IRC Logs for #crux-devel Monday, 2015-03-09

*** diverse has quit IRC01:48
*** diverse has joined #crux-devel01:55
*** nilp has joined #crux-devel02:53
*** nilp has quit IRC03:04
*** mavrick61 has quit IRC03:24
*** mavrick61 has joined #crux-devel03:25
*** crash_ has joined #crux-devel04:29
*** crash_ has quit IRC04:30
*** crash_ has joined #crux-devel05:00
*** ryuo has quit IRC05:01
*** ryuo has joined #crux-devel05:01
*** ryuo has quit IRC05:32
*** crash_ has quit IRC06:02
*** crash_ has joined #crux-devel06:13
*** crash_ has quit IRC06:29
*** crash_ has joined #crux-devel06:30
Romsterfrinnst, can you explain why we need /etc/ld.so.conf.d/firefox.conf08:34
Romsterdon't thunderbird and firefox have shell scripts to set the paths for there own libraries. firefox and thunderbird work finewithout that /etc/ld.so.conf.d/firefox.conf file and yes i did ldconfig after.08:42
Romsterbefore ldd showed that thunderbird binaries were resolving to firefox libraries.08:44
frinnstI think it was added before I took maintainership08:46
frinnsti've not bothered to remove it. but iirc it was added to shut revdep up. more or less08:46
Romsteri thik it's causing more hrm than good but i got a person working on a C implementation of revdep08:48
Romsterand it takes 1 minute 23 seconds to run compared to the shell script revdep takes over 20 minutes on my desktop08:49
Romsterhttps://gist.github.com/therealromster/63f5ea1e86abd7f25ce508:49
Romsterthat firefox.conf file is really screwing over resolving libraries.08:49
Romsteri am even guilty of it in firefox-pgo and firefox-gst08:50
*** nilp has joined #crux-devel09:01
Romster21.08user 10.28system 0:37.70elapsed 83%CPU (0avgtext+0avgdata 89232maxresident)k09:03
Romster127184inputs+0outputs (1719major+2156716minor)pagefaults 0swaps09:03
Romstereven faster when the kernel has it cached.09:04
*** nilp has quit IRC09:13
frinnstit was added with the release of firefox 4.009:15
frinnstwhere we started using the bundled xulrunner libs09:16
Romsterah09:17
Romsteri removed that file from my system, firefox and thunderbird is still working (after ldconfig) but i haven't rebooted yet.09:18
frinnstIt was added 4 years ago. Dont remember that far back :)09:20
Romsterneather do i but i'll reboot later and test without that file and report back.09:25
pitillowhat a mess with llvm and clang...09:25
frinnstwhat *needs* clang btw?09:30
frinnstchromium seems to work without it. but not sure if plugins from chrome (flash) still works09:30
frinnstnope. still works09:31
pitilloclang, llvm, mesa3d...09:32
frinnstmesa3d needs llvm, not clang09:33
pitillotrue, sorry09:33
frinnstnp :-). its a bit confusing09:33
pitilloyeah... I've read it backwards, my bad :)09:34
Romsterwhat mess?09:42
Romstermesa3d only sues llvm09:42
Romsterand that's for only the raideon driver09:43
Romster(i know i can never remember how raideon is spelled)09:43
Romsterchromium needs clang apparently.09:44
pitillois radeon (gallium) the only driver which relays in llvm?09:44
frinnstno all gallium drivers do09:48
frinnstbut radeon is perhaps the only hardware that *requires* a gallium driver. intel etc already have native drivers09:49
frinnstim not really sure09:50
pitillointeresting... I'll check it slower to build mesa3d on ARM without gallium, since I'm not sure svga/egl uses it09:50
Romsteras far as i know yes09:50
Romsterunless mesa3d-gallium is seperated from mesa3d?09:51
Romsterto migrate the llvm09:51
Romsterpersonally you don't build llvm all that often09:52
frinnstI dont think there is any future in that though09:52
Romsterand added work of keeping 2 ports in sync isn't worth the effort.09:52
Romsterpitillo, it is half the tiem considering teK_ split out clang from llvm09:53
Romstertime*09:53
Romsterand there is opt/ccache09:53
frinnstclang is a "very optional port". "llvm" is pretty mandatory09:53
pitillopretty mandatory under x86... and it isn't probably under ARM gpu's (if I understood all the info right)09:55
Romsterso arm does not need gallium?09:58
Romsterall i can suggest there pitillo is to grab mesa3d remove llvm and disable gallium for arm.09:59
