IRC Logs for #crux-devel Friday, 2011-12-16

*** mike_k has joined #crux-devel02:14
Romsterwas the iso made from frinns_t's 64bit pure?03:07
Romsteror you used the official 32bit? though that would make no sense. unless you rebuilt the official crux iso with 64bit.03:08
juejaeger: thanks, will have a look at your multilib stuff03:42
juefrinnst: btw, do you have an idea how often your iso has been downloaded?03:46
juejaeger: is glibc-32 the only "special" package we need for a working system which is multilib capable?03:50
pitilloand a ready gcc multilib capable if I'm right (those at least)03:55
Romsterthought you also need mpfr gmp mpc binutils to go with that mix.04:04
juesorry, was a bit unclear, I meant "additional" with "special"04:06
Romsterdo't those also need to be 64bit too?04:08
Romsterdon't*04:08
juesure, but I'd like to know how big the impact between pure/multilib x86_64 is04:10
pitilloI think first libgmp should be fixed to use the right triplet, then go ahead with toolchain, letting gcc support multilib and then building glibc for both abis (adapting the 32b version)04:10
Romsterie entire tool chain libs glibc binutils mpfr gmp mpc gcc all need 32bit and 64bit builds to work and going by what jaeger has i see there is also gdbm?04:10
juedo you need all of this if you don't need 32bit?04:11
jueI don't think so ;)04:11
Romsterthought only perl used gdbm04:11
juewell, to be clear04:11
Romsternot really sure to be honest.04:11
Romsteri onyl got so far and jaeger builds the entie thing befor ei've even learned how to get past the first stage of gcc04:12
Romsterthat took me like a days worth of stuffing around to build in total.04:12
Romsterwith all the bugs i kept hitting.04:12
juethe only additionla package on the jaeger's ISO is glic-32, so I'd say nothing more is needed for a working system04:13
Romsternever mind i said anything i clearly don't understand this enough04:13
jueof course, if you _need_ 32-bit support you have to install a lot more04:14
Romsterin the case of wine i do.04:14
Romstermight be some other emulators for atari or something that only has 32bit as well. but i haven't hit those yet04:15
juesure, but most people will not04:15
Romstermost wont need 32bit but maybe browser for flash04:15
pitillojaeger will clarify all. For a working system it's only needed glibc32, but if you want to build both abi's you need to fix some toolchain components (here I haven't much experience with multilib support but that's what I understood from what I readed)04:16
Romstercan't think of anything else.04:16
juethe reason why I ask for that is because I'm looking forward how a official CRUX 64 should look04:17
Romsterjaeger, probably dropped gmp mpc mpfr into the gcc source directory and let it bootstrap that way, but didn't provide the 32bit libs of those on the system after... or something along those lines.04:17
Romsteror those libs are now statically linked into gcc multilib, i don't quite understand that part yet.04:18
pitillojue: that sounds really interesting. I hope 64b goes official too. There are a lot of work behind 64b scene (pure/multilib)04:19
juedon't think that gcc needs 32bit versions of that libraries04:19
Romsterthat would also mean existing contrib/opt other major ports xfce etc would need to be on 64bit to be able to test that.04:19
Romstergcc would need 32bit versions to be able to handle 32bit on a 64bit system *shrugs*04:20
Romsteri'll wait for jaeger to clarify04:20
Romsteri'm fine for moving to 64bit and been wanting too for quite some time but wine has been holding me back.04:21
Romsterif it could be pure 64bit and some overlay ports tree for multilib that would be the best of both worlds for those that need some 32bit support.04:21
juehehe, that's exactly the reason for my first question above ;)04:22
Romsterwhich i obviously missed :D lol04:22
pitillothat's the aim of x64 multilib project, isn't it?04:23
Romsteri'm hoping so jaeger's doen all the work, don't know if the little bit of the first stage build commands i let jaeger see helped in any way. but i sure would help maintain wine for 32bit support on 64bit crux.04:24
Romsteraim being able to build needed libs and ports to support some software that wont work in 64bit. but those that don't need it can leave that off there system. i'm hoping that's jaeger's plan.04:25
