IRC channel logs

2015-04-06.log

back to list of logs

<mark_weaver>civodul: indeed, it might break some things, but so do many of the other changes we make
<mark_weaver>there are two problems with letting it use 'config.guess'
<mark_weaver>one is that many programs have an old config.guess
<mark_weaver>the other more important problem is that 'config.guess', by design, is based on whatever the build machine happens to be, instead of the baseline that we should be targetting, as least for purposes of binary substitutes
<mark_weaver>s/is based on/returns a triplet based on/
<mark_weaver>fwiw, I think it would be great to provide a way for users to build binaries optimized to their hardware
<mark_weaver>but the binaries produced by hydra should be targetting a chosen baseline based on the system name
<mark_weaver>the other thing is that it's an impurity
<civodul>yes, agreed
<mark_weaver>builds made for armv7 should have a different hash than those for armv8
<mark_weaver>good! :)
<civodul>let's try passing --build
<mark_weaver>okay
<mark_weaver>thanks :)
<civodul>mark_weaver: more precisely ;-), could you send a patch?
<mark_weaver>will do
<civodul>thanks!
<civodul>dunno if that can work for this core-updates
<civodul>maybe that's doable
<mark_weaver>it's not a huge rush, I don't mind waiting for the next cycle on that one
<mark_weaver>if you prefer
<mark_weaver>what potential difficulty are you thinking of, besides the occasional build system that breaks?
<mark_weaver>otoh, it would be nice not to delay the current cycle any further, especially since it would be good to get wingo's gettext patch deployed
<civodul>yes, i'm testing it right now
<civodul>yeah, maybe next cycle is better for --build, then
<mark_weaver>okay
<mark_weaver>sounds good
<mark_weaver>I should probably delete this 'glib-rebuild' branch and jobset
<mark_weaver>at the time, I didn't realize that the gettext change would also be needed
<mark_weaver>wdyt?
<civodul>the gettext change is needed to rebuild glib?
<civodul>then yes, indeed
<mark_weaver>no, but it's needed for the new gtk, which was the reason for updating glib and cairo
<civodul>ah ok
<mark_weaver>should we apply its two patches (update glib and cairo) to core-updates?
<civodul>yes, that's what i was about to suggest
<mark_weaver>(they are the same as wingo's posted patches, but I changed the commit logs to conform to our conventions)
<civodul>ok
<mark_weaver>okay, will do
<zacts>hey mark_weaver have you heard of The UNIX Haters Handbook?
<zacts> http://web.mit.edu/~simsong/www/ugh.pdf
<mark_weaver>yes, I know it, but I haven't read it
<zacts>ah ok, cool
<mark_weaver>I've hung out with Simpson a few times, although it was a long time ago
<zacts>I'm finding it to actually be an interesting book
<zacts>oh neat
<ijp>zacts: physical copies come with a barf bag
<zacts>ijp: really? :-D
<ijp>really
<mark_weaver>We watched "The Matrix" together, when it first came out
<zacts>heh
<mark_weaver>Simson I mean
<zacts>yeah
<zacts>mark_weaver: you and bart simpson were the best of friends... hehe
<mark_weaver>hehe :)
*zacts is reading the info to dmd
<kete>zacts, what kind of systems do they like?
<zacts>kete: they mention lispy systems
<kete>oh wow
<davexunit>zacts: this book is pretty entertaining
<davexunit>some things never change
<davexunit>"You know how a LispM is always jumping into that awful, hairy debugger with the confusing backtrace display, and expecting you to tell it how to proceed? Well, Suns ALWAYS know how to proceed. They dump a core file and kill the offending process."
<davexunit>reading this makes me sad in some ways.
<davexunit>programmers have become more accepting of dynamic languages in recent years, but there are still many features that some people were used to in the 80s on LispMs that still haven't become commonplace
<davexunit>like incremental recompilation. We're used to it in our Lisp implementations, but it's basically unheard of in Ruby/Python/etc.
<davexunit>the Ruby web applications I've worked on can re-evaluate entire source files when I save them to disk, but that's a really weak form of incremental recompilation.
