJ>attempting to download results in resource not found
<janneke>OriansJ: hi!
<OriansJ>Hopefully you are having fun janneke
<janneke>OriansJ: Your labeled hex is a great idea, and timing is perfect!
<OriansJ>good :)
<janneke>i had just implemented dumping some flavour of `.o' file (my complicated lambdas)
<janneke>and i'm now working to link multiple .o's into one elf
<janneke>as soon as that's done, i'd like to see if i can change to your labeled hex
<OriansJ>well just output multiple .hex2 files (labeled hex) and cat them together to create the desired elf
<OriansJ>aka elf header is first file (as labeled hex) and the compiled lambdas just are labeled hex files that are appended.
<rain1>hello :)
<OriansJ>rain1: Great to see you, how are things?
<janneke>sure...but i'd like to take small steps and possibly go throug a heterogenous lambda/labeled-hex stage
<janneke>hi rain1!
<rain1>I had an idea.. might not be good but maybe an interpreter for pure lambda calculus could be useful. It can be implemented with reference counting instead of a full gc
<OriansJ>janneke: I trust you to do what you think is the better way of implementing :D
<rain1>I started a wiki to take notes as i read the code not much there yet though
<janneke>now that we have ascii .o format/hex2, transition should be painless
<OriansJ>rain1: You missed C500
<janneke>rain1: gives me warm fuzzy feelings!
<rain1>:D oh great, let me add it
<janneke>rain1: i think prescheme deserves a place there too?
<rain1>true, adding. and feel free to make an account and edit this too if you like!
<OriansJ>actually I was thinking about this today:
<rain1>my compiler produces a kind of prescheme, lambdas are hoisted and nested expressions are denest to produce long LET*s
<OriansJ>rain1: We are going to have to work on your bootstrap but I think that'll be fun
<sturm>thanks alezost: based on your suggestions I switched root's "guix-latest" to point to the seemingly working version from my regular user account. The "sudo guix pull" then ran smoothly. My "system reconfigure" seems to be building webkitgtk though so I might have to come back tomorrow to see if it's all worked out :P
<janneke>ACTION just linked libc-mes.o + m.o -> m executable :-)
<janneke>now to update the build system
<OriansJ>great job janneke :D
<OriansJ>I'm currently stuck debugging a minor regression with the garbage collector that is only affecting (define foo (let ((..)) (lambda ...))) expressions
<janneke>OriansJ: i have never been stuck more often than this past year, hacking on mes
<janneke>and on almost every instance i got unstuck the next day
<OriansJ>janneke: I complete agree, garbage collection is like the ultimate reminder that we as humans keep forgetting tiny details that always end up biting us
<janneke>OriansJ: if you would talk about the stuckiness, feel free...talking has helped me a lot
<OriansJ>(define foo (lambda (x) (let ((a 4)) (+ a x)))) works just fine with garbage collection both on and off
<OriansJ>yet (define foo (let ((a 4)) (lambda (x) (+ a x)))) works only when garbage collection is turned off
<OriansJ>I might have to implement lazy unmarking to prevent infinite recursion to get it to work
<rain1>about hex2, the labels are written as 1 byte?
<rain1>relative jump?
<rain1>ah, 16 bit
<OriansJ>rain1: @label is 16bit relative $label is 16 bit absolute and &label is 32bit absolute
<OriansJ>and the labels are written as :label for ease of parsing
<rain1>very elegant how the assembly language is built up, using line macros
<jlicht>very exciting to see y´all have this back and forth on these interesting project
<rain1>but this hex code language, it's a kind of virtual instruction set - how is it executed?
<OriansJ>rain1: using the vm included in stage0 with is made with make vm
<OriansJ>or if you prefer make vm-minimal
<OriansJ>rain1: the core design is in vm.c, the implementation of the instructions are in vm_instructions.c and the decoding logic is in vm_decode.c
<OriansJ>tty.c is for allowing raw access to key presses in modern environments
<rain1>would there be a name for this vm language/instruction set?
<rain1>oh lilith ?
<OriansJ>I just call it stage0 vm
<OriansJ>but it was based on the lilith implementation of the Knight platform (circa 1967 if I remember correctly)
<rain1>someone showed me this JIT engine they made, pretty neat and simple
<OriansJ>rain1: if you did make libvm it would create a library of the virtual machine, whose interfaces are in wrapper.c and was leveraged by the Web IDE
<rain1>i wonder if a different line assembler could be used to compile the virtual instruction set to x86 instead of stage0vm
<OriansJ>rain1: actually you could in the M0 line macro processor simply put the x86 hex codes for instructions and it would produce a binary that worked on x86 instead
<OriansJ>aka instead of using the stage0 defs file, create a new one with x86 instruction mappings.
