IRC Logs for #crux-devel Thursday, 2014-03-06

jaegeryeesh, crazy busy last two days01:06
prologicjoin the club :)01:07
jaegerok, finally looking at llvm now01:14
jaegerSo yesterday I had to work on a really bad setup at work01:17
jaegerA workstation that's hooked up to an MRI scanner. It's an HP WX8400 but the scanner manufacturer, GE, built it and I'm not allowed to touch it most of the time01:18
jaegeryesterday it was broken and we couldn't run any scans so the guy in charge asked me to look at it01:18
jaegerIt had 4 drives in it: 73GB SAS 6GB/s x2, 300GB SATA, 1TB SATA01:19
jaegerthe SAS drives have SAS->SATA converters on them and the motherboard is 7 years old, it only supports SATA1 speeds01:19
jaegerAnyway, that's dumb but the problem was that the boot drive started to fail. clicking, etc.01:19
jaegerIt wasn't backed up, of course01:19
jaegerFortunately ddrescue was able to recover it all, though that took a while01:20
jaegerNext problem is that the second SAS drive and the 300GB SATA drive were configured in a RAID0 (yes, 0) array for the data storage01:20
jaegerthe 300GB drive lots its partition table and took the data with it01:20
jaegerI recreated the RAID array but of course couldn't save any data01:21
jaegerSuch a stupid build... and they deploy it to hospitals01:21
jaegerThey even said if they were at a hospital they'd just have replaced and rebuilt the entire system but since we're a research organization and the guy in charge has *heavily* modified the scanning software, the GE guys just watched and chatted with us while I rebuilt the damn thing01:21
jaegerFortunately it ended up saving the guy in charge a lot of time since he didn't have to rebuild all of his modifications and whatnot01:24
*** deus_ex has quit IRC01:33
*** deus_ex has joined #crux-devel01:35
Romsterjaeger, i got a raid recovery tool but it's windows based should you wanna try that. iv'e recovered a failing hdd out of a 2 raid0 set.01:43
Romsterand got most of the files back01:43
Romsteralso ddrescue is awesome.01:44
jaegeryeah01:51
jaegerunrelated to that machine, I love how solid mdadm is these days... I've got a 14-drive mdadm RAID10 made of 10krpm 146GB SAS drives for some full packet capture stuff at work and it's working great01:54
prologicwow02:09
prologicnice story :)02:09
prologicwhat a crappy system though02:09
prologicheh02:09
jaegerI guess the idea is that your dicom data from MRI scans should go immediately to a PACS system but still, I'd have made it a bit more robust than that if I were the designer02:20
jaegeralso had fun yesterday fixing a cups install on an old ubuntu 9.04 install that they won't let me update due to some customized software02:22
jaegerhad to remove and recreate a printer queue via the command line because the web GUI didn't respond to button clicks02:23
prologicheh02:26
prologiccan't upgrade because of software requirements/02:26
prologicwhat crap :)02:26
prologicit'd had better be written in C/C++02:27
prologicusing tight dependencies on old glib ;)02:27
jaegerto be more clear, it's a parallel port kernel driver for use with the MRI scanner02:27
prologicoh I see02:27
prologicand so what, the standard kernel parallel port drivers don't work good enough? :)02:28
jaegerthat's what they tell me02:28
prologichard to believe really02:28
jaegerI finally convinced them today (after a couple years) to at least try it on a current ubuntu install02:28
prologicwell that's a start :)02:28
jaegerso I built a machine with ubuntu 12.04 today that has the exact same hardware, going to give it to them tomorrow to test when they can02:28
prologic*nods*02:29
jaegerRomster: I should have something for you to test soon with llvm/llvm-32, been back and forth away from the computer so it's slowing me down02:54
jaegerjue: I ran into an odd issue when I did the bash/readline updates on a machine earlier... I should have logged it but was so busy I forgot02:59
jaegerjue: it seems like updating readline before bash broke bash (and gawk, incidentally) because they use a symbol that was removed from readline03:00
prologicahh yes03:04
prologicbash depends on deadline and ncurses afaik03:04
prologicand I can see how that could happen :)03:04
jaegerindeed03:04
jaegerI'll have to test it again in a VM and see what the exact sequence was03:05
prologicI could test it right now for you in a CRUX Docker container :)03:06
jaegerfeel free if you like :)03:07
prologicsure :)03:07
prologicit's just so quick and convenient03:07
prologicI'll paste a terse output of the results03:07
prologicincluding exact steps taken03:07
prologicso first try, update readline before updating bash?03:07
prologicsee if bash breaks - which it probably will03:08
jaegeryeah.. and since awk/gawk break as well it gets hard to update things03:09
prologic*nods*03:10
prologicI wonder if ed/sed still work after the breakage03:10
prologic:)03:10
jaegeryeesh, llvm's libdir stuff is a mess03:20
jaegermore work than I expected03:21
prologic$ docker run -i -t crux:3.0 /bin/bash03:21
prologicbash-4.2# ports -u &> log && prt-get cache03:21
prologicbash-4.2# prt-get diff | egrep "(bash|readline)"03:21
prologicbash-4.2#03:21
