IRC channel logs


back to list of logs

<mthl`>davexunit: have you look to the website->sxml thing?
<mthl`>I tried to convert the top index.html,
<mthl`>you will notice that i stop trying formatting at some point when i realize that i can't make it pretty.
<mthl`>I will continue the work tomorrow, if it's not done yet.
***tschwing_ is now known as tschwinge
<rekado_>hmpf, after repeated requests for a license declaration of a bioinfo tool I intended to package, the university department sent me a silly "standard academic license" with tons of restrictions on use, modification and copying.
<rekado_>I find it especially annoying because they use LGPL code.
<rekado_>is this even permitted under the terms of the LGPL? The code in question is compiled to objects and linked into the final executable.
<rekado_>I miss the "infectiousness" of the regular GPL.
<rekado_>I'm so disgusted by this "standard academic license".
<civodul>hm, hello Guix!
<rekado_>hello civodul!
<civodul>"infectious" or "contagious" to describe copyleft is propaganda
<civodul>call it "liberating"
<civodul>it's propaganda as well, but the right one ;-)
<civodul>"diffusive" is more neutral
<Phlogistique>rekado_: (ianal) it's ok iff they provide a way to replace the LGPL code in the final executable with your own version
<Phlogistique>(and ofc provide the source to their version of the LGPL code)
<Phlogistique>both "infections" and "liberating" give the wrong idea that the licence is somehow able to move on its own and cover other programs
<Phlogistique>whereas the reality is that developpers chose to accept a licence when they are using a GPL or LGPL product to build their own product
<rekado_>What I intended to say was, however, that I'm sad that the license of the code they include does not override their ridiculously restrictive terms for the combined work. I see the restrictive license as the disease, not the virus delivering the cure.
<rekado_>(In general I think that bacteria and viruses have a needlessly terrible reputation.)
<rekado_>what I find somewhat disturbing about this is that the license isn't even published. I wonder if the people who have contributed to this piece of software are aware of its license.
<rekado_>I usually assume "all rights reserved" unless shown otherwise, but the people I've met seem to think "it's on github/bitbucket, so it's 'open source'".
<civodul>which is wrong, legally
<civodul>but yeah, github encourages this
<rekado_>yes; but I suppose the term "open source" is vague enough to make it seem obvious that what is published is "open source".
<rekado_>I don't like these terms that prevent thinking about concepts clearly.
<rekado_>anyway, I have lost all interest in this piece of software.
<civodul>what university was that, if that's something you can say?
<civodul>iyzsong: i'm taking the $GUIX_PROFILE approach in the end
<civodul>the least bad, it seems
<rekado_>civodul: I was wrong. It's actually a division of a publicly funded health care provider in British Columbia, not a university.
<iyzsong>civodul: ah, ok ;-)
<rekado_>I can pass "--listen=some-socket" to guix-daemon, but how can I tell the guix client to use the specified socket file instead of the one in localstatedir?
<civodul>rekado_: oh, ok
<civodul>rekado_: clients honor $GUIX_DAEMON_SOCKET
<civodul>see (guix store)
<civodul>fun: while running 'git pull' on Nix: "GitHub is offline for maintenance. See for more info."
<civodul>we should invent distributed VCSs
<Juhani>git is THE distributed version control system
<civodul>that was ironic
<rekado_>recently an acquaintance was complaining about not being able to create private git repositories for personal use. (In their mind git = github.)
<mthl>Hello Guix!
<civodul>hi, mthl!
<civodul>GCC 5 uncovers a bug in the daemon:
<civodul>(C++ is horrible)
<mthl>civodul: Have you looked at ?
<civodul>mthl: no i hadn't, but that looks good
<civodul>i think davexunit was looking at it too
<civodul>hello davexunit :-)
<civodul>mthl was showing an initial conversion of one of the web pages:
<davexunit>civodul, mthl: nice!
<davexunit>looking good!
<davexunit>mthl: feel free to carry on, I haven't done any work on this.
<phant0mas>I think it's time I start making modifications in make-bootstrap.scm to try to produce the hurd bootstrap-tarballs
<mthl>have you any opinion on the indentation?
<davexunit>mthl: looks fine to me.
<davexunit>does anyone know how to deal with cmake scripts that search in hardcoded directories for headers and libs?
<davexunit>is doing that
<davexunit>and preventing me from building some software
<civodul>mthl: yeah, whatever pretty-print or emacs produces
<civodul>davexunit: i don't know, but probably there are other packages already doing that?
<civodul>mthl: ideally you could make one module per web page
<civodul>like: (define-module (www) #:export (main-page)) (define main-page '(html ...))
<civodul>and then similarly for (www download)
<civodul>that would be great :-)
<davexunit>civodul: I haven't seen anyone else deal with it.
