IRC Logs for #crux-devel Thursday, 2015-03-12

Romsterjaeger, ah i'll look into that.00:46
*** leo-unglaub has joined #crux-devel01:33
*** Feksclaus has quit IRC02:04
*** deus_ex has quit IRC02:23
*** deus_ex1 has joined #crux-devel02:23
*** mavrick61 has quit IRC03:33
*** mavrick61 has joined #crux-devel03:34
*** leo-unglaub has quit IRC03:44
*** c0x` has joined #crux-devel05:02
*** c0x has quit IRC05:09
*** deus_ex1 has quit IRC05:14
*** ryuo has left #crux-devel ("Leaving")09:04
teK_jue: will do09:04
teK_I also feel partially responsible for this mess; for future committers we should see that the 'stick around, adjust you attitude' part does not fall short09:06
jueteK_: thanks and yes, your are right, this is probably the most important part at all09:09
Romsterit's ok teK_ we don't blame you.09:13
Romsterso whats the current situation? Do we just go back to only sepen again or is leo going to sort out his differences?09:14
jueno, of course not :)09:14
frinnstleo was very clear that he didnt want any part of it09:14
frinnstso back to normal i guess09:15
Romsteryes but you know people say stuff when they are angry and don't always mean it.09:15
frinnstAnyways, I dont need this kind of drama in my free time and I doubt anybody else does09:16
Romsteri think i was over it before it began to be honest.09:16
Romsterso who's the prt-utils maintainer these days jue ?09:17
Romsterabout new revdep i haven't found any more text issues. i think it's ready for mainstream use. would it get accepted into prt-utils or is it best as it's own package?09:18
juesure, I'll include it in prt-utils09:20
juedon't think that we should create a new ports for it09:21
Romsteras for xfce i could fix stuff up in my spare time and work with sepen, if it's ok with sepen. i do have enough on now but i can help out like i do in xorg from time to time.09:21
RomsterI can change the packaging if you like or how it's presented. where it's hosted etc.09:21
Romsterjust name it09:22
juelet's see, but propably best to put it into a subdir of tools/prt-utils.git09:24
jueI'll do some test and can do a new prt-release over the weekend09:26
Romsterlet me know if anything is wrong. it has been tested on a few systems now.09:27
Romsterbut you know it's hard to test all use cases.09:27
Romsterthe only issue i can see is it's incomplete for ppc and arm use. but we don't target them arch types.09:28
Romsterbe sure to grab again was some text fixups in man page and on revdep -h09:29
Romsteryou might want to do LDFLAGS='-static' make09:30
Romsterupto you.09:30
Romsterdoes depend on zlib and elfutils.09:30
jueindeed, a static binary make sense for a tool like revdep09:35
juehave you done it? size of the binary?09:36
jueguess a min. of 800K?09:36
jueIIRC, that was the min. size for a binary with glibc09:37
juesometimes in the past I was a big fan of dietlibc and very small binaries, but nowadays I don't care anymore about09:39
frinnstyeah that is a huuuge uphill battle09:39
Romsterabout 800KB for static09:39
frinnst-rwxr-xr-x 1 root root  74M Feb 25 09:23 libxul.so09:39
Romstermin was like 8KiB for a binary with glibc?09:40
frinnsti remember when i had enabled nostrip and all sorts of debugflags in gcc09:40
frinnstand forgotten about it...09:40
Romsteroh i haven't got nostrip in the Makefile but pkgmk does this anyways09:41
Romsteroh boy everything be 10 times as large frinnst10:13
Romsterand slower with debugging on10:13
frinnstsure is10:28
frinnstbtw, have any customers running panda romster?10:28
Romsternot that i'm aware of.10:29
frinnstone of our msp partners just dropped by. He has ~1000 panda clients. they switched from karspersky a month ago10:29
frinnsthe looked suicidal10:29
Romsterkarspersky is pretty good why would you switch10:29
frinnstlicense reasons10:30
frinnstthe corporate windows world is quite fucked up10:30
Romsteroh geez.10:31
Romsterall over a false positive10:31
frinnstthe city counsil where i live also run panda10:32
frinnst*everything* is fucked :D10:32
Romsteroh geez, um cna't they just whitelist it or turn off the resident shield and restore?10:33
frinnstperhaps if you only had one machine to deal with.. but ~1000 ?10:34
Romsteroh so much pain10:34
Romstertoo bad they can't just use shadow copy?10:35
Romsteror system restore.10:35
Romsteror even the last backup point.10:35
Romsterhas this hit you to frinnst ?10:37
frinnstnot really. I am restoring a backup for a server we host but not manage10:38
frinnstI get to enjoy it from the sidelines10:38
Romstercan't really blame one company i'm sure others have had similar false positive in the past.10:38
frinnstoh, this is #crux-devel10:38
Romsterlol yeah10:38
*** Feksclaus has joined #crux-devel11:32
*** leo-unglaub has joined #crux-devel12:48
frinnstseems the docbook things are still semi broken13:35
jaegerperhaps the new revdep could use a config file with some default "broken" packages listed, such as syslinux, jdk, jre, thunderbird14:55
jueyeah, sounds useful14:59
