IRC channel logs
2014-11-06.log
back to list of logs
<jxself>It is usually better to just ask your question on IRC. :) <jxself>Lots of people use HURD already. Some use it in QEMU. Some on older hardware. I recall reading that 75% of the packages in Debian main compile & run just fine. <jxself>AFAIK, there is still no SATA support. That doesn't always matter though because some boards have a legacy compatibility mode. <DusXMT>that's not true, there is sata support <DusXMT>it was added around the time of the wheezy release <DusXMT>they get recognised as SCSI discs, so you'll find them in /dev/sd* <toxemicsquire4>I can't install Wheezy on my SATA computer because it can't find the disk drive <DusXMT>Then there are two possibilities: The developers of d-i forgot to take SCSI drives into consideration, or: Your BIOS lacks a feature Mach uses for getting the device list (my netbook had this problem, I don't remember exactny how the feature was called though) <DusXMT>Only some anicent PCMCIA cards ;) <jxself>GNU has two kernels nowadays, you know, so completing work on HURD doesn't seem as important as it once was though. <jxself>Linux-libre was made a GNU package package about 2.5 years ago. <toxemicsquire4>HURD development has skyrocketed for about the last 2 years. It has alot of development compared what it used to have had a couple years ago <jxself>Yes, and that's great IMO. The more GNU kernels the merrier IMO. :) <DusXMT>Note: It's important to note that the Hurd isn't exactly a kernel. It is `A replacement for the UNIX kernel', but it itself isn't a kernel, but a collection of servers that run on top of a microkernel (currently only Mach) to provide POSIXy APIs, and the translator API of course <toxemicsquire4>I know that, but I'd rather call it just HURD/Mach kernel that HURD multi-server on top of Mach microkernel <DusXMT>Simply saying `The Hurd' does the job, imho, no need to bring confusing terms :) <toxemicsquire4>Yea pretty much. Do you ever think it will be "done", and by that I mean being able to be considered stable enough for end user use <jxself>Calling it "done" sounds like a lot of work. The computer industry isn't standing still. <jxself>But it's nice to hope that one day it might be. :) <jxself>At the same time, though, we have a free kernel already - Linux-libre - so it takes off some of the urgency IMO. <jxself>I can imagine. HURD has some interesting technology. <jxself>But we can get to the goal of free operating system, and that's something. :) <jxself>Time to me to get to bed though. ZZZzzz... <civodul>alezost: in the package-info buffer, it would be nice to have a "view source" button :-) <civodul>that would use package-source-derivation, and then drop you in a dired buffer in the source tarball/directory <alezost>civodul: thanks for the idea! I'll look at it. IIUC the source is available only if a package was built locally, isn't it? <civodul>alezost: no it's always available; the command-line equivalent is "guix build -S" <civodul>often it's a tarball, sometimes it's a directory (git checkout, etc.) <alezost>civodul: oh, great! I didn't know about that <civodul>sneek: later tell ph4nt0mas try running "make clean-go && make", to account for the recent ABI change *cyril_ managed to talk about Guix while at the OpenStack summit \\o/ <civodul>did they have to stop and repurpose the conference or something? <cyril_>no, they just made sure I left the premises immediatly <cyril_>I have a coleague who's got the weirdest system <cyril_>he's only got a very tiny installation of Gentoo <cyril_>and every single process is run in a VM <cyril_>so he's got tons of qemu running <cyril_>but they only take a small amount of CPU <civodul>VMs are the outcome of poor OS design <cyril_>I mean, I don't care, it's just so funny <civodul>the "Scratchbox" cross-compilation environment did (does?) funny stuff with emulators <cyril_>and he said "well, the 'real' system only needs to upgrade qemu" <cyril_>and if another program suffers from a bad upgrade, I can just use a snapshot to "downgrade" <cyril_>so I told him that the right design was to have rollbacks as part of your package manager <cyril_>Lots of people are talking about Docker here, and I'm wondering whether some of the problems it is supposed to solve could not be adressd by Guix as well <cyril_>though I don't even understand what problems it solves exactly :) <civodul>it's not completely clear to me, but it seems to be lower-level <civodul>however, Docker has hype, which has brought millions of bucks <cyril_>it seems mostly sysadmin-oriented <cyril_>but it seems to make it "easy" to dsitribute applications <cyril_>mostly because most people cant use package managers properly :D <civodul>and it's really the thing where "applications" may well mean "proprietary blob" <cyril_>you know, I'm surrounded