IRC Logs for #crux-devel Monday, 2012-07-02

juegood morning02:30
jueFYI, I'll be offline for the next 5-6 days, take care02:32
teK_you too :)02:33
juethanks :)02:38
*** jue has quit IRC02:39
teK_Romster: retried pushing?02:39 sucks:02:43
frinnstSorry, an error has occurred. Reason: That is an invalid ID, or the post has expired.02:43
Romsterthank you teK_ that fixed it02:52
Romsteri had similer errors pushing to opt-x86_64 too thats why i've never done anything in there either02:53
*** horrorSt1uck has joined #crux-devel03:00
*** horrorStruck has quit IRC03:03
teK_does itwork no?03:05
teK_(opt64 i mean)03:05
*** Romster has quit IRC03:29
*** mike_k has joined #crux-devel03:34
frinnstfuck your space-key tek!03:45
frinnst(i'm being nice!)03:46
teK_still loving you frinnst <3 <303:46
teK_so no worries03:46
frinnstpermissions being changed on the fs?03:47
teK_for contrib? yes03:47
frinnstset it sticky?03:48
*** Romster has joined #crux-devel04:08
*** Romster has joined #crux-devel04:08
teK_btw have  there ever been considerations for 'offical' "dist-upgrades"?05:57
teK_there were some suggestions (evenby you frinnst?) on how to do it without a physical CD + reboot to it05:57
frinnstusually just editing the repo-files and doing a sysup works without issues06:15
frinnstadding / removing new packages etc needs to be done manually06:15
frinnstbut there are potential potholes and "supporting" it is probably a pita06:16
teK_even if one focuses on core and opt only?06:16
frinnsta user blindly upgrading without being aware of major changes might run into trouble06:17
teK_we're (well jue, jaeger ,...) are in the progress of figuring things out right now. If this would happen better documented it should not be too hard06:17
frinnstif we were to change the init-scripts and not merging due to changes in /etc/pkgadd.conf, udev/mdev etc06:18
frinnstbut yeah, everything is doable06:18
teK_it just sucks to be unable to do a simple prt-get dist-upgrade. And yes I know that even Debian and especially Ubuntu don't always get that right. too06:19
frinnstdunno, add a new port in 2.7 once we release 2.8 containing the new repos?06:19
frinnstthat would require a user to manually install the repos06:20
teK_you mean s/2.7./2.8/g   /etc/ports/{core,opt,contrib}* ?06:21
frinnstyeah or add new {core,opt,xorg}-2.8.rsync06:22
teK_it should be more intelligent and handle problematic transitions (just as the ISO does in upgrade mode?)06:23
frinnstis there a possibility that there might be runtime issues? think libarchive/xz without running a static pkg tools?06:25
teK_yeah these cases should be covered06:26
teK_problem is: this is 2.7 -> 2.8 specifici and can only be created right before the2.8release process06:27
teK_maybe you cannot guarantee this to work in a  year either because there were other changes in the 2.8 ports that break 2.7 -> 2.8 transitions but not 2.8 -> 2.8 updates and therefore go unnoticed06:28
teK_I need a new (huge) root-Server. :-)06:39
teK_something like this:
frinnst10TB data /month?06:55
frinnstmy stupid german-skills are very lacking06:55
teK_exceeding this is 6,90EUR / TB06:55
frinnstor lowered to 10mbit unless you pay?06:56
teK_we use a fraction of 10TB at present06:56
frinnst~price ?06:57
teK_but most important for me: Administrationssoftware Webmin (bei LAMP verf├╝gbar)06:57
frinnstah, 49EUR06:57
frinnstpretty nice06:57
teK_we're paying the same (@hetzner, too) for:06:57
teK_AMD Athlon(tm) 64 Processor 3800+ single Core and 1GB RAM06:57
teK_i.e. we're stupid not to switch06:58
teK_you have to pay 49,- once but well..06:58
frinnstbut not xeon/ecc?06:58
frinnstI wonder how they run it06:58
frinnstare those physical servers?06:58
teK_Xeon/ECC starts at EX6:
teK_that's why they are called root-Servers I guess06:59
teK_funny thing.. some months ago they relocated our server .. but I missed the notice06:59
frinnsthaha, nice06:59