juehaving 32bit support is for sure a big pain, because you need a lot of additional 32bit packages, but I wouldn't care about that if the basic multilib system is nearly the same as a pure 64bit system04:29
Romstermost of it other than the tool chain can be in there own ports trees as jaeger has it currently.04:30
jueif you need 32bit grafic support you will end up with most libraries doubled04:30
Romsterelse it's going to end up that most use pure 64bit and the minority will be on multilib.04:30
Romstera long time ago i did think of a build64{} function in each lib witht he build{} but that's more complexity that crux doens't need.04:31
Romsterif it comes to that then either myself jaeger or someone will have to provide/maintain those as well.04:32
juewell, as always the aim of CRUX will be to provide that basic04:33
jueif have a 64bit system that can be easily extended to a full blown multilib system that's fine04:34
jueto be clear, the focus has to be "extended" here04:34
Romstersupport to extend to 32bit with other repositories should you require 32bit support ie wine.04:35
Romsterwine will run on 64bit but it will only run 64bit PEs04:36
Romsterand isn't as well tested yet.04:36
Romsterlots of stuff is 32bit PEs still.04:36
frinnstgrep -c crux-2.7.1-x86_64.iso *06:18
frinnstaccess_log:26306:18
frinnstthough that is probably not a reliable number :D06:18
frinnstlog is from the 10th06:22
juethat's lot, and for 2.7?06:33
frinnstlet me sort out the unique downloads before you get too excited :D06:34
jueok :)06:34
frinnst25 downloads the last five days06:41
frinnst909 since oct 1506:43
frinnstdoubt everyone uses it though :D06:43
jueeven if only 10% are used, that's more than I thought06:56
Romster64bit is all the rage these days07:02
juesure, but I thought that we have less than 100 users at all ;)07:05
juefor both archs07:06
Romsteri really do wonder how many re really have.07:26
Romsters/re/we07:26
Romsterthough crux isn't about a large userbase anyways07:26
frinnstrsync logs might shed some light on it?07:57
Romsterquite possibly.08:13
Romsternumber of unique ip address would make it close08:13
rmullYou could add a "popcon" style survey to the installer to have an opt-in "phone home" stage so you can tally actual installs :P08:27
rmulljk08:27
jaegerjue: glibc-32 should be the only one, unless I've misunderstood somewhere (which is entirely possible)08:28
jaegergmp, mpc, mpfr 32-bit versions were not needed to create the toolchain08:28
jaegerI still do not have them installed and I'm able to build all the 32-bit stuff I've needed so far, including wine's deps.08:29
jaegerGoing to work on the deps for 32-bit firefox and flash today08:29
jaegeras well as the 32-bit glibc linker fix (which is just a symlink I need to add)08:29
jaegerI built the ISO over the 32-bit official ISO repo but did include frinnst's 64-bit repos in it08:30
jaegerupdated the kernel and initramfs and init08:30
jaegeras for the pure vs. multilib question, part of my intent from the beginning was to make it NOT diverge from pure64 unless needed08:32
jaegerso lib64 is a symlink to lib and lib32 is its own dir08:32
jaegerit is contrary to the FHS but fits our pure64 setup better08:32
jaegerotherwise we'd have to build everything with --libdir=/usr/lib64 or similar08:32
jaegeralso, along with that intent is the intent to be able to install a pure64 CRUX and convert it to multilib easily, like we used to be able to do with Lucas' version08:42
jaegerwhich works, I've been doing most of the development on a laptop that was installed in exactly that way08:42
jaegerThe one in that screenshot I linked (http://jaeger.morpheus.net/linux/crux/tmp/c27x64ml.png) is that system08:43
*** acrux|G4 has joined #crux-devel10:09
*** acrux|G4 has joined #crux-devel10:09
juejaeger: thanks, that sounds good to me10:18
jueI think that we sooner or later should switch CRUX to x86_64 and consolidate all our efforts10:20
jaegerI would like that as well10:21
rmull+110:21
jaegerI've gotten 32-bit firefox (binary from mozilla.org) working now, though flash doesn't work yet10:23
jaegerNot sure why, I don't get much useful from the terminal output10:23
