IRC channel logs


back to list of logs

***hopses is now known as hospes
***hopses is now known as hospes
*davexunit really hope the static guile he builds does the right thing so he doesn't have to wait through the build again
<davexunit>damn it, I've gotta build it again.
<sneek>davexunit, you have 1 message.
<sneek>davexunit, civodul says: thumbs up for 'guix environment --container'!
<davexunit>my idea for the above ^ is that I will ask the not-yet-built container daemon to spawn a new container with $PWD and /gnu/store bind mounted in.
<davexunit>and then the command passed with -E will be run as PID 1
<davexunit>this way we'll be able to quickly make environments that are very similar to what guix-daemon is creating.
<davexunit>and then I suppose in the glorious future when guix-daemon uses pure guile they will be unified.
<ewemoa>anyone working on gparted?
<mdln>My GuixSD install went smoothly! :D
<ewemoa>here's an attempt at gparted:
<civodul>Hello Guix!
<ewemoa>still not much luck jumping to symbol defs -- as an interim work-around, in the guix src dir: find . -type f -iname "*.scm" | etags - seems to lead to a somewhat useful TAGS file
<civodul>ewemoa: did you try 'guix edit'?
<ewemoa>civodul: i haven't tried that -- will test
<ewemoa>civodul: that is nice for finding a package definition -- tnx! -- what i've been looking for is a quick way to get to defs of things such as wrap-program, substitute*, and the like though
<amz3>if jump to def doesn't work, M-x rgrep can be help
<amz3>all do all the time
<amz3>I've added (global-set-key (kbd "C-x C-r") 'rgrep) to .emacs
<ewemoa>amz3: i have been using grep and friends but find having to look through results tiring :)
<amz3>look at the bright side, you get a glimpse of some other code :)
<civodul>ewemoa: if you use Emacs, Geiser is the best thing
<ewemoa>he he -- it's true that i appreciate not finding an answer immediately -- there is definitely value to the "extra" stuff one ends up seeing while searching
<civodul>alternately you could try GNU GLOBAL
<ewemoa>civodul: i use emacs, but i haven't gotten geiser's jumping to def to work it seems
<ewemoa>i tried gnu global but it dies unfortunately
<civodul>Geiser can only jump to files that have been compiled in the inferior guile process
<civodul>so you first need to C-c C-k a bunch of files
<ewemoa>yes -- it seems so, yet, i have failed to get that to work so far
<ewemoa>at least find-tag is working with the tags table for the moment -- thanks any way :)
<amz3>I did a script to generate a dot file, I tested it with rekado qsynth, but the output of my program and the output of guix package -n -i qsynth is different
<amz3>my output is missing ruby
<amz3>I will update it to work with sigmajs this afternoon
<amz3>ewemoa: what are you hacking on guix?
<ewemoa>amz3: i am just packaging various things and trying to learn some internals
<civodul>amz3: nice! that's actually more readable that i expected
<civodul>the good thing is that you used the package level rather than the derivation level
<civodul>so it's less "noisy"
<civodul>graphviz could handle that, i guess
<amz3>some package are missing, like ruby. maybe-create-edge might be buggy
<civodul>i think that's because it's an implicit input
<civodul>so it's never explicitly listed in the 'inputs' fields of packages
<civodul>the up side is that the graph is more concise
<civodul>the downsize is that it's less accurate
<amz3>it was in the graph before I introduce maybe-create-edge
<civodul>well dunno, it didn't look at the code
<rekado>amz3: nice!
<rekado>I find qt scary.
<civodul>actually we could have three levels of details: packages, bags, and derivations
<civodul>the last one is too much for graphviz to hanlde
<rekado>why does Qt need to have postgresql and mysql as inputs?
<civodul>heh, who knows ;-)
<rekado>and cups...
<rekado>I wonder if it would be possible to split Qt.
<ewemoa>it's secretly trying to become systemd
<civodul>well GTK+ also depends on CUPS, for its printing dialog
<rekado>what bothers me about this is that I think I *know* what parts of Qt qsynth doesn't use.
<rekado>it's unfortunate that it "depends" on so many unused things.
<rekado>pulseaudio for example. Qsynth is just a frontend. The audio part is handled by fluidsynth, which uses alsa or JACK.