*mark_weaver works on updating nss and icecat
<mark_weaver>I'm rather surprised to see no digital signature from Mozilla on the nss tarball
<mark_weaver>this thing contains the CA trust store, so it rather security sensitive
<mark_weaver>am I missing something?
<mark_weaver>it just seems ... incomprehensible to me that the upstream developers wouldn't sign nss
<mark_weaver>it seems that Ruben can't decide whether Icecat will put its user settings in ~/.mozilla or ~/.gnu
<mark_weaver>31.4.0 put them in ~/.mozilla, 31.5.0 put them in ~/.gnu, and 31.6.0 puts them back in ~/.mozilla. I have to keep moving my data back and forth to keep my settings.
<mark_weaver>settings / bookmarks / saved sessions
<Sleep_Walker>just create symlink
<mark_weaver>heh, yeah
<mark_weaver>I didn't expect it to change again so soon
*wingo did a guix system build and it succeeded, yay
<mark_weaver>and I think I'll be optimistic and hope that 31.5.0 was an aberation
<mark_weaver>wingo: sweet!
<wingo>wondering tho why guix system reconfigure decides to rebuild inkscape tho (!)
<wingo>i guess for the guix logo svg?
<mark_weaver>wingo: right
<mark_weaver>that bugs me too.
<mark_weaver>we should use librsvg, maybe via Guile's FFI
<wingo>there is guile-rsvg
<mark_weaver>ah, nice!
<wingo>or guile-librsvg, i don't recall which
*wingo uses it
<wingo>it can render to a cairo surface
<mark_weaver>also, the way it's currently done is to convert to a pixmap and then scaling it after, which obviously leads to poorer results than could be achieved
<mark_weaver>see 'grub-background-image' in gnu/system/grub.scm
<mark_weaver>patches welcome! :)
<mark_weaver>wingo: btw, can you send the patch that you needed to fix the circularity problem?
<mark_weaver>I guess that's a prerequisite to the gtk+ update that you already sent
<wingo>mark_weaver, sure
<mark_weaver>thanks!
<wingo>done
<wingo>not much more comments as i have to go
<mark_weaver>okay, bye!
<mark_weaver>and I need to sleep :)
<mark_weaver>I was hoping to find evidence of the ~/.mozilla -> ~/.gnu -> ~/.mozilla changes in the gnuzilla git repo that contains the 'makeicecat' script, but alas it is not there at all. so it must have been something Ruben did by hand that's not in the script at all :-/
<mark_weaver>well, unless I missed something.
<mark_weaver>actually, I guess I probably did miss something. d4913be..: Ruben Rodriguez 2015-03-15 Fixed profile path to .mozilla/icecat
<freaj>mark_weaver: you didn't miss anything
<freaj>Few days ago it was on .gnu/icecat, it moved to .mozilla/icecat :-)
<freaj>(When I updated my icecat, it was a vanilla profile)
<mark_weaver>same here
<freaj>But I enjoyed the .gnu/ folder, that's sad. :P
<bavier>if source lists only "GNU General Public License" as the license, can we pick any version?
<mark_weaver>look in the source files
<mark_weaver>not just the COPYING or README or whatever
<mark_weaver>does it really not give a version number anywhere?
<bavier>mark_weaver: actually, the README is the last place I just looked
<bavier>and it has the "under the same terms as Perl 5" that I was looking for
<bavier>sorry for the noise
<civodul>Hello Guix!
<davexunit>hey civodul
<Sleep_Walker>it seems that default lsof mirror may cause problems with access - http://sprunge.us/CIPN
<Sleep_Walker>do you mind using some other mirror for that?
<civodul>"guix lint -c source lsof" works for me
<Sleep_Walker>author mentions that on his web page - http://people.freebsd.org/~abe/
<civodul>but we could add one of these mirrors to the source
<Sleep_Walker>that would be great
<civodul>can you send a patch for that?
<Sleep_Walker>I'm not on Guix yet and I don't want to send patch without test
<Sleep_Walker>I'm trying your recommendation (btw. thanks!) from the 20037
<civodul>ok, i can do it if you want
<civodul>while i'm at it
<Sleep_Walker>right now I don't know how I could do that
<Steap>can I interactively run code as a guix-builder ?