Romsterif that is what you really need. and want to keep the version in sync.09:59
pitillothat's what I'm trying to confirm to avoid building llvm under the toys09:59
pitillomesa3d should be overlayed under ARM (no ati/nouveau/intel cards... then footprint mistmatches)10:00
frinnstits not like there is a universal gpu driver for arm10:05
pitillonot universal drivers, but universal EGL implementations used by drivers10:08
pitilloand it seems to be gallium dependent (EGL/gles)10:09
Romsteryou can spit mesa3d up to different drivers.10:21
Romsterbut effort.10:21
Romstermesa3d-driver-foo10:22
Romsterand adjust dependencies for each10:22
pitillothere is no real need for that (more maintenance). ATM I'm playiing a bit with all options (enabling by default gles and currently trying to add openvg too)10:30
*** crash_ has quit IRC10:37
juefrinnst: you found the reason for your nfs issue?11:02
frinnstnot looked into it yet11:20
jueok, please tell me if we have to fix something11:21
frinnstyeah will do.11:24
frinnstcurrently i start nfs followed by nfsclient. nfs shouldnt exit before everything is completed, no?11:25
Romsterpkginfo -o libxul.so | grep thunderbird11:30
Romsterthunderbird  usr/lib/thunderbird-devel/sdk/lib/libxul.so11:30
Romsterthunderbird  usr/lib/thunderbird/libxul.so11:30
Romstershould we not be removing /usr/lib/thunderbird-devel ?11:30
Romsterisn't that a build time thing?11:30
Romsterarch does this11:32
Romster# We don't want the development stuff11:32
Romsterrm -r "$pkgdir"/usr/{include,lib/thunderbird-devel-*,share/idl}11:32
Romsteri'll file a bug report after testing this myself.11:32
frinnstDont arch have a bunch of -devel packages like rhel and debian?11:32
Romsternope not as far as i know.11:33
Romsterhttps://www.archlinux.org/packages/extra/x86_64/mesa/ .la .pc .h so files exist so nope.11:33
Romsteri think it's the case of what will ever build against thunderbird?11:34
Romsterevery time poppler gets bumped it breaks inkscape -_-11:37
Romsterand i mean not just recompile fix but changes.11:38
juefrinnst: yes, it should wait until idmapd and statd are started12:27
jueRomster: inkscape shouldn't link against poppler but poppler-glib, there's nothing we can do besides recompiling12:29
juewell, I'll do a [notify] commit for the next update12:30
Romsterits nottoo bad but i have to patch inkscape again which i'll do shortly12:32
Romsterjue, i hope your keen on a C implementation of revdep to replace the shell one in prt-utils. that runs 1/20th of the time?12:33
Romsterits not quite ready but not far off being done.12:33
Romsterhttps://gist.github.com/therealromster/d02f5f72b33a3066d2d412:34
juesure, if it works and we can trust the author12:34
Romsterstill cleaning up the output and info levels.12:34
Romsterme and ryuo from frugalware are working on it. mostly ryuo programming and me on advice.12:35
Romsterruns in 1 minute 21 seconds compared to 21 minutes.12:35
Romsteron my desktop12:36
Romsterand i think we got the last bug out of it for the logic.12:36
Romsterjue, queston be wuld you like it a stand alone archive or jsut the files in a directory to add to prt-utils/devdep/...12:37
Romsteri was thinking of just tracking it in prt-utils.git12:37
Romsteri'll submit a feature request when it's ready.12:38
Romsterthen everyone can test it12:38
*** Feksclaus has joined #crux-devel12:39
*** ryuo has joined #crux-devel12:39
ryuoRomster, been talking about me have we? :P12:40
Romsterhaha yeah12:40
Romsterbusted12:40
Romsterjue, i'm sure you may remember ryuo12:40
Romsterex crux user.12:41
ryuohow would he? it's been 6+ years.12:41
Romsteri dunno.12:41
Romsterwas it that long?12:41
ryuoI started with frugalware in 2008.12:41
Romsteranyways i suggested a rewrite of revdep because it's so damn slow.12:41
Romsterlater find out ldd is a shell script itself.12:42
ryuoi've done something similar for FW before. not too hard, but i had to adapt to the different design.12:42
Romsterso we call /lib32/ld-linux.so.2 and /lib/ld-linux-x86-64.so.2 directly12:42
Romsterand parse the database to a hash table.12:42
ryuoRomster, we? WE? lol12:43
Romstershoulnd't be that much different.12:43
