IRC channel logs

2015-10-28.log

back to list of logs

<civodul>ACTION watched https://media.ccc.de/v/camp2015-6657-how_to_make_your_software_build_reproducibly
<civodul>at 51mn, interesting Q&A about Nix/Guix
<efraim>sneek: tell civodul later there is a git-annex repo on github with a bunch of conference videos, which is great for perusing and downloading to watch later https://github.com/RichiH/conference_proceedings
<sneek>civodul, efraim says: later there is a git-annex repo on github with a bunch of conference videos, which is great for perusing and downloading to watch later https://github.com/RichiH/conference_proceedings
<efraim>gah, have to figure out sneek
<efraim>sneek help
<efraim>well that actually worked well
<efraim>sneek: later ask civodul there is a git-annex repo on github with a bunch of conference videos, which is great for perusing and downloading to watch later https://github.com/RichiH/conference_proceedings
<sneek>Okay.
<efraim>sneek botsnack
<sneek>:)
<civodul>Hello Guix!
<sneek>civodul, you have 1 message.
<sneek>civodul, efraim says: there is a git-annex repo on github with a bunch of conference videos, which is great for perusing and downloading to watch later https://github.com/RichiH/conference_proceedings
<civodul>nice
<civodul>we don't have git-annex though :-)
<wingo>i have guix on nixos on one machine. it is carping about locales. what do i need to do to fix it?
<wingo>i think it's the glibc 2.22-on-2.21 thing
<wingo>and should guix 'fix' this by itself, setting GUIX_LOCALEPATH or whatever?
<civodul>GUIX_LOCPATH
<civodul>Guix cannot define environment variables for you though
<wingo>why wouldn't that var be present in ~/.guix-profile/etc/profile?
<civodul>it's only if both glibc and glibc-*locales are installed in the profile
<civodul>which is admittedly suboptimal, but that's tricky to fix
<wingo>i guess i just don't know the fix and i feel that it will become a FAQ and should end up in the documentation.
<wingo>heh, i seem to have my "whiny demanding user" persona turned up to 11, excuse my rudeness ;-)
<civodul>haha :-)
<civodul>np
<civodul>it's already documented but yeah, it's not as smooth as we could hope for
<wingo>ah somehow i missed that!
<wingo>i am doing well at "clueless user" too :)
<civodul>in retaliation, i filed a bug for stexi->shtml :-)
<civodul> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=21772
<wingo>:)
<rekado>I wonder if I'm doing something wrong, but Python setuptools is weird. Can't seem to fix pbtranscript-tofu and that I need to replace setuptools with distutils.core in other packages is troubling.
<civodul>what's the problem with pbtranscript-tofu?
<rekado>I get this unhelpful error: "AttributeError: Extension instance has no attribute '_convert_pyx_sources_to_lang'"
<rekado>this relates to install_requires=[...]
<rekado>in setup.py
<civodul>ouch
<rekado>it contains a list of dependencies.
<civodul>yes
<civodul>i've seen install_requires, but never saw that error
<rekado>but it seems that as part of "build_ext" these dependencies are attempted to be built.
<civodul>built or downloaded?
<civodul>what i've seen is that it would attempt to download them
<rekado>it does not try to download them.
<civodul>ok
<rekado>I see "running build_ext" and then a backtrace showing that "self.build_extension(ext)" somehow failed.
<rekado>and then the AttributeError.
<civodul>presumably it tries to build a bundled version because the system's one is not found?
<rekado>there are no bundled dependencies.
<rekado>I'm playing with the inputs now. Maybe there's something wrong with them instead.
<rekado>python2-bx-python doesn't seem to work :(
<civodul>according to http://hydra.gnu.org/job/gnu/master/pbtranscript-tofu-0.4.1.c7bbd5472.x86_64-linux it broke between commits 4179f9 and e15565
<rekado>I suspect 62a9a23bf9743a1078c4fd860edc3bbdbebde034
<rekado>this was a version bump from 12.1 to 18.3.1.
<rekado>I'm going through the changelog now.
<tyrion-mx>does guixsd support encrypted partitions yet?
<sneek>tyrion-mx, you have 1 message.
<sneek>tyrion-mx, mark_weaver says: your home directory is probably on the root partition, obscured by the mounted filesystem. you should copy the dot files from your home directory as created by guix. you can get to your hidden home directory by bind-mounting the root filesystem somewhere else, using a command like "mount --bind / /rootfs"
<rekado>pbtranscript-tofu can be built with python2-setuptools at version 12.1 (I added a variant locally).