prologichmm no bash/readline updates from 3.003:21
prologicshould I bump it by hand?03:22
prologicor update to 3.1.dev repos?03:22
jaegeryou could just grab those two ports from 3.1, yeah03:22
prologickk03:22
jaegerit's not updated in 3.0. I should have mentioned that03:22
prologicahh nps :)03:22
prologicdo we have 3.1 resync ports sources?03:22
jaegeryeah, in the standard location03:23
prologickk03:23
prologicso just s/3.0/3.1/g03:23
jaegeryep03:23
jaegerIn the end it may not matter because upgrading from 3.0 to 3.1 isn't a "supported" method anyway and the updated versions will be bootstrapped properly on the ISO03:24
jaegerodd issue, though03:24
*** ____mavrick61 has quit IRC03:38
jaegerhrmm... I probably won't have llvm-32 ready tonight, it's being a pain03:38
*** ___mavrick61548 has joined #crux-devel03:39
jaegerQuestion - llvm installs clang, do we actually NEED clang for anything?03:47
prologichere's an idea though04:10
prologicif we created # Depends on: for all our core ports04:10
prologicI know we don't for whatever reason04:10
prologicbut if we did04:10
prologicwouldn't pet-get update them in the proper order then?04:11
prologichttp://codepad.org/0BWeoaos04:11
prologicIn any case, upgrading readline before bash worked just fine04:12
prologicno breakage04:12
jaegerok. not sure what happened in my case, then04:12
jaegerI'll keep an eye out for it again, I guess04:15
jaegerthanks for testing04:16
prologicwant me to try anything different?04:18
prologicdifferent sequence?04:18
jaegernot that I can think of at the moment but thanks04:19
jaegerI'm testing building llvm and llvm-32 without clang04:23
jaegerIf that's a problem for anyone tell me04:23
prologicdon't care :)04:24
prologicgcc/clang04:24
prologicblah :)04:24
jaegerwell, I didn't see anything that actually uses clang as a separate port with prt-get dependent and it's just a part of the llvm port anyway. so removing it could save some build time04:26
jaegerInteresting that llvm builds with clang if it's installed04:33
jaegerman, I'm tired of rebuilding llvm so many times, heh05:08
prologichah05:08
jaegerThink I'm done with llvm itself, back to llvm-32 now05:09
jaegerif that builds successfully I'll try mesa3d-32 again in the morning but probably going to bed after llvm-32 is sorted05:09
prologicyeah get some sleep :)05:11
prologicmust be late for you!05:11
jaeger23:11, not terrible... but I'm definitely slowing down05:11
prologicheh05:12
prologicnothing depend on llvm?05:12
prologicwhy do we even bother? :)05:12
prologics/depend/depends05:12
jaegerin 3.0, no. in 3.1, yes05:12
jaegerIt's listed in the TODO31 doc05:13
prologicahh05:13
prologicso mesa3d05:14
prologicgotcha05:14
prologicbeen thinking too05:14
prologicperhaps we should consider enabling contrib by default?05:14
prologicor maybe I should consider this for the Docker images of CRUX I maintain05:14
prologicit's a bit of a PITA to do a bunch of sed work on /etc/prt-get.conf and rename the .inactive rsync port def in /etc/ports/05:15
prologicPITA programmatically that is05:15
prologicConsider: https://github.com/therealprologic/docker-crux-python/blob/master/Dockerfile05:15
jaegerteK_: http://jaeger.morpheus.net/linux/crux/files/llvm.diff <-- here's the llvm diff, any thoughts/comments/objections?05:16
jaegerhow's that a PITA? It seems like that's programmtically trivial unless I'm misunderstanding what you mean05:16
prologicwell okay it's not that much of a pain :)05:17
prologicI just think perhaps it would be nice if this was the default05:17
prologicand we were a little more confident in our contrib ports/05:17
prologicdunno just thinking out loud05:17
prologicI've never not enabled contrib rots on any of my systems I don't think05:17
jaegerRomster: http://jaeger.morpheus.net/linux/crux/files/llvm-32.diff05:20
jaegerNOT tested with mesa3d-32 yet05:20
jaegergoing to bed, will investigate mesa3d-32 tomorrow. Night05:24
juejaeger: was a bit hard, but I can reproduce the issue and found the reason for it09:19
juejaeger: we run in the same problem as here -> http://thread.gmane.org/gmane.comp.gnu.readline/48509:20
juejaeger: what happens is that in the install phase libreadline.so.6.3 is rebuild, but _not_ linked against ncurses09:21
juejaeger: see here, important is the 'gcc ...' line -> http://f672438c7d3b1d59.paste.se/09:22
juejaeger: note the missing -lncurses09:22
juejaeger: fix is trival though, after adding -j1 to the make calls I was unable to reproduce the error09:23
frinnstou, a backlog09:30
frinnstslept to 6am - i needed sleep09:30
juefrinnst: good morning :)09:46
frinnst:=09:47
frinnst)09:47
jaegerjue: good to know, thanks :)13:41
jaegerwe seem to be making quite good progress on TODO3113:48
jaegerhrmm... guess I'll have to install mesa3d-32 on real hardware to test, the vmware xorg video driver doesn't compile properly13:57
jaegerunfortunately my 3.1 real hardware is at home and powered off right now13:59
frinnstwhats left?14:22
frinnstlvm to eudev?14:23
jaegerI think that's still needed, yeah14:26
*** kipling has joined #crux-devel18:13
*** kipling has quit IRC18:45

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