IRC channel logs
2013-11-15.log
back to list of logs
<civodul>dmd can reboot, power-off, and all that, woo! <civodul>yeah, not really a distinguishing feature tho *gzg hates when sites default to non-static binary tarballs. <gzg>I hate that I can't seemingly have my expression grab from Slim's actual website and flee to sourceforge for this. <civodul>yeah, some packages have this sort of problem <jxself>civodul: Can it do those with a delay specified? <civodul>there's no stand-alone 'shutdown' command yet <jxself>I could just do sleep 2h first though, so not really an important thin probbably. <civodul>now it's just 'deco stop dmd', 'deco power-off dmd', etc. <civodul>well it'd be cool to have a real 'shutdown' command like the one people are used to <gzg>civodul: What does "deco" even stand for? :^P <jxself>What is the diff between 'stop' and 'power-off'? The later is clear but is stop supposed to reboot? Seems an off thing to call it. <gzg>Oh yeah, Slim was the expression I got this little spew "CMake Error at /nix/store/b2gji5mkk6wc4qkdi4bw8ssd5nmnl7s5-cmake-2.8.12/share/cmake-2.8/Modules/FindX11.cmake:427 (message): Could not find X11" and was looking for the right configure flag. It's kind-of coming back to me! <civodul>jxself: 'deco stop dmd' stops dmd; before it would panic, now it reboots <jxself>Ah, dmd reboots - not the machine rebooting, I see. <civodul>it's PID 1, so when it quits, the kernel panics <civodul>so instead i chose to make it reboot, which makes more sense than quitting when we're PID 1 <gzg>civodul: Hey! Almost finished with SLiM, but Guix doesn't seem to automatically copy slim's conf file where it needs to be (currently attempts /etc/slim.conf.). Also, started on Bluez. <civodul>gzg: perhaps /etc/slim.conf is hardcoded somewhere? <jmd>I wonder if the guix build environment, should be revealed by config.guess ? <jmd>This way, guix built packages would have a build triplet like i686-pc-guix <gzg>civodul: What do we do if something is dual-licensed? Libical states "You may choose the terms of either: * The Mozilla Public License (MPL) v1.0 or * The GNU Library General Public License (LGPL) v2.1". <gzg>civodul: Also, I'll look into where slim.conf's export is set, probably sometime later today. <civodul>jmd: no, because it's just normal linux-gnu <civodul>gzg: you could have a list of licenses <gzg>civodul: Just literally something like (license `(lgpl2.1 mpl2.0))? <civodul>gzg: (license (list lgpl2.1 mpl2.0)) <civodul>i think there are already occurrences of that somewhere <civodul>also add a comment to say that it's dual-licensed, at the user's discretion <gzg>civodul: Okay, it's nearly 04:00 here, heading to bed -- I'll likely pop in-here later today and try to finish off slim. Peace. o/ <jmd>So far as I can tell, guix-daemon doesn't actually do what daemons should (the way that Stevens describes). Would a patch be accepted? <civodul>what doesn't it do that it should do? <jmd>It doesn't fork, detatch itself from the terminal, become its session leader, cd to / ... <jmd>At least I don't think it does - I haven't looked in detail. <jmd>It depends how people are going to use it. <civodul>it's going to be started by the init system <jmd>It might help me find out why mine keeps crashing. <civodul>feel free to email bug-guix@ with all the details if you want <civodul>so 'make check' didn't pass i suppose? <jmd>I think it did. I can't be absolutely sure that I ran make check <viric>Where I worked previously, I used core dumps often... and I was laughted at. <viric>as if I used tech from the 70s <viric>the new tech was to start all again in gdb, and tell the people to reproduce the problem. <jmd>What did they use? Crstall balls. <viric>if they couldn't reproduce it, it wasn't a problem. <jmd>It might take hours to reproduce. Or eternity. <viric>exactly. it means it is not a problem. <civodul>good strategy from the developer's POV <jmd>I know that kind of workplace. <viric>well, some people like to develop features, and don't like fixing bugs. <viric>I like most fixing bugs :) but I was quite alone in that. <jmd>I once spent 3 months developing a whole new suite of test procedures in responce to concerns about quality. But they were soon abandoned because "too many products failed compared to when using the old procedures". <viric>exactly. It's like the frustration caused by phredom's system <viric>jmd: or the famous thousands of segfaults found in all debian programs <jmd>Debian is not that bad! <viric>that had to become thousands of bug reports <jmd>I thought my daemon was crashing. I was wrong. It is exiting quite normally. Why should that be? <mark_weaver>civodul: fyi, jmd reports that when he tries to build his pspp with guix, he sees "guix package: error: Connection reset by peer" and the daemon exits normally (without crashing). *civodul struggles with gcc in core-updates <civodul>maybe i'll