jaegerjue: one thing which I have NOT tested yet is this: I would think that the "official" 64-bit crux could use the multilib toolchain but no -32 packages would be required if you want to run pure6410:28
jaegerThough if that bugs people who want a 100% pure64 toolchain I'm also fine with the "install pure64 and then overlay" method10:29
jaegerI will maintain a multilib ISO regardless, most likely10:30
jaegerI have a feeling the messages I'm seeing that say things like "assertion 'GTK_IS_VISUAL' (visual) failed" are related10:42
jaegernot sure, though10:42
juejaeger: if all the "overhead" is the glibc-32 package I'd go with the multilib toolchain for our ISO11:03
jueseems most flexible to me and the user can start building 32bit packages as required11:04
jaegerIt should be, I believe11:04
jaegeryeah, that was my thought as well11:04
mike_khmm, how the same binary for different archs is going to live in /usr/bin?12:44
mike_kI just don't remember how other multilib distros handle that12:44
frinnstso, firefox segfaults when compiled with -mavx (included with -march=native on a few cpus). best way to handle it in the port? "export CFLAGS="$CFLAGS -mno-avx"" ?14:10
jaegermike_k: for most software the binaries don't need to be duplicated - in most cases all you need in the 32-bit version is /usr/lib32/*14:14
jaegersome cases that doesn't work, like freetype, pango, gtk, gdk-pixbuf14:14
jaegerI copy the binaries that are needed to <binary>-3214:14
jaegerNot sure how everyone else handles it but I think Arch does it similarly14:15
jaegermike_k: for example, http://morpheus.net/gitweb/?p=jaeger/opt-multilib.git;a=blob;f=gtk-32/.footprint;h=fa6d8f4b25d16546e0f8943d12d9e9ddd6828c4a;hb=2.714:16
jaegerfrinnst: that's probably how I would do it, I guess14:17
mike_kjaeger: thanks. would be pretty odd for any widely used (and directly run) apps.14:29
mike_kjust hope there wouldn't be many of them14:29
jaegerWell, there aren't many cases for which you need a 32-bit app and no 64-bit one is available14:31
jaegerAt least not many that I've experienced14:31
*** acrux|G4 has quit IRC14:43
jaegeroops... I seem to have left grub on the multilib ISO15:36
jaegeroh well, not a big deal, it just isn't useful without patches15:37
mike_kwould be no grub at all on the iso? maybe statically linked 32-bit one would fit?15:43
jaegerIt would probably work fine, I just didn't mess with it since I don't use it15:43
mike_kI've used that on some release from obra.se 3 years ago15:48
jaegerlilo works fine from the ISO but I use grub2 from my repo anyway15:48
mike_koh, grub2. haven't tried that yet15:49
rmullugh15:49
rmullgrub2 is pain15:49
jaegerIt doesn't give me any trouble at the moment. It's overkill but works15:50
mike_kI used to play with and boot some fancy things like QNX or NetBSD. So, I've got used to old grub a bit.15:50
*** mike_k_ has joined #crux-devel16:37
*** mike_k has quit IRC16:40
*** mike_k_ has quit IRC16:46
jaegeroops, found another problem with the ISO, gonna take it down for tonight17:04
jaegerWill start a new one when I get home17:04
Romsterdamn it i alraedy downlaoed the multilib iso already to guess i should hold off doing it next weekend.18:55
jaegerYou can go ahead and use it if you want, just need to update pkgutils before you install any 32-bit stuff18:57
jaegerI left a line out of pkgmk.conf18:57
jaegerthe new ISO won't be a huge difference18:57
jaegerThat problem doesn't affect 64-bit ports anyway, only 3219:03
jaegerI just thought it would confuse people if I left that one up with the bad port19:05
Romsterwhat line is it missing i intend to play with 32bit on it.19:09
jaegerin the 32 section of the PKGMK_ARCH case, add: export PKG_CONFIG_LIBDIR="/usr/lib32/pkgconfig"19:09
Romsterah19:09
Romsterguess that would confuse some.19:09
jaegerotherwise it looks in /usr/lib/pkgconfig and gets very confused19:09
jaegerI'm going to set up rsyncable repos tonight as well19:10
jaegerwill save some hassle19:10
Romsterk19:10
jaegerhttp://morpheus.net/gitweb/?p=jaeger/core-multilib.git;a=commitdiff;h=65bf76a5712d97f86bcf1da51b75b77e85e60242 <-- here's the commit, though, so if you pull the current git repos you'll have the proper port ready to build19:11
Romsterso can i just boot off it pkgadd -u everything and rebuild kernel reboot? dunno if the setup script would be ok for my deviated system.19:11
jaegerIs your system pure64?19:11