<amz3>(it's rhytmbox that pulls mesa
<amz3>see you later
<rekado>It's frustrating to read mailing list discussions about Qt packaging.
<civodul>rekado: does that means that solving the bundling problem is hard?
<rekado>it means that upstream doesn't think it's much of a problem.
<rekado>"I understand that you have certain policies, but the reality out there is
<rekado>that a certain level of pragmatism is required if you want to get a usable
<rekado>and competitive product."
<davexunit>good morning geeks
<civodul>rekado: terrible
<civodul>rekado: we should probably just remove QtWebEngine for now, as Fedora does
<civodul>hey, davexunit
<davexunit>hey civodul
<davexunit>thanks for patch review
<davexunit>static guile and I are happier now
<civodul>you live in perfect harmony?
<davexunit>hehe yes
<davexunit>now that I can have cgroups
<davexunit>civodul: my 'clone' syscall procedure seems to have an issue. I can't successfully call 'waitpid' on the pid that it returns to the parent process.
<davexunit>I get a "no child processes" error
<civodul>clone(2) says it returns the "thread ID of the child process"
<civodul>so that's not exactly a PID
<davexunit>need to figure out the new process id somehow...
<civodul>well i don't know if it really is different
<civodul>davexunit: ignore what i wrote: the example in clone(2) treats the return value as a pid
***hopses is now known as hospes
<davexunit>civodul: I just checked and it does return a valid pid
<davexunit>I pk'd the pid had the container process sleep for awhile
<civodul>and still waitpid fails with ECHILD?
<davexunit>and it was indeed in the process list
<davexunit>hmm, does it's special status as PID 1 in a new PID namespace change things?
<davexunit>could be the issue...
<civodul> (based on the example in clone(2)) works as expected
<civodul>that is, waitpid succeeds
<davexunit>I must be missing something in my clone implementation
<davexunit>I am directly making the syscall
<davexunit>not going through glibc
<davexunit>due to all that stack nonsense
<civodul>ah right
<davexunit>it could as simple as the flags, though
<davexunit>I'm not using SIGCHLD
<davexunit>that's probably the issue
<davexunit>thanks for helping me see the obvious, civodul.
<davexunit>one problem down.
<davexunit>time for "real work" now, but I'm going to dig into making interactive container processes next.
<civodul>well, enjoy your daytime work!
<davexunit>docker containers by day, guix containers by night.
<davexunit>sudo guix environment --ad-hoc --pure --container guile -E "guile -c '(sleep 5)'"
<davexunit>totally worked.
<civodul>that's really cool
<civodul>--totally-pure could be an alias for --pure --container
<davexunit>civodul: haa yeah
<davexunit>I don't know if this is the greatest idea, but to make things "just work", I bind-mound $PWD as /env within the container
<davexunit>so that someone can just run it from the root of a source tree
<davexunit>and have their build environment.
<davexunit>actually, --container makes --pure redundant in some ways
<davexunit>--container should imply pure, because the current environment variables don't really apply to the container process
*davexunit reads
<davexunit>currently using phusion's "350 line my_init system" at work, I want to rip it out and put dmd in its place.
<civodul>davexunit: but by default the environment variables are inherited by the child process
<davexunit>civodul: yeah, but this doesn't make sense for a container. we ought to cleanse the environment first.
<davexunit>having $HOME/.guix-profile/bin on $PATH within the container doesn't make much sense.
<civodul>seen in a real piece of software:
<cehteh>wow .. c++ ftw m(
<davexunit>civodul: we're pleased to announce the release of UnknownDistro 0.8.3
<cehteh>HowTo "Portable Programmers with a crowbar"
<civodul>i find it hard to believe that such a "high-profile" project has this kind of hacks
<cehteh>i just was amazed that this pastebin has a nice brace/parens highlighting ... but then .. ohno its a lisp pastebin :))
<cehteh>and nice spam there
<davexunit>yeah, has a spam problem
<paron_remote>civodul: wowee
<paron_remote>re: that paste
<paron_remote>I'm thinking in a month or two I might want to get a gluglug x200 while that's still possible and dual boot with guixsd and debian on it
<paron_remote>I have an x220 but I feel like I can't trust it ever since I learned about intel amt
<paron_remote> so scary :(
<paron_remote>maybe I'll give the x220 to my roommate
<paron_remote>it still works nicely
<paron_remote>I just can't trust it
<paron_remote>my roommate already uses an ipad so I doubt she has the same level of paranoia :)
<Mikko->ugh, got hit by kernel panic after install ~.~
*civodul uses 'guix environment' for continuous integration
<civodul>Mikko-: of GuixSD 0.8.2?