commit what i have so far <civodul>problem is it takes several hours to check if i got it right <civodul>probably i'm not being very smart either <mark_weaver>at some point, it might be worth working on improving the debugging experience, somehow. <mark_weaver>e.g., you could set up builds not in a chroot, but in the normal environment where everything is installed, and where you could modify a few files and then type "make" to rebuild more efficiently... get that working, and then do the pure builds again from scratch. <mark_weaver>also, I wonder if it would be possible to set up ccache for use with guix, at least for "impure" builds during debugging. <civodul>what i typically do in non-trivial cases is build with --keep-failed, and then hack from there <civodul>does ccache have any benefit on an SSD? <mark_weaver>I would expect so, because it eliminates the compilation step entirely if the same thing is compiled more than once. <mark_weaver>I guess it works by looking up the preprocessed input in a hash table, or something along those lines. <mark_weaver>If we could convince ourselves that it doesn't introduce impurities, I suspect it could improve the performance of guix builds quite significantly, although I guess it gets dicey when you're debugging the toolchain builds itself. <mark_weaver>(obviously we need to preserve purity above all else, especially the importance of our security goals) <civodul>pthread_cancel & libgcc_s.so, see? :-) <civodul>with GCC 4.7 our "lib" spec adds gcc's libs to the RUNPATH <civodul>actually i was trying to make it smarter <mark_weaver>maybe you should hold off on making it smarter until you get the basics working? dunno. <zacts>civodul: how long have you hacked scheme? <zacts>oh, cool. what would you recommend for a newbie? <zacts>I want to eventually work on guile scheme projects, and learn comp sci at the same time. <mark_weaver>SICP is what created my love for scheme in the first place, around a dozen years ago I think. <zacts>I'm working on remembering all of my math from highschool for SICP <civodul>i had a very "bottom-up" approach to Scheme <civodul>i came to it from the GNU extensibility idea <civodul>and then discovered the beauty of the thing <mark_weaver>zacts: the little schemer, the seasoned schemer, and the reasoned schemer are good books. <civodul>one of the few computer books i have <civodul>but it's not about getting started with Scheme <zacts>I remember how to do geometric proofs, but not mathematical induction <zacts>I can remember only the basics of complex numbers <zacts>I'm just getting started in school again, college. <zacts>I took a college math class a few years ago, I have to review everything because I forgot it all. <mark_weaver>well, you don't need to know much about complex numbers to read SICP, but it certainly is true that it assumes a reasonably strong math background, and yet the math isn't needed to learn programming. <mark_weaver>(it's just that SICP was aimed at MIT engineering students) <zacts>I will be taking more math classes soon. actually, that will be my main study point for a while is mathematics <mark_weaver>well, you don't need to know much about complex numbers to read SICP, but it certainly is true that it assumes a reasonably strong math background, and yet the math isn't needed to learn programming. <mark_weaver>(it's just that SICP was aimed at MIT engineering students) <zacts>maybe in a year or so I'll be ready for it <zacts>but that is a lot of time in between that I could be learning scheme stuff <zacts>I own the little schemer books, and need to read them. <zacts>but I was thinking of starting with HTdP <zacts>then apply it to real projects like guix / guilemacs / etc.. <mark_weaver>HTdP is probably a good way to learn programming, but the language it teaches (racket) has diverged quite a bit from standard scheme, for better or worse. <zacts>I don't know what to do then.. HTdP would probably be a good start, then I would eventually want to learn standard scheme. <zacts>the current scheme standard isn't large is it? <zacts>I could possibly take what I learn from HTdP and read the scheme reference and standard? <zacts>I think I'll start with that. I will then have the goal of reading SICP in about 1 - 1 1/2 years from now <zacts>(after which I will have completed several college math courses) <zacts>for now, I'll just add simple packages to guix <zacts>also coursera offers an HTDP course, I'll do that this next semester. <zacts>civodul: mark_weaver do you know other languages besides (scheme)? <zacts>and what other projects have you worked on? <mark_weaver>zacts: I'd love to talk about this, but right now is not a good time. <mark_weaver>I have just a few minutes before I have to leave, and I'm hoping to have civodul's attention (on #guile) for a few more minutes if possible :) <zacts>and maybe I should transfer this to #scheme