<janneke>o wow, i didn't catch that labeled hex2 isn't x86...hmm
<OriansJ>janneke: actually it is platform independent :D
<OriansJ>rain1: shows how simple it is
<rain1>I'm quite stunned by how simple
<rain1>just building up an assembler from nothing :D
<rain1>janneke: the .o file you use, is that a standard file format?
<OriansJ>rain1: although for x86, you might want to make a couple minor changes to better map single characters to things you want
<janneke>rain1:'s very mes-specific *and* x86
<janneke>:-( :-(
<OriansJ>octal is easier for x86 than hex but honestly it doesn't matter once you get past
<rain1>janneke: that,s okay :) im just learning..
<janneke>haha, me oo
<OriansJ>janneke: actually it wouldn't be much work to make your output platform independent
<rain1> neatld by ali rudi may be helpful in some way, he is one the best programmers at creating short code that does a lot
<janneke>rain1: i am planning to go to something much like hex2, preferrably exactly hex2
<rain1>it seems to me that the technique of hex2 can be used to create assemblers for most instruction sets -- maybe a little tricky with binary packing since it works at a hexadecimal pitch..
<rain1>although it may be better to limit attention to x64 at the moment
<OriansJ>rain1: actually all instruction sets map to a multiple of the hexadecimal pitch
<OriansJ>8, 16, 24, 32, ... byte long instructions can all be encoded in Hex
<OriansJ>rain1: as you can see here
<OriansJ>the real kicker for things like x86 is things like this 48 c7 c6 99 01 60 00 # mov $0x600199,%rsi
<OriansJ>48 c7 c0 3c 00 00 00 # mov $0x3c,%rax
<OriansJ>x86 mov is Turing complete
<OriansJ>there is a very important reason why every instruction in the stage0 vm has exactly 1 encoding
<janneke>OriansJ: that's excitingly good news...
<rain1>whats the reason for 1 encoding only?
<OriansJ>rain1: so that the logic required to implement an assembler becomes trivial
<OriansJ>it literally becomes a trivial pattern matcher (DEFINE FOO FF00CCAA)
<OriansJ>where as for x86, you are going to have to create things like (DEFINE MOV_RSII 48c7c6) (DEFINE MOV_RAXI 48c7c0)
<OriansJ>aka x86 will require about 1,900 DEFINE lines to cover the encodings you probably want but stage0 only needs less than 250 lines
<rain1>hmm thats just too much, maybe instead of assembling x86 a langauge could be made from composite operations
<OriansJ>rain1: that is literally what a vm/bytecode interpreter does at run time and a O-code/bytecode compiler does at compile time
<ng0>I have just generated an initial x86_64 image of pragmaOS, with lots of parts still missing. Weights in at around 4GB already because of some of the graphical applications. My wish to use a different libc and compare sizes is justified. I hope this will not grow to 8GB when all applications are finished
<ng0>It looks so easy to maintain the build instructions and the whole process. disk-image is a cool function :)
<rain1>ng0: which libc will you try?
<ng0>I want to compare musl and uclibc-ng builds. It has no high priority, I keep hacking on uclibc-ng on and off for months now.. the issue with it is that it's not just a package but is similar to a kernel build, but more compatible to glibc. But musl needs lots of patches, at least back o nGentoo
<janneke>what's the status on arm? running `guix environment guix' on latest master, i get
<janneke>builder for `/gnu/store/qf5ynngs5hkcjg4p9fn7v0i6ggg296vd-libgpg-error-1.24.drv' failed due to signal 4 (Illegal instruction)
<OriansJ>FOUND IT!!! Finally fixed the garbage collection bug in the High level prototype...
<janneke>OriansJ: yeah!, good job!
<rain1>awesome :) congrats
<OriansJ>now for the real fun programming, back porting those changes to the stage0 assembly implementation
<apteryx[m]>rekado_: not sure if it's worth mentioning, but managed are now indexed so they can be searched with man -k or apropos.
<AndreySuchev>ACTION listen
<ng0>ouch. something is pulling in texfm when I added another package.. I hope the plan to make it modular works out :)
<AndreySuchev>htmlm5 anon
<civodul>hey hey!
<OriansJ>and with that release 0.0.6 of stage0 is complete
<ng0>civodul: I experience the same error with guix system vm since yesterday. I can't really send emails atm, which is why this wasn't reported
<ng0>guix system reconfigure is okay, and guix system disk-image and other too. at least they spit no errors
<ng0>well, okay... I did not start the disk-image, so there is an error with disk-image, jus tread the emai lcloser
<ng0>we had some qemu related commits, maybe there's something with qemu to fix?