teK_at 2 a.m. they just ripped it out the rack06:59
teK_I _really_ got nervous then :P06:59
frinnstwhat a lovely thing to wake up to07:00
teK_still awake _working_ !07:00
frinnstso, all fine then I guess :D07:00
teK_you know, good persons have to work to make a living07:00
frinnstffs, how can a fresh windows install encounter "unknown errors" when running windows update for the first time?07:01
teK_maybe something flame-related? ;D07:01
RomsterhorrorSt1uck, has a nice upgrade method. but i don't believe it'll work every time.08:17
teK_an up-to-date core/opt installation is a known scneario so I'd love to see that feature integrated08:18
*** horrorSt1uck is now known as horrorStruck08:32
jaegerIs the new tzdata supposed to go in /etc/zoneinfo or does it matter?08:42
frinnstjue has a port for it08:42
frinnstdont remember off hand08:42
jaegerlooks like he did it the exact same way I did08:44
frinnstgreat minds, eh? :)08:44
* jaeger builds gcc 4.7.1 a few times10:40
*** sepen has joined #crux-devel10:49
sepenjaeger: $ cd initramfs-dir && find . | cpio -H newc -o > ../initramfs.cpio10:58
jaegersepen: yep, that's one of the ways to create them10:59
sepenproblem is #KERNEL_VERSION# right?10:59
jaegernot necessarily, you could sort that out *before* the cpio10:59
jaegerthe "problem" is more that with the .lst file you can be 100% specific about what gets included10:59
jaegerotherwise you'd have to remove things you don't want manually before cpio11:00
jaegerboth methods work fine, matter of preference11:00
sepenwell, busybox --install at least avoids some maintainement11:03
jaegersome, yeah11:03
jaegerMore importantly, though, the initramfs.lst includes only things needed for block devices and filesystems (plus their dependencies)11:04
jaegernot network drivers, graphics card drivers, etc.11:04
sepenthe objective is to find_and_mount the install media11:04
sepenrootfs could contains the rest of modules11:05
jaegerSo you definitely *could* use the find | cpio method to create it but I think that would be more work than the initramfs.lst11:05
sepenhmm in their favor this file give us more info about contents of initramfs and what changed from release to release11:06
sepenhere I tried to put dropbear statically linked into the initramfs, but size is around 3.5 for each binary (dropbear, dbclient, ...)11:39
sepen3.9M  work/dropbear-2012.55/_install/usr/sbin/dropbear11:39
sepenso I think that rootfs.tar.xz would be the best11:39
sepenhmmm wait, now I stripped binaries ;)11:41
sependropbear 736K, dbclient 692K11:41
sepenmy fault11:42
jaegerthat sounds more correct :D11:42
jaegereither way, though, it's more use in rootfs... no real need for ssh/dhcp before then11:42
jaeger <-- gcc 4.7.1, glibc 2.16.0 (multilib)11:50
jaegerwanted to get that built before I go out of town11:50
*** horrorStruck has quit IRC12:12
*** horrorStruck has joined #crux-devel12:14
jaegerglibc 2.16.0 breaks cpio (and others that use 'gets', presumably)12:58
horrorStruckjaeger: i guess it's gcc 4.7.1 that does13:03
horrorStruckplease god kill me13:04
jaegerwell, gets got removed from glibc, so.. :)13:04
jaegerdiffutils is another13:05
horrorStrucktook me less than 1 min to realize. given the fact i'm 3/4 drunk it's not too bad. i have to find a new nickname now anyway.13:05
horrorStruckmy birthday today :) gimme internets hugs13:06
jaegerhappy birthday13:06
horrorStruckah shit. it was yesterday already13:07
horrorStruckjaeger: --enable-stackguard-randomization in glibc configure sounds like a good stuff to have, dont you think?13:10
jaegerI have no idea. I see the option but it means nothing to me13:11
horrorStrucka security measure basically:
jaegerhow many applications are compiled with -fstack-protector?13:16
horrorStrucki have it in my pkgmk.conf on my crux router machine :)13:17
jaegerSo it's something that you would have to enable, not on by default13:17
horrorStruckFWIW, it's default on most distros13:17