<Steap># su guix-builder1
<Steap>This account is currently not available.
<Sleep_Walker># su guix-builder1 -c /bin/sh
<Sleep_Walker>maybe
<Sleep_Walker>but better question is - why?
<civodul>Sleep_Walker: ok, i've added the URLs to lsof, thanks
<Sleep_Walker>thanks
<civodul>Steap: that is a strange thing to do though, there's nothing special about this account
<civodul>hey!, BTW ;-)
<bavier>Only about 45-50 packages to go for hydra!
<davexunit>only! :P
<bavier>davexunit: perspective ;)
<davexunit>that's how I feel trying to package mediagoblin
<civodul>bavier: this is craaazy!
<civodul>i'd never thought there were so many deps
<civodul>you're a hero :-)
<bavier>civodul: thanks. I'm not looking forward to updates :P
<civodul>heh
<civodul>we need a generalized 'guix refresh'
<bavier>civodul: definitely
<Steap>wanted to try and debug an issue without relaunching a build
<Steap>the bug might have something to do with permissions, so I would have liked to be able to use the same account
<civodul>Steap: i typically do "sudo chown -R $USER /tmp/nix-build-*" and then work from there
<civodul>and run "source environment-variables"
<civodul>hopefully that'll be enough?
<Steap>hum, not sure
*davexunit would like an "orchestration" system built into guix
<davexunit>to control guix machines across a network :)
***pfo_ is now known as pfo
<davexunit>been reading about the hot new thing from google: http://kubernetes.io/
<civodul>oh
<cmhobbs_>what's a good metric for knowing if guix is stable enough for regular use?
<cmhobbs_>i heard xfce was available now
<cmhobbs_>is it based on the software i use or is it a stability issue currently?
<cmhobbs_>i'm chomping at the bit to give guix a try somewhere other than a vm :D
<civodul>cmhobbs_: GuixSD still has a bunch of limitations and rough edges
<civodul>it wants you to get involved
<civodul>;-)
<davexunit>civodul: thanks for the guix-web patches
<cmhobbs_>i'm using the package manager on my netbook
<civodul>davexunit: you're welcome!
<cmhobbs_>but i'd like to run it as an os
<cmhobbs_>speaking of which, i need to upgrade it on my netbook. i think i'm behind a version or two
<cmhobbs_>i have a desktop i could use that was a home server for a while
<cmhobbs_>but i'm stoked about the potential of guix as an os
<davexunit>civodul: I've done some code changes since FOSDEM, so some of these might not apply, but I'll figure it out.
<davexunit>one big change is that I moved everything into the (guix ...) namespace in anticipation of it being upstreamed
<civodul>nice!
<davexunit>so you can run 'guix web', instead of 'guix-web'
<civodul>cmhobbs_: if you're used to a desktop environment, you may be disappointed
<civodul>many things probably won't work out of the box, etc.
<cmhobbs_>well i read that xfce was available
<cmhobbs_>my needs aren't very many
<davexunit>xfce support is slowly getting better
<civodul>if you've seen wingo's experience report a few days ago, it was pretty difficult
<civodul>(and instructive from our POV)
<davexunit>yeah
<cmhobbs_>xfce, emacs, midori
<cmhobbs_>and a handful of smaller things
<davexunit>cmhobbs_: you could package midori :)
<cmhobbs_>i did not see it
<cmhobbs_>davexunit, i'd have to work in the time for that
<cmhobbs_>at least i'm using guile in a skunkworks project that'll be libre before it's all said and done
<cmhobbs_>might be good guile practice to learn to package, heh
<davexunit>if you a machine to spare, give it a shot.
<davexunit>expect rough edges, and let us know what was pleasant and what really sucked. :)
<davexunit>I really need to get back to work on guix web. I have some concerns about upstreaming it, due to the inclusion of so much javascript and stuff.
<cmhobbs_>i guess i'll have to toss another monitor on my desk. my son is using my only spare
<cmhobbs_>how's sshd support in guix?
<cmhobbs_>would it make sense to install it and run it headless for packaging?
<cmhobbs_>do i need to run guix as an os to package? i have guix as a package manager on my netbook
<cmhobbs_>guix web?