<ng0>*related to
<quiliro>how can i specify swap on config.scm?
<quiliro>swap paritition
<quiliro>why is qwmu being installed on a default bare-bones.scm guix system init ?
<lfam>quiliro: To create swap with config.scm, you must first create a partition for swap in the normal way. Then, you can add (swap-devices '("/dev/foo")) to your config.scm. You'll need to change 'foo' to the right name.
<lfam>quiliro: I haven't done this before, but I think it should work
<quiliro>lfam: you do not use swap?
<lfam>quiliro: I haven't needed it yet on my GuixSD machine. I do use it on some other machines
<quiliro>when is it necessary?
<lfam>quiliro: If you are running out of memory, then you'll need it.
<lfam>I can't say if sda2 is the right partition. That depends on your machine.
<quiliro>how to know if you run out of emory?
<quiliro>yes it is the secondçi was just confirming it is not /dev/sda
<lfam>quiliro: There will be messages in /var/log/messages about "out of memory" and "OOM"
<lfam>And things will stop working randomly
<quiliro>oh! great info :-)
<lfam>The (swap-devices) field is documented briefly here:
<quiliro>nad for hybernation, i suppose
<lfam>Actually, I recently noticed that my GuixSD machine does need swap, because I saw the OOM messages in my /var/log/messages. But, I haven't noticed anything broken, so I'm ignoring it for now :)
<lfam>Yes, also for hibernation, but be careful if you set that up. There are caveats in a variety of cases. I think that GuixSD should handle this for the user so they have fewer opportunities to make mistakes.
<lfam>There is work-in-progress code for hibernating on GuixSD, but I don't think it's been tested by many people:
<quiliro>i had a power outage
<quiliro>the mnt partition has been recovered including /mnt/tmp
<quiliro>is there a proceedure to start again?
<lfam>quiliro: If the /gnu/store still exists, you should just run the same command again
<quiliro>lfam: /gnu/store never existed
<quiliro>only /tmp/gnu7store i think
<lfam>That's what I meant
<quiliro>lfam: oh...good
<destt_>could I get some help on making a package for guix?
<destt_>my end goal is haxe ( ), but it looks like I'll need a couple other packages before I get there
<destt_>Here's what I have so far with the haxe package, and I'm almost entirely sure I'm doing it wrong
<lfam>destt_: Hi! What happens when you build that package?
<lfam>That is, why do you think something is wrong?
<destt_>because I'm a complete newb to guix :)
<destt_>ice-9/boot-9.scm:2870:6: no code for module (guix packages ocaml)
<destt_> is the result of guix build -f haxe.scm
<lfam>destt_: You need to import (gnu packages ocaml), not (guix package ocaml). Likewise for the rest of the package modules
<destt_>ok, I have done that. (gnu packages camlp4) isn't working although I see when I run guix packages -s camlp4
<lfam>destt_: You need to import the module that contains camlp4. That is (gnu packages ocaml)
<destt_>oh, thanks
<Sleep_Walker>it seems that documentation and code is out of sync - documentation use menu-entry, but it seems to be internal for gnu/bootloader/grub.scm and there is some new boot-parameters record which is undocumented
<lfam>Sleep_Walker: The web-based documentation or the documentation in the Git repo?
<Sleep_Walker>lfam: I tried both
<Sleep_Walker>Git one is easier to grep
<janneke>system reconfigure gives me: janneke/dundal.scm:49:5: janneke/dundal.scm:49:5: In procedure module-lookup: Unbound variable: menu-entry
<lfam>The web-based documentation will almost always be out of sync, because it corresponds to the latest release. If the latest documentation in the Git repo is out of sync, that's a bug. Can you report it?
<Sleep_Walker>lfam: sure
<civodul>janneke: could you check on the version-0.13.0 branch as well?
<civodul>that would be a regression
<janneke>civodul: this is master, will check now
<lfam>Speaking of the web-based documentation, I'd like to be able to update it to fix "real" mistakes. That is, when it doesn't even correspond to the latest release.
<lfam>I think I've had to explain that '-net default' is a typo in 'Running GuixSD in a VM' at least 5 times now
<Sleep_Walker>janneke: that is the problem I got
<civodul>lfam: yeah i think we should set up a cron job like you suggested a while back
<lfam>civodul: Can you remind me what I suggested? :)
<civodul>setting up a cron job :-)
<lfam>To keep the web-based manual up to date with the latest code? I do think we should offer the latest manual on the web. But it's also useful to keep the release version online.