<rekado>I'll try to identify with which version number things broke.
<rekado>setuptools 17.1.1 is the last working version; 18.0 is the first to break. It's not fixed in 18.4 (latest release).
<rekado>fixed!
<petter_>tyrion-mx, yes.
<tyrion-mx>petter_, also root? :O
<petter_>yes
<petter_>if your boot loader is up for it, you can have /boot encrypted as well
<tyrion-mx>petter_, I currently have that setup with grub and debian
<tyrion-mx>this is super nice news
<tyrion-mx>petter_, network manager is not there yet, right? (I did not see it in the packages list)
<davexunit>nope
<civodul>there's a branch for Network Manager though
<civodul>we should Do Something about it
<avoine>Hello, I was looking for the documentation of the --system and --load kernel arguments to change the root and boot path at boot time but I couldn't found it.
<avoine> Anyone knows if it's specific to guix/libre-linux or if it's a undocumented feature of the linux kernel?
<civodul>avoine: these are arguments passed to GuixSD's boot scripts
<civodul>they're indeed undocumented
<civodul>normally you don't have to touch them, though :-)
<civodul>what are you trying to do?
<avoine>I'm just trying to understand how things works
<avoine>when you say GuixSD's boot scripts your taking about dmd?
<civodul>there's a boot script that initializes a couple of things and then spawns dmd
<avoine>ok
<civodul>if you're on master, this is boot-service-type in the (gnu services) module
<avoine>ok, I'll check that out
<civodul>in (gnu system) there's an 'operating-system-boot-script' procedure that returns the boot script of the given OS
<civodul>now you know where to look for these things :-)
<civodul>ACTION sends tarball to the Translation Project
<rekado>hmm, GCJ passes its tests on MIPS, but building ant fails there with bogus errors.
<rekado>GCJ fails tests on arm.
<rekado>ran ",enter-store-monad" in the REPL, got this error: "ERROR: Throw to key `srfi-34' with args `(#<condition &message [message: "guile-arm-fixes.patch: patch not found"] 35cfa00>)'."
<civodul>ACTION pictures a desperate rekado 
<davexunit>:(
<civodul>rekado: is this from ./pre-inst-env?
<rekado>no, it's with geiser (C-c C-z in a Scheme/Guix buffer)
<civodul>in a "Guix REPL" or "Guile REPL"?
<rekado>oh, a Guile REPL.
<civodul>that should work
<civodul>could you check the value of (%patch-path)?
<rekado>unbound.
<civodul>from (gnu packages)
<rekado>I don't know what modules I have to load in the REPL before I can use it.
<civodul>just (gnu) and (guix), roughly
<rekado>(use-modules (gnu)) fails with a different "patch not found" error.
<rekado>(use-modules (guix)) works.
<rekado>oh well. I take a look at this tomorrow.
<rekado>ACTION leaves
<civodul>weird
<aeva>a
<mark_weaver>rekado: I chown'd that directory on hydra-slave0. sorry for the delay.
<mark_weaver>I also chown'd the icedtea6 failed build directory while I was at it
<davexunit>similar to my "ruby on guix" article, there's a new "nix for python developers" article: http://datakurre.pandala.org/2015/10/nix-for-python-developers.html
<rekado>mark_weaver: thank you!
<efraim>adding that to the read queue
<sprang>davexunit: your hacker news posts are probably getting penalized by the voting ring detector: https://news.ycombinator.com/item?id=7543910
<davexunit>sprang: that must be it
<davexunit>every site works differently. slashdot asks that you get friends and such to upvote your post.
<sprang>when you ask for upvotes in here a bunch of people appear out of nowhere and upvote, likely without even following the HN link
<sprang>anyway, just a heads up :)
<davexunit>thanks sprang
<davexunit>99% sure that was the reason now.
<davexunit>maybe someone can re-submit it in a week or so
<davexunit>and see if it gets anywhere
<civodul>interesting
<civodul>so we must stop bothering, basically, and wait until people upvote
<civodul>it's actually quite sane
<davexunit>yeah
<civodul>less clicking for us, too ;-)
<davexunit>;)
<civodul>davexunit: interesting Nix/Python post
<davexunit>civodul: the hashbang thing that nix-shell does is cool
<civodul>yes, nice
<davexunit>we should add that
<civodul>yep!