<civodul>cmhobbs_: you can develop Guix packages without using GuixSD
<davexunit>cmhobbs_: I do most of my work from a Debian machine with guix on it.
<davexunit>you may find that things behave differently when run on your host OS vs GuixSD
<davexunit>for example, the Guix minetest build runs fine for me on Debian, but on GuixSD it's messed up because of an OpenGL issue.
<cmhobbs_>yeah, my netbook is running wheezy
<cmhobbs_>ok, back to my original question, is sshd functional and does it make sense to run a guix rig headless?
<cmhobbs_>i could steal my son's monitor
<cmhobbs_>for a couple of hours, heh
<davexunit>cmhobbs_: yes, we have sshd.
<civodul>davexunit: didn't you have patches for core-updates, from a month back or so?
<davexunit>we have lsh and openssh, but only an lsh system service.
<davexunit>no openssh daemon service yet.
<davexunit>civodul: I can't remember :)
<davexunit>libxml stuff?
<davexunit>I never finished that work
<civodul>dunno
<civodul>i just had a vague recollection
<civodul>not sure
<davexunit>too many irons in the fire...
<davexunit>I bounce around between many things to fight burn out.
<davexunit>cmhobbs_: and to answer another earlier question, 'guix web' is a little side project of mine that provides a web client for guix.
<davexunit>so you can do package management from the comfort of your web browser.
<cmhobbs_>look at you bein' all modern
<cmhobbs_>\\o/
<davexunit>yeah, it tweets and posts to facebook when you install stuff.
<civodul>:-)
<davexunit>and it uses tinder's API to match you up with other guix users that use similar software.
<civodul>ah ah
<civodul>i can imagine why Guix geeks would remain lonely ;-)
<davexunit>and, in the future when we have a more distributed build farm, it will call Uber to drive you closer to the nearest collocation facility with a guix server.
<davexunit>for faster downloading of substitutes
<davexunit>okay I'm done. :P
<civodul>davexunit: it's time you create a startup in anticipation of all this :-)
<davexunit>civodul: I've already got the LLC!
<cmhobbs_>don't tell my wife about the tinder part
<cmhobbs_>she might try to hook up with me
<davexunit>haha
<davexunit>there's also iGuix, our iOS client
<davexunit>it will be $2.99 on the app store
<davexunit>or rather, it will be $0, but transactions with guix-daemon require micropayments.
<civodul>1ยข per RPC
<civodul>and it would install stuff on a remote free machine, while preserving you from freedom
<davexunit>yes exactly
<davexunit>that's the next step of the business, really.
<davexunit>it's like Netflix, but for packages
<cmhobbs_>i like where this is going
<davexunit>you pay $10/month, and stream packages to your computer to use
<cmhobbs_>we may just disrupt the paradigm outside the box
<cmhobbs_>thereby moving the needle
<davexunit>back into the haystack
<davexunit>so you get on-demand applications, without the hassle of needing everything installed to a local disk
<cmhobbs_>do you guys have a link to wingo's report? was it here in chat or was it a blog post or something?
<davexunit> https://gnunet.org/bot/log/guix
<davexunit>is the general place for channel logs
<davexunit>just have to find the right day
<davexunit>one sec...
<cmhobbs_>kk
<davexunit> https://gnunet.org/bot/log/guix/2015-04-02
<davexunit>cmhobbs_ ^
<davexunit>probably a lot of text to sift through
<cmhobbs_>thanks!
<davexunit>but the saga starts there
<cmhobbs_>i'll read through it now
<mark_weaver>civodul: fyi, I send the dumpspecs output for the armhf issue
<mark_weaver>*sent
<mark_weaver>if we were able to find the problem in time for this core-updates cycle, that would be great of course
<mark_weaver>(in case the fix would involve a rebuild)
<mark_weaver>and also so I could start a build of core-updates on armhf to find other issues
<davexunit>civodul: have you given any thought to managing multiple guix machines remotely a la NixOps?
<civodul>mark_weaver: ok!
<civodul>davexunit: not really
<civodul>sounds like a logical next step though
*civodul -> zZz
<civodul>good night/day!
***pfo_ is now known as pfo