jaegerif so, it'll be rather similar to http://crux.nu/Wiki/ConvertPure64ToMultilib19:12
Romsterno it's nearly crux 2.7.1 32bit currently other than a patch to binutils and gcc 4.6.219:12
jaegerthose are Lucas' old instructions19:12
jaegerAh, ok19:12
jaegerI have not tried to install it over a 32-bit install but I would imagine it works if you simply replace everything19:12
Romsterthat's what i assume too.19:13
jaegerI *have* installed over a pure64 install without trouble19:13
jaegerthe new bootstrap is started19:24
Romstergoing out for lunch soon if i'm in the mood later i'll start redoing my system.19:25
Romsterneed to test that rc-ng too19:25
jaegerhave a good lunch19:26
teK_yeah19:26
teK_upstream found a new annoying habit19:27
Romsterwill do hmm thinking and seeing that PKGMK_ARCH would someone try to later on extend it to CRUX PPL too.19:27
Romsternot that i use that.19:27
teK_how do I download (via pkgmk) the newest archive from: https://github.com/seb-m/pyinotify/tags19:27
teK_it will name the file 0.9.2 m(19:27
RomsterSorry, there aren't any downloads for this repository.19:28
Romsterwhat the hell.19:28
jaegerhttps://github.com/seb-m/pyinotify/tarball/$version ?19:28
jaegerpart of pkgmk's download is: --output-document=$LOCAL_FILENAME_PARTIAL19:28
teK_it will download to 2011-12-17 02:28:55 (93.0 KB/s) - `0.9.2' saved [59258/59258]19:29
teK_oops19:29
teK_to 0.9.219:29
jaegerso pkgmk should be trying to rename the download to the right thing19:29
teK_it won't.19:29
Romsteror even https://github.com/seb-m/pyinotify/snapshot/$name-$version.tar.xz19:29
teK_I never got the point in using $name for this but I will try19:30
Romsteri have thought about the entire rename thing i have a few ports that i do a mv to the correct name and extract manually. in the build{} function19:30
jaegeryeah, perhaps a snapshot would be preferable19:30
Romsterthe sites should offer one really.19:31
Romstertaged git tree should have a snapshot file19:31
teK_https://github.com/seb-m/pyinotify/snapshot/pyinotify-0.9.2.tar.xz is a 40419:31
teK_maaan19:31
Romsteror bz2?19:31
jaegerah, I guess since the download link doesn't have an extension it gets confused19:31
Romstergz dunno what they default too.19:32
teK_of course not19:32
teK_this would be too easy19:32
teK_so it took me < 1 Minute to create the Pkgfile and 10 to donwload that crap :P19:32
teK_their Downloads section is empty, of course ;>19:33
Romsteryou can shallow download it to the depth of 119:33
Romsterwhat i do is mirror it on my site then point source to that.19:34
teK_craaaap ;)19:34
Romstermore irritating for updating though19:34
teK_oh19:34
teK_seb-m-pyinotify-36a7f9419:35
teK_instead of cd $name-$version19:35
teK_hihi ;)19:35
Romsterwhere did you find that?19:35
Romsterusualyl you can't list the darn files19:35
Romsterbut if you know the name you can download them19:35
teK_I did that after downloading the archive19:35
teK_i.e. in build()19:36
teK_and this is just:19:36
teK_$author-$name-$shortref19:36
Romsterwhy cant' they use the darn tag instead of shortref argh19:37
jaegerI'd say just keep your own copy and point to that... less complicated19:38
jaegerone more step for each release but at least it's only one time per release19:38
Romsteri tell ya linux is getting to the point of being a bitch more so than downloading a file for windows.19:38
teK_well seb-m could have filled the Downloads Section of github, too19:38
jaegerThat's got nothing to do with linux, it's all that author19:38
Romsterin general google github soruceforge how many times have you had broken 404 urls?19:39
jaegertrue, they all suck, but none are linux-specific :)19:39
Romsterok downloads sites then.19:40
jaegersourceforge is the best now that it's been for years but man, was it a pain in the ass for so long19:40
Romsterdownload*19:40
jaegerStill has problems, they all do19:40
jaegersorta like browsers... they're all a pain19:40
Romsterindeed19:41
Romsteri was using fonts on some sties and then they decide to break that on the browsers.19:41
jaegerhrmm.... I need to fix openssl, it seems to want to build 64-bit binaries no matter what19:46
jaegershould be an easy patch19:46
jaegerah, much better19:55
Romsterthunderstorm lovely23:13

Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!