<davexunit>ooh got something to work, maybe.
<civodul>davexunit: i suspect iyzsong would know the answer
<iyzsong>oops, you're mean the FindX11.cmake from cmake package. Nope, I don't know how it works..
<davexunit>passing -DX11_FOUND=1 to cmake has helped somewhat
<davexunit>along with another flag that the FindX11 script sets
<davexunit>yay I got it!
<davexunit>though, to enjoy this program I just built, I need to fix mesa...
<vmlinuz88>davexunit Are there any suggestions/tips for running GuixSD in a vm?
<davexunit>vmlinuz88: I can't think of anything specific to that
<davexunit>I use qemu/kvm and generate my VMs with 'guix system vm' for quick tests (temp fs, shared store)
<vmlinuz88>Ah, I see.
<vmlinuz88>davexunit Where is the vm template file located?
<vmlinuz88>ah, never mind, found it
<cehteh>a 'guix locate ... ' would be nice somtimes :D
<davexunit>what would that do?
<cehteh>solve all problems :) .. DWIM
<cehteh>really i am not sure yet, but locating things in the store, find the package definitions etc
<cehteh>prolly a 'guix apropos' would be more approbiate. as being new to guix i have some problems to figure out where things are defined
<davexunit>'guix package' tells you where package definitions are
<davexunit>for example, 'guix package -A emacs' will give source locations for all emacs packages
<vmlinuz88>davexunit: what is the difference between gnu-system.scm and guix.scm in the build-aux/hydra/ directory?
<cehteh>havent played with guix these days, setting up my quadcopter :D
<cehteh>will be back on guix in a few days
<davexunit>vmlinuz88: I don't know, those are used for our build farm, it seems.
<davexunit>seems that guix.scm is for the guix itself, while gnu-system.scm is for the packages
<mark_weaver>gnu-system.scm is used to describe to hydra what jobs to build for
<mark_weaver>guix.scm is used to describe what jobs to build for
<davexunit>thanks mark_weaver
<mark_weaver>although I've never seen a successful build for and I seem to have forgotten its purpose
<civodul>mark_weaver: it's to do a 'make distcheck' of Guix itself
<mark_weaver>ah, okay.
<mark_weaver>well, I've never seen it succeed, but I suppose I rarely look :)
<vmlinuz88>what file do I use for a guix system disk-image argument?
<davexunit>vmlinuz88: you write one.
*mark_weaver wonders what broke the test on mips.
<vmlinuz88>ahh, nice
<davexunit>vmlinuz88: neither of those files you were wondering about have anything to do with OS configs.
<vmlinuz88>davexunit I figured as much, based on what was mentioned about them above.
<vmlinuz88>could something like configuration-template.scm be used?
<vmlinuz88>I'm on my Fedora host system, not on GuixSD
<davexunit>vmlinuz88: there's build-aux/hydra/demo-os.scm
<davexunit>to use as an example
<vmlinuz88>Ah, okay cool
<davexunit>the manual also has configuration examples
<civodul>mark_weaver: could it be a non-determinstic issue?
*civodul hopes not...
<mark_weaver>civodul: it's possible
<mark_weaver>civodul: it would be very helpful if the relevant test log(s) were printed when 'make check' fails, especially on MIPS where we have so little available build capacity.
<mark_weaver>it's over an hour of build time wasted, which now has to be repeated to recover the test log
<mark_weaver>and it seems fairly common to have guix test failures on non-intel platforms
<mark_weaver>anyway, I have to go afk for most of the day...
<civodul>it's hard for me to tell if it's "fairly common" if there are no reports, hint hint
<mark_weaver>heh, well, I've chosen to just work on fixing the bugs rather than report a bug that no one else is likely to work on
<civodul>heh :-)
<mark_weaver>and part of the issue is that by the time I know enough details to file a report, I already know how to fix the bug.
<mark_weaver>most of the work is rebuilding guix to recover the test log
<mark_weaver>and I've turned every decent MIPS box I have access to into a hydra build slave
<mark_weaver>anyway, I have to go. happy hacking, all!
<vmlinuz88>davexunit: when building a disk-image, I get an error that says "Could not access KVM kernel module: Permission denied."
<davexunit>vmlinuz88: you need permissions to access /dev/kvm
<davexunit>on your system
<davexunit>as a hack, a lot of us just 'chmod 666 /dev/kvm'
<vmlinuz88>Okay, thank you
<civodul>yeah i think the manual has a footnote on that topic ;-)
<civodul>i've just pushed the thing to build the profile's profile file (sic)
<civodul>comments & testing welcome!
<mthl>"Make check" doesn't succeed on my machine
<taylanub>mthl: could you paste the contents of profiles.log?
<mthl>of course
<taylanub>mthl: which guix version is that?