by people who "love open-source" <cyril_>but also, we have lots of "package managers" that are language-specific <cyril_>because your application works with these versions of library foo, and these versions of library bar <cyril_>and since it's a pain to have an up-to-date package manager, yo uend up using pip <civodul>yeah, it boils down to each community wanting to have more control over the distribution workflow <cyril_>so with docker, you can create an "image" that will "just work" onyour machine <cyril_>though it sucks, because I want to use what my distribution provides <cyril_>because I know that Debian/Guix only provide Free Software <cyril_>but I do not trust other sources <cyril_>btw, do we have this guix-shell thingy yet ? <cyril_>or any utility that would allow users to create a new profile and "source" it <cyril_>because in Python, we have those "requirements" files <cyril_>with a list of requirements and a range of acceptable versions <cyril_>and we end up creating "virtual environments" that use pip to recreate /usr/lib/python2.7/ <cyril_>it'd be fun to have a Guix command that does something similar <civodul>davexunit would know better, but i think 'guix environment' is a reasonable replacement <cyril_>that is, creating a new profile used for testing the application <cyril_>the worst thing is that these "requirements" only specify Python libraries, obviously <cyril_>so if you need to run a program/library that is not written in Python so that the app works, well, you have to do it manually <cyril_>but with Guix, you could just specify it and have it installed as well <iyzsong>hi, any way to get guix environment set a shell prompt? <civodul>iyzsong: that's what it does by default, if everything goes well <civodul>just run "guix environment foo", and it starts a shell <civodul>cyril_: npm manages to deal with non-JS dependencies, but it really relies on you having the "right" environment ready, with a C++ compiler and everything <iyzsong>hm, I use zsh with prompt='%# ' in my .zshrc, after guix environment, the shell prompt remains '% ' <civodul>the default is to invoke either $SHELL or /bin/sh <sneek>Welcome back davexunit, you have 3 messages. <sneek>davexunit, ArneBab says: How do you do the matrix multiplication in sly? It’s something I’d like to try, too. <sneek>davexunit, lloda` says: however if you do only 4x4 matrices, those calls may have too much overhead. You should try to multiply bigger matrices. <sneek>the, civodul says: difference <davexunit>I can answer some questions you have about it. <cyril_>civodul: yes, some python packages can also build Python extensions written in C <cyril_>but it won't build MySQL for instance :) <davexunit>you're familiar with nix shell, so maybe you have some improvements in mind :) <iyzsong>davexunit: yeah, I want to get a different shell prompt to ensure I'm in guix's env :) <davexunit>iyzsong: yes, I wasn't quite sure how to handle that. <davexunit>then yes, we should set a special PS1 for guix environment. it would be nice if that could be customized, too. <iyzsong>I think we should avoid spawning a $SHELL, even with --pure, I still get unwanted env with my zsh. <davexunit>hmm, how do we prevent such behavior for all shells? <iyzsong>I see there is --noprofile for bash, maybe we could just use bash for a pure env. <davexunit>it would be a shame to force the pure env to use bash <civodul>we could allow -E to take a command, instead of a file name <davexunit>and it will run that instead of $SHELL or /bin/sh <davexunit>so, it's not automagic, but it will get you what you want, I think. <davexunit>sometimes spawning a shell isn't desired, so -E is there for running one-off scripts in the env and then returning to normal. <iyzsong>guix environment --pure -E "PS1='@ ' /bin/zsh --no-rcs" gcc <davexunit>iyzsong: soon enough, sly will have the equivalent of a default.nix file in the repo, and you'll be able to run `guix environment -l package.scm' to create a dev env <davexunit>just need guile-sdl's test suite to pass, and the dependencies will be complete <davexunit>hmm, I wonder if that's part of the problem. I didn't get the chance to investigate much. <davexunit>I will look into that first when I get a chance to hack on it <iyzsong>I haven't look into tests, but I do admire guix's attitude for testing. <davexunit>yeah, we try to have passing test suites as much as possible. if the package has tests, of course. *davexunit needs to settle on some API decisions and prepare a release <davexunit>I've spent months writing and rewriting scene graph implementations. <civodul>i should play with it one of these days <davexunit>try out a release, when I make one. there will be less rough edges. <davexunit>and it will be easier to try when you can 'guix package -i sly' <civodul>would be wonderful if someone updated Mesa