Romsterwell ryuo sorry12:43
ryuoRomster, you helped test it, but i wrote all of it so far. :P12:43
Romsterall i did was test and give direction12:43
Romsteryes12:43
ryuoi know what you meant, but i just find it funny the words people use.12:44
Romsterits still beyond my programming level for now.12:44
Romsterwell i mean we as a group working on it.12:44
Romsterbut i did mention before you joined that you are doing all the programming12:44
Romsterunder my direction and some of yours.12:45
ryuoRomster, as for license, you can choose that yourself. i don't really care.12:45
Romsteranyways when the formatting and licence is added, it should be ready for testing.12:45
ryuoi'm not surprised by the speedup.12:46
Romsteri was thinking GPLv2 anyone here have a preference? on what license to use.12:46
ryuoshell is a very poor choice for designs where speed is desirable.12:46
Romsterespecially when it's forked a few hundred times.12:47
ryuonot to mention file has poor performance. it wasn't designed to be used as a programming tool.12:47
ryuoby switching to a whitelist, that performance was greatly improved.12:48
Romsterand this is why i noticed thynderbirds dev/sdk directory with the /some/ of the dame libraries as in /usr/lib/thunderbird/ ad that /etc/ld.so.conf.d/firefox.conf messing LD resolving thunderbird to firefox libraries.12:49
Romsters/dame/same12:49
Romsters/ad/and12:49
ryuostill the shell script was useful for helping diagram the problem solution.12:49
Romstergreatly i was only expecting a 50% speed up if that.12:50
ryuoit should be greater than that.12:50
Romster1:23 compared to 21 minutes12:50
Romsterso like 1/2012:50
ryuoi'm surprised nobody wanted to optimize it before now...12:51
Romsterguess no one cared/bothered or real life more important who knows.12:51
Romsterryuo, wanted something to teach me with and i suggested revdep12:51
ryuoRomster, eh? i remember it as you talking about it being slow. no mention of the learning until later?12:52
ryuooh well.12:52
Romsteri'll start on cutting my teeth on wc etc.12:52
ryuoyou sure C is what you wish to start with?12:52
Romsterwell i do better by example so i can pick it apart and learn from it.12:53
Romsterwell i wish to know enough to knock some basic stuff out.12:53
ryuorevdep required about intermediate skill to write...12:53
ryuoin C that is.12:53
Romsteranyways let me know what licence you'd prefer for revdep guys.12:53
*** Feksclaus has quit IRC12:55
*** Feksclaus has joined #crux-devel12:56
*** leo-unglaub has joined #crux-devel13:33
*** heroux has quit IRC14:23
leo-unglaubwhat would you guys think about providing also binary packages on crux.nu?15:20
frinnstnot from crux.nu. too much bandwidth and space15:21
Romsteri already keep some on mine as your aware. but i have only half done the CI15:22
leo-unglaubRomster, yeah, you have something, i want to build something15:22
leo-unglaubmaybe we could centralize it a little bit :)15:22
leo-unglaubfrinnst, bandwitdth is not a problem for me, my company could provide the hardware behind download.crux.nu15:23
Romsterwell i can build any port with 2 commands in a container.15:23
Romsteri can jenkins it15:23
frinnstAs could I and im sure jaeger could too. Problem is if/when we change jobs15:24
Romsternot an issue when there is more than one mirror15:24
leo-unglaubfrinnst, then we switch the dns entry :)15:24
Romsterdo you plan to go back to your cave?15:24
frinnstdidnt prologic fiddle with a binary build system?15:24
frinnstwhat happened to that?15:25
Romsterstill is i got one in the works too15:25
Romsterdunno he is on so many things currently.15:25
Romsterask him when he wakes up15:25
leo-unglaubi am going to write my ideas about it to the mailinglist15:46
leo-unglaubbecause i think, even if i love a source based distro !!!!, sometimes official binaries could be great!15:47
leo-unglaubfor example if you install crux on an very old computer where building glibc just would take 5 houts ..15:47
*** crash_ has joined #crux-devel15:51
*** crash_ has quit IRC16:03
*** crash_ has joined #crux-devel16:06
*** sepen has joined #crux-devel16:12
*** heroux has joined #crux-devel16:12
pitilloleo-unglaub: the problem is maintaining that "system", btw, it'll be good to read about it (if I remember right, it was discussed)17:06