<taylanub>ah never mind, it's a newly added test
<taylanub>(I was confused because I thought I have the latest master and couldn't find the test)
<taylanub>mthl: when you open the file tests/profiles.scm and change the string "source " to ". " it will probably succeed
<taylanub>I'm preparing a patch, assuming that is the bug. (it could also be that the test should explicitly use GNU Bash instead of any /bin/sh, I can't tell right now.)
<mthl>taylanub: thanks for looking into it!
<taylanub>np :)
<taylanub>bugs are there to be fixed
<mthl>"make check" is running again but the test fails with ". " too
<taylanub>can you paste the log again?
<mthl>i annotate the previous one with only the interesting part
<taylanub>can you paste the make check output too? (where it was saying "/bin/sh: 1: source: not found")
<paroneayea>forgot to add #guix to my autojoins, corrected :)
<paroneayea>hello #guix!
<davexunit>hey paroneayea
*taylanub tries 'make check' with no ~/.guix-profile in any *PATH variables and /bin/sh set to dash
<taylanub>aha, I can reproduce the failure
<taylanub>ah, found the problem: dash's output for set(1) is like FOO='value' with the single quotes
<phant0mas>civodul: added a macro in base.scm so make-bootstrap.scm will choose the right glibc based on target system
<phant0mas>and made sure it works with everything else :P
<phant0mas>now I am working on actually making guix build --target=i686-pc-gnu bootstrap-tarballs to produce the tarballs
<civodul>taylanub: maybe that test should just use Bash, actually
<civodul>the syntax of the profile's etc/profile is not POSIX Bourne shell
<civodul>it's rather Bash, but zsh supports it too
<civodul>phant0mas: yay, sounds good!
<civodul>the macro to select the right libc is a good idea
<civodul>davexunit: i noticed that nix-shell has a -p option, which is like what you proposed before:
<taylanub>civodul: hm, it actually worked with dash after fixing a second bashism. by the way there was also a third problem that was sh/bash-unrelated; I just sent a patch fixing all three.
<civodul>ok, well even better if we can get it to work with dash!
<phant0mas>civodul: even though "guix build gnumach-headers" works just fine to produce the cross-toolchain, when I try "guix build --target=i686-pc-gnu gnumach-headers" to build them using that toolchain it fails at configure
<civodul>phant0mas: probably due to "--build=i686-pc-gnu"
<civodul>"--build=i686-pc-gnu" means "the build machine runs GNU/Hurd", which is not the case
<civodul>so here maybe configure thinks it's not cross-compiling because host = build
<phant0mas>civodul: that's exactly what happens
<civodul>so maybe you can conditionalize the --build hack on whether it's being cross-compiled
<civodul>like (if (%current-target-system) '() '("--build=i686-pc-gnu"))
<phant0mas>btw tomorrow I will report to both mailing list what I did this week, so we can keep everyone up to date and happy :-)
<civodul>cool! :-)
*davexunit attempts to fix nouveau in mesa package
<davexunit>can't play any video games on guixsd without working 3d accel
<davexunit>I think I got it. :)
<davexunit>yay that worked.
<davexunit>the mesa package got cleaned up quite a bit and made this easy. :)
<paroneayea>whoo nouveau!
<davexunit>minetest runs at an acceptable speed now. ;)
<davexunit>civodul: a change to mesa will affect a good number of packages, I think. should I push the fix to master or core-updates?
<davexunit>the patch is trivial.
<mark_weaver>davexunit: oooh, nice! I think it should be built on a separate branch
<bavier>davexunit: 'guix refresh -l mesa' says 206 packages
*davexunit always forgets about 'guix refresh'
<mark_weaver>guix refresh is very unreliable though.
<bavier>mark_weaver: in this case it should be fine
<davexunit>I could make a 'mesa-nouveau' branch.
<bavier>mark_weaver: I've done a bit of work to increase the reliability, but it's still too slow to be acceptable IMO
<mark_weaver>it doesn't take into account any implicit inputs, in this case the gtk+ implicit input added the gtk/glib build system
<mark_weaver>I remember that mesa causes a lot of rebuilding. imo it should be done on another branch
<davexunit>yeah, anything that uses OpenGL needs to be rebuilt now.
<bavier>mark_weaver: I see, I didn't know about gtk+
*mark_weaver goes afk
<davexunit>so if I make the branch, will someone else do whatever needs to be done such that hydra builds it?
<mark_weaver>yes, I'll do it
<davexunit>mark_weaver: done
<mark_weaver>davexunit: okay, jobset created:
<davexunit>thank you!
<davexunit>this means I should have webgl support in icecat now :)
<davexunit>I'm not sure what to do, but I need to figure out a way to get a more up-to-date mozilla browser.