horrorStruck(but we dont really care about others i know <- why i love CRUX)13:18
horrorStruckif it doesnt hurt performance we could enable it anyway, no?13:19
jaegerGot any benchmarks for it? I've got zero experience with it so I'm wary13:19
horrorStrucki dont but i could do some13:19
jaegerA link to some analysis that's already been done would be fine, if there is such a thing13:20
horrorStrucksometime in the future (before dinosaurs rule the world again)13:20
horrorStruckhow about ebaling graphite? it's mentioned on 2.8 TODO IIRC?13:21
jaegeryeah, that's planned for 2.813:22
horrorStruckah nice13:22
jaegerwell, assuming it doesn't cause any problems. I've built toolchains with it after jue and I discussed it but we haven't done a lot of testing yet13:23
horrorStruckFWIW again, gold is pretty robust, what about enabling that?13:23
jaeger <--- got some here if you run multilib13:23
horrorStruckgraphite is really really smooth to use13:24
jaegergold has never been stable for me in the past. What's the benefit?13:24
horrorStrucki havent got any single problem with that in years13:24
jaegerIt's probably been a year since I tested it last13:25
horrorStrucksorry, o was talking about grpahite13:25
jaegerI like the idea behind gold but it needs testing13:25
jaegerI haven't had problems with graphite13:25
horrorStruckgold can be built as a optional plugin13:25
horrorStruckmost ports will build with gold13:26
horrorStruckmain benefit is speed13:27
horrorStruckyeah i got some failure13:27
horrorStruckdon't use it as default13:27
horrorStruckbut you can enable it without being default13:28
horrorStruckwhile building binutils+gcc13:28
horrorStruckand keep ld.bfd as default13:29
horrorStruckjaeger: any reason why crux can't be 100% rolling release? upgrading toolchain is not as hazardous as i thought it would be. that would make maintainers' life easier no?13:38
jaegerdon't know, I haven't been around much for discussions of that. I was retired for a long time, don't know if it was considered much during that time13:48
horrorStruckretired but still on irc :P i dont think this has been discussed much TBH13:51
jaegerThat doesn't mean I paid attention 24/7 :D13:51
jaegerfar from it13:51
horrorStruckliar :O13:51
horrorStruck(difficult to be as mean as frinnst)13:52
horrorStruckteK_: OOC, any input on 100% RR CRUX>13:54
teK_I have a dpaste13:55
teK_for you.13:55
teK_it's from today13:56
* horrorStruck pretends to be afk13:56
teK_this kind of feature should be _thorougly_ tested13:57
teK_I noted  that I need a new, bigger root-server13:57
teK_just to do that13:57
teK_wtf @ML Post by Don Cupp13:58
jaegerhe just doesn't understand that "awesome" and "lots of users" aren't the same thing13:58
teK_awesome == few problems == few traffic13:58
jaegeralso doesn't know that most devel chat happens in IRC, I guess13:58
horrorStruckteK_: interesting. for sure it will require some user interaction but that may be nice to at least being able to do so.14:01
teK_it shouldn't14:01
teK_prt-get distupgrade14:01
teK_maybe reboot14:01
horrorStruckagreed. i'm already at crux 3.0 TBH14:04
horrorStruckhuh. forgot that: :P14:05
horrorStruckif you're up-to-date than distupgrade shouldnt be too painful one would think. RR distros do require manual interaction. wrt installing from iso, we may want to build (not so tested) weekly iso maybe, if that's not too painful.14:19
jaeger11 ports failed to build with the new toolchain/glibc14:55
jaegerer, 10. util-linux-ng was the normal footprint issue14:59
*** mike_k has quit IRC15:12
*** joe9 has quit IRC16:05
*** sepen has quit IRC19:57
horrorStruckonly 16 (but with 5 ignore so maybe more) with gold+lto, no too bad19:58
*** mavrick61 has quit IRC21:44
*** mavrick61 has joined #crux-devel21:45
Romsterif i had more time to work on testing late lunch at 2:40pm...23:41

Generated by 2.11.0 by Marius Gedminas - find it at!