jueI don't think that we should officially provide binary packages, if we ever do that we need strict dependencies17:06
jueI think it's ok to have the big monsters somewhere but at all really optinal ;)17:08
juesepen: btw, I've monit 5.12.1 running here, works without issues for me17:09
leo-unglaubjue, the install iso already provides "official binaries"17:27
leo-unglaubi mean it has absolutly no priority for me, i am still hacking on my initramfs binary :)17:28
diverseI was also going to ask you about that17:28
diverseleo-unglaub: how's the progress on your implementation?17:30
leo-unglaubdiverse, i would say 50%17:31
diversewow, fast!17:31
leo-unglaubif you boot into my initramfs you get a list of all found harddrives17:31
leo-unglaubthen you have to specify what you want to do17:31
leo-unglaubb <- shortcut for boot17:31
leo-unglauband then the number from the partition of the list17:32
leo-unglaubb1 would boot from /dev/sda117:32
leo-unglauband so on17:32
diverseso, b is whatever the boot partition of /dev/sdx17:32
diverses/partition/device/17:33
leo-unglaubyou get a list printed like that: http://paste.debian.net/plain/16040417:34
leo-unglaubdepending on all your drives17:34
leo-unglaubthen you just use "b <number>"17:34
leo-unglaubto boot from that partition17:34
leo-unglaubthen you will be prompt for a password17:34
diverseah I see17:35
diversekind like a boot menu17:35
leo-unglaubyes, i need that step for denyability17:36
leo-unglaubif the initramfs is configured to always boot from X someone could say you have an OS on X17:36
leo-unglaubbut if you just get propmt for an action without a default17:36
leo-unglaubno one can proove that you actually have an os encrypted on that device17:36
leo-unglaubbecause without luks header it just looks like random data17:37
leo-unglaubthat way you always have to type "b <number>17:37
leo-unglaubbut i think thats okay ....17:37
diverseI suppose the need to be explicit is necessary17:43
jueleo-unglaub: that's a quite different story, but you know that17:44
leo-unglaubjue, hmm, actually no17:44
leo-unglaubwhat would be the big difference?17:45
leo-unglaubi mean it would be the exact same binary17:45
leo-unglaubor do i miss something?17:45
jueyeah, they are build in very controlled enviroment17:48
jueonly the ports from the iso are installed17:49
leo-unglaubyes, thats exactly what i wanted for crux.nu as well17:49
leo-unglaubcontrolled environment, signed packages and only the core17:50
leo-unglaubmaybe even determistic builds ... even if i am not the largest fan of that17:51
sepenjue: glad to know, I forgot to update it, thanks17:52
sepenleo-unglaub: ping17:52
leo-unglaubsepen, pong17:53
sepenhttps://crux.nu/gitweb/?p=ports/xfce.git;a=commitdiff;h=5e3fc0669e6eee5abcedc37e0d6443b96c7a9f5d17:53
sepenwhy are u using an extra variable in Pkgfiles?17:53
sepennameBuild=${name/-icon-theme/}17:53
sepenis strictly necessary?17:53
leo-unglaubsepen, i cannot see the content of that link because i get an OCSP cert error17:54
sepenand why to move { to next line?17:54
leo-unglaubbut i know what you mean17:54
leo-unglaubi use an extra variable if i have to calculate the same function more than one time17:54
leo-unglaubin this case 3 times17:54
juesepen: np17:54
sepenpfffff17:54
leo-unglaubthen i put that stuff in a variable17:54
sepenI don't agree with taht17:54
sepen*that17:54
sepenplease read: https://crux.nu/Main/Handbook3-1#ntoc2717:55
leo-unglaubso you prefer to have ${name/-icon-theme/} 3 times in your code than once in a variable?17:56
sepen"Do not add new variables to the Pkgfile...."17:56
sepenyes, I prefer that, because your new variable can break other scripts that read Pkgfiles17:56
leo-unglaubyes, and now the secound part17:56
leo-unglaub Only in very few cases does this actually improve the readability or the quality of the package17:56
leo-unglauband in this case it does17:56
sepenok17:56
leo-unglaubi even checked for name collisions before i did that17:57