<civodul>we're evil copiers!
<civodul>:-)
<davexunit>goes both ways :)
<efraim>i think my https-everywhere is interfering with the youtube-dl urls
<efraim>the site is available with and without https, but the certificate isn't actually for youtube-dl
<efraim>DNS Name: *.aries.uberspace.de
<efraim>definately not youtube-dl.org
<civodul>youtube-dl.org is available for me over HTTP over Tor
<efraim>currently in video.scm the origin uri has https and the home-page has http. i'll fix that to http next time I update it
<efraim>interestingly wget fails with https and guix download doesn't complain
<aeva>davexunit I just read your ruby on guix article - would be really handy if there was some way to generate the environment package.scm from an existing gemfile
<aeva>as, I agree with your sentiments against rvm and bundler, and use them at work and would enjoy being able to use guix instead without having to recruit the all of the different tech teams >_<
***atrapa is now known as atrapado
<davexunit>aeva: yes, that would be a useful feature, but tricky to implement.
<davexunit>and it couldn't possibly be done completely automated, either, due to some gems needing non-ruby deps that aren't specified in ruby packages.
<aeva>I think I sorta figured it would take that and determine which ruby packages in guix exist and which don't and error out when something is missing from guix...
<aeva>and if everything is in guix, then it would generate the scm file?
<aeva>ACTION is not terribly familiar with the nuance of either of these systems
<civodul>alezost: what do you think of adding guix-build-log-mode to auto-mode-alist?
<karhunguixi>unfortunate timing there
<civodul>indeed
<civodul>sneek: later tell alezost what do you think of adding guix-build-log-mode to auto-mode-alist?
<sneek>Will do.
<civodul>guix-build-log-mode is really awesome, if you've never tried
<karhunguixi>i haven't; i have actually just barely tried Emacs (i assume you're talking about Emacs)
<civodul>yes
<rekado>would this be used in shell-mode when running "guix build"?
<paroneayea>hello!
<civodul>rekado: rather when opening a log file as returned by "guix build --log-file"
<civodul>you do M-x guix-build-log-mode there
<civodul>then you can fold/unfold things with Tab/M-Tab, etc.
<civodul>cool stuff!
<davexunit>aeva: that could work! good idea.
<efraim>i just noticed that we have xf86-video-geode packaged, and apparently it has major problems with nonroot Xserver since it doesn't use KMS
<civodul>"The GNU project as a whole has the feel of a relic" <- https://news.ycombinator.com/item?id=10444628
<cehteh>gnu is not hipster anymore? ... omg :D
<efraim>bluez 5.35 has a new config option AutoEnable=true so you don't have to mess with hciconfig hci0 up
<efraim>i had a script I used almost 2 years ago when I went commandline only networking and bluetooth tethering was always the big downfall
<civodul>cehteh: it actually is! the commenter just hasn't realized yet
<cehteh>civodul: part being sarcastic, part ignoring this, it is just as it is and everyone lives in his own cave, you, me, the commenter there. so mostly i give a sh*t about such discussions
<cehteh>(besides the free software world would be much more dire if rms wouldnt be that polarizing)
<cehteh>he fights, he takes the blame, and and often enough he wins
<civodul>he's been very consistent and able to clearly formulate his ideas
<cehteh>yes
<cehteh>but his ideas are not everywhere accepted :)
<rekado>yesterday a new version of openblas has been released.
<cehteh>puh .. file server here back online .. at noon it died with scrolling endless filesystem errors on the terminal
<rekado>with the current version (and an additional make flag) I can build it on MIPS, but at least one test fails.
<cehteh>i thought the filesystem was toast, but thanks to errors=remount-ro nothing happend, just a looooong fsck with 2 minor problems
<efraim>python as an input for the tests, needed or no?
<efraim>i built it twice, once with and once without
<efraim>both passed the exact same number of tests
<rekado>what package?
<efraim>bluez
<efraim>tests with python were 120 seconds vs 95 without
<efraim>could python be leaking in from /usr/bin? should I be building in a pure environment?
<efraim>no difference when build in an environment, same output hash as building without python
<efraim>ACTION goes to sleep
***Xe is now known as \\section{Xe}
***\\section{Xe} is now known as Xe