<janneke>civodul: yup, broken on version-0.13 too
<lfam>For example, the Borg project offers both: <> and <>
<lfam>But, this does complicate things. Your users must know which manual to look at. And people who need the release manual are the people least able to figure that out :/
<lfam>So... a trade-off
<janneke>civodul: no, waitaminute...i rebased from master onto 0.13...brb
<janneke>civodul: Sleep_Walker: version-0.13.0 has menu-entry; you need to include (gnu system grub)
<janneke>only master has this problem
<janneke>still cannot reconfigure, now i'm looking at: guix system: error: build failed: directory `/homeless-shelter' exists; please remove it
<janneke>i never know how to figure out what package had a download problem here
<janneke>ah, using --keep-failed i get: note: keeping build directory `/tmp/guix-build-enchant-1.6.0.tar.gz.drv-0'
<Sleep_Walker>janneke: thanks
<civodul>janneke: ok, thanks for testing
<civodul>janneke: could you report the menu-entry problem in the bootloader debbugs entry?
<civodul>or on guix-devel
<Sleep_Walker>I already reported that as 27007
<Sleep_Walker>but I thought that documentation is missing
<civodul>oh good
<destt_>As a guix newb, how would I package up neko ( )? This is what I have so far:
<lfam>destt_: You should try to make a neko package that includes the listed packages as inputs and uses the cmake-build-system.
<lfam>I think we have all of those dependencies packaged already
<destt_>lfam: I can't seem to find mbedtls
<lfam>destt_: Try `guix package --search=mbedtls`
<lfam>It seems like you are on the right track, but you'll need to use the cmake-build-system instead
<lfam>And, you did find mbedtls-apache. So, try building neko with cmake, and see what happens :)
<destt_>Hmm. guix package -s mbedtls doesn't return anything for me
<lfam>destt_: It should show you mbedtls-apache
<lfam>destt_: It's possible your copy of Guix pre-dates the inclusion of mbedtls-apache. You can update it with `guix pull`
<destt_>lfam: trying that, you're probably right. I installed release and never ran `guix pull`
<lfam>Yikes :)
<AndreySuchev>ACTION eat anon fn
<lfam>I recommend frequent updates. You can always roll-back
<destt_>lfam: it's only been a day :)
<lfam>We push security updates to the packages almost daily, so I want everyone to be safe :)
<lfam>I mean, perhaps it doesn't matter that much. The high frequency of important security updates implies that there are *many* undiscovered or undisclosed bugs still there
<destt_>Still, better to be as secure as possible
<AndreySuchev>ACTION eat anon fn
<rekado_>argh, I now know why libstdc++ didn’t work in my arm-none-eabi toolchain
<rekado_>I had only passed ‘--target’ but not ‘--host’, so it didn’t actually cross-build the thing.
<rekado_>ng0: why bother with something other than the GNU libc at the bottom? Is this for a router?
<quiliro1>please remind me how i can make the new version of the usb beyond 0.12 that is published
<quiliro1>so i can boot in efi with guixsd installer
<ng0>rekado: yes, among other ideas. and also adding pieces into the puzzle framework of extending possibilities for Guix and GuixSD deployment
<rekado_>ng0: would you also set up a build farm to rebuild everything with that other libc? Or would users have to build everything from source?
<ng0>depends on our resources at that point and how we end up deploying pragmaOS builds, but the goal is to have an build-farm at some point to automate certain things. I don't expect users to build anything.
<quiliro>lfam: you told me that if /mnt/tmp was intact after power outage happened during installation, installation could be continued....should i follow the process in the guixsd installation manual as is (except for partitioning and partition formating)?
<ng0>alternative libc is at the very end of list of priorities.
<lfam>quiliro: Yes
<quiliro>of course i do not need to create /mnt/etc/config.scm or /mnt/etc
<quiliro>lfam: it is connecting to hydra now
<ng0>I'm also looking to either re-implement prep or write a parser for prep and add a reader for it to Haunt so that multi language pages generated from one document are possibel without the pain of pure xml
<ng0>as Haunt seems to have no opinion on anything it seems better than to continue writing my own simple generator
<ng0>which right now is using smu, make and optionally Guix
<quiliro>lfam: it is redownloading
<quiliro>and i already have 3,2GB
<quiliro>3.2 in English
<quiliro>how can i make it check before downloading
<quiliro>and download if the files are not found
<quiliro>it has been 2 days of installation now
<quiliro>and i cannot hold sleeping in the sofa any more
<quiliro> stopped
<quiliro>downloading and it is compiling now
<quiliro>i think it will download some more latter, right?
<quiliro>in the meantime how can i go into my guix installation over debian
<quiliro>it is done now
<quiliro>oh...just open a terminal and use it
<quiliro>guix package -i nginx
<quiliro>it works
<quiliro>error now