leo-unglaubi greped the pkgmk source to check if there would be a collition17:57
sepenI do not agree to introduce these changes without consensus17:57
sepencan someone from @core @opt review that please17:57
sepenI'm not going to start breaking the rules now17:58
leo-unglaubwhat rules?17:58
leo-unglaubthe rules state that you can add variables if it improves the code quality17:58
sepenrules in CRUX's Handbook17:58
leo-unglauband sorry, but variables exist to store stuff in them and before i copy the same logic 3 times in a file i store them in a variable17:59
leo-unglaubthats also why we use $name in there so we dont have to dublicate stuff17:59
sepenhow that improves it?17:59
sepennah17:59
leo-unglaubbecause you never dublicate code?17:59
sepenand why not keep 'build () {' ??18:00
sepenI'm not agree with some of your changesets18:00
sepenand also I'm tired to discuss that, feel free to do whatever you want18:01
leo-unglaubare you trolling me right now??? are you really suggesting that i should copy ${name/-icon-theme/} 3 times on that script instead of using a variable??18:01
sependo it like hell you want18:01
leo-unglaubi am sorry, but that discussion is over. i am not engaging in a discussion about dublicating code18:01
sepenI'm busy, bbl18:01
leo-unglaubif am looking forward to get feedback on thing i have done wrong and bug reports, but the reduction of code dublications is clearly not a bug18:02
*** crash_ has quit IRC18:05
teK_I suggest to use as few variables as possible18:05
teK_we are ought to wirte Pkgfiles, not shell-novels18:06
teK_:-)18:06
teK_new discussion in 3..2..1..18:06
leo-unglaubi totally agree with keeping things small and clean. but if a variable is needed to keep things faster, easier and cleaner i am going to use one after i made sure there is no collision18:07
sepenI just not see the point to refactor 3 lines of code18:14
sepenand more when there is risk of break other scripts that use pkgfiles18:14
leo-unglaubthere is no risk in it, the build is spawned into it's own shell, so there is no overlapping with other pkgfiles18:14
sepenif all your changes will be oriented in this style I prefer to look away18:15
leo-unglaubyou mean clean, simple and working?18:15
sepenyep18:15
leo-unglaubwhen i do something i like it to work 100% .. thats why i am going over all those old Pkgfiles.18:17
leo-unglaubi dont do that to be mean to you18:17
leo-unglaubbut i just want it to work perfectly and when afer a couple of years an icon package still has no script to regenerate the icon cahce18:18
leo-unglaubthen i am adding one18:18
leo-unglaubbecause i care about that package to work when installed18:18
leo-unglaubi have no idea why you think i do that to show you off or something18:18
sepenI'm just surprised that neither got to discuss these changes before commit them18:18
sepentime to left out the office18:19
sepenbbl18:19
leo-unglaubthat is actually not true18:19
*** sepen has quit IRC18:19
leo-unglaubwhen i changed an api i talked to the people here18:19
leo-unglaubtwice btw ...18:19
leo-unglaubabout splitting the vcs plugin and about the suspend function18:20
leo-unglaubah, he left in the middle of a discussion ... i hate that ...18:20
*** leo-unglaub has quit IRC18:43
diverse...18:46
pitillocan someone take a look to ppp and perl sources? no way to download them from here18:53
*** Claus_ has joined #crux-devel19:18
*** Feksclaus has quit IRC19:21
*** crash_ has joined #crux-devel20:02
*** c0x has quit IRC20:11
*** crash_ has quit IRC20:18
*** c0x has joined #crux-devel20:18
teK_frinnst: how did you fill up the emptiness inside after binge watching House of Cards?20:27
frinnstwatch the miniseries right away20:58
frinnstthen watch hell on wheels20:58
frinnst4 seasons20:59
teK_not bad21:19
teK_I managed to watch gone girl21:20
teK_interstellar and wolf of wall street are still on my lit21:20
teK_s21:20
*** jaeger has joined #crux-devel22:46
frinnstyeah gone girl was pretty cool23:31
frinnstwolf of wall street was .. meh. not my thing23:31
frinnstnot seen interstellar yet23:31
*** heroux has quit IRC23:41

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