<civodul>could you post your OS config?
<paron_remote>civodul: guix size is cool, thanks for adding it :)
<Mikko->mostly left it unmodified from the default desktop.scm
*civodul tries it
<civodul>we need a web service that builds an OS config and boots it in the browser
<civodul>that would be fun
<civodul>Mikko-: for your root file system, you have (device "/dev/sda1") (title 'label)
<civodul>the 'title' should be changed to 'device
<civodul>or simply omitted
<Mikko->okay... retry
<civodul>Mikko-: so that was a kernel panic, not a Guile prompt?
<Mikko->yes, it was a panic
<civodul>something's fishy, normally one would get a prompt
<Mikko->didn't see a prompt, probably should've smashed the keyboard :S
<civodul>instead of reinstalling, you can also edit the GRUB line and modify the --root argument
*civodul has to go
<anthk_>paron_remote, I think I'll try to yse coreboot
<paron_remote>anthk_: good idea, though not sure why you're pinging me about it :)
<anthk_>you made me rethink about intel's remote spying
<paron_remote>anthk_: ah... I think the problem is that even installing coreboot might not disable AMT :(
<anthk_>sometimes my screen turns itself on at night, over 60 minutos, idk if it's an isse with DPMS or...
<Mikko->your screen is haunted? o.o
<anthk_>Mikko-, idk, but it's scary, it flashes for 1-2 seconds, and then it turns off
<mark_weaver>paron_remote: yes, coreboot normally includes the AMT blob.
<mark_weaver>at present, we don't know how to make an X220 work without AMT.
<mark_weaver>the folks on #libreboot can tell you more about this.
<fchmmr>paron_remote, if you're going Intel, get an X200
<fchmmr>X220 will never be supported in libreboot.
<paron_remote>fchmmr: yeah, that's what I was saying above
<paron_remote>I got an x220 before I knew about AMT.
<fchmmr>Libreboot is focussing on AMD and ARM now.
<fchmmr>Intel is dead.
<paron_remote>dead to freedom? :)
<fchmmr>AMD's heading in a bad direction too. I'm doing what little I can to try to persuade them not to.
<paron_remote>everything is heading in a bad direction
<paron_remote><bkuhn> free software dark ages
<fchmmr>AMD always lags behind Intel, though.
<paron_remote>we're going to have little choice soon but for hackers to figure out how to get more deeply involved in hardware hacking
<fchmmr>Fortunately, this also applies to going proprietary.
<fchmmr>They've taken longer than Intel to get worse.
<fchmmr>And unlike Intel, have actually tried to be nice.
<fchmmr>Right now, AMD can still be influenced.
<paron_remote>it's too bad. Intel at least had free graphics drivers for a while there
<fchmmr>Especially since the company is in serious trouble.
<paron_remote>even now that is gone :(
<fchmmr>AMD will literally do anything to make more money.
<fchmmr>That means they can be swayed by us, potentially.
<fchmmr>We just need to demonstrate that respecting user freedom is profitable (it actually is. I know this for a fact).
<paron_remote>it's tough to prove that free software users are much of an economic force
<fchmmr>I disagree.
<fchmmr>We're a force to be reckoned with (not in all areas, but in many we are).
<mark_weaver>it could be one of the few ways that they can distinguish themselves in a positive way compared to intel at this point.
<fchmmr>servers, for instance.
<fchmmr>ISPs, hosting providers, sysadmins, etc.
<Mikko->no dice, I think it failed at mounting the filesystem
<Mikko->oh, it booted
***schjetne` is now known as schjetne
***init is now known as exio4
***exio4 is now known as init
***hopses is now known as hospes
<rekado>git is very big with 224MiB.
<davexunit>rekado: I've been meaning to see how we can split that up
<davexunit>now we can run 'guix size' on it and see what's up!
<davexunit>^ my life at the moment