jaegerIn the output it could list them with an asterisk or something that indicates there's a problem but that it's expected15:02
juebut on the other side, the packages are not really broken, but most people miss some libs/ports like ffmpeg-compat for jre15:03
jaegerright. I'm just thinking it'll save some time for users who don't understand why they show up in revdep's output even after a rebuild15:04
juejust look at libreoffice, I can see stuff like kdefilepicker or ;)15:04
jaegerIt's a small thing, not too important15:05
RomsterI was against a ignore file.15:11
jaegerwhy's that?15:12
jaegerI personally would prefer it to adding -i blah every time15:12
Romsterif a user goes and adds stuff willy nilly and not realize what they are doing?15:13
Romsterup to you lot.15:13
Romsterideally stuff shouldn't show.15:13
jaegerI don't want it to ignore those packages, I want it to make the output obvious that they're expected to be broken15:13
jaegersomething like "package *" or "package -", nothing big15:14
*** ryuo has joined #crux-devel15:14
Romsterhey ryuo looks like they want some way to show some packages can be expected to fail.15:15
jaeger"they" is me in this case15:15
jaegerfor example, jdk/jre always show in revdep's output15:15
Romsteronly because of it needing ffmpeg-compat15:16
jaegerwe expect them to, I'm just suggesting that revdep have a config file that lists those packages15:16
jaegerit doesn't "need" ffmpeg-compat15:16
Romsterthunderbird i've submitted a bug fix.15:16
jaegerit has libs that link to it that are not required for most daily use15:16
jaegerlibreoffice is the same15:16
ryuoso basically you want something similar to an ignore list?15:16
Romsterlooking into libreoffice15:16
jaegerryuo: similar but not quite. I'm not suggesting those packages be removed from the list, just something added to the line to note them as different15:17
Romsterjaeger> I don't want it to ignore those packages, I want it to make the output obvious that they're expected to be broken15:17
Romster<jaeger> something like "package *" or "package -", nothing big15:17
Romstersomething like that ryuo.15:17
ryuostill process but signal that this may be a known routine error?15:17
ryuoi suppose i could add a static table. or did you have something better in mind?15:18
jaegerI suggested a conf file but Romster doesn't like that idea15:18
Romsterwell it can't be on revdep standard output. it would have to be on stderr with -v levels would work. to show the user and not upset ports list for prt-get use?15:18
Romsterwell a user can add anything and then what happens when it is really broken?15:19
ryuoi would need to add a config parser.15:19
Romsterbut if you want that.15:19
jaegerIt seems unlikely to me a user would be adding shit there for no reason, for what that's worth15:20
ryuobut if you want this, it should be prepopulated with known examples.15:20
ryuobut still editable?15:20
Romsteroy say way would be a /etc/revdep.d/ that a Pkgfile can add to it for jre/jdk/libreoffice...15:21
jaegerThat would be my suggestion, yes. If this is something the rest of you don't really want, then ignore me. It's a small thing15:21
jaegernot revdep.d, the known examples15:21
jaegerjdk, jre, syslinux, libreoffice. thunderbird we can fix.15:21
Romsteri managed to fix every error on my system without that. but *shrugs*15:22
Romsterlooking into libreoffice once i install that in docker15:22
jaegerlibreoffice is a binary port, don't look too hard.15:22
Romsteri know that.15:23
ryuoyou can build libreoffice, but better take a day trip. :)15:23
ryuowhen we've built it for FW, it takes build servers at least half a day to build.15:23
Romsterrevdep was made to check a port for libs not in the ld path.15:23
jaegerI built a source port of it as a PoC, it worked fine. It's just not worth the build time for features people don't use much15:23
ryuosad thing is GCC compiles faster XD15:24
jaegerI built with as many system libs as I could based on what was available in ports; it downloads and builds the rest on its own15:25
Romsteryou basically want to pre-populate a conf file it with all the common binary ports that report broken.15:25
Romsterrevdep managed to resolve nvidia even15:25
jaegernvidia hasn't been a problem for it15:26
Romsterthough there is one libarey in there that wont resolve if gtk3 isn't installed.15:26
ryuothis would be easiest to add with shell aliases and a new option.15:26
ryuobut if it must be a config file i can do that.15:26
jaegerryuo: don't worry about it, it's not important15:26
jaegerthe current revdep doesn't handle them in any special way15:26
jaegerprevious, I should say now, I guess15:27
ryuoi wonder what other slow utilities crux relies :P15:27
Romsterhow about we just test this a bit longer and think it over a bit if it's really needed?15:27
Romsterthen if its a pain then add a feather to solve it.15:27
jaegerfrinnst: I'm running thunderbird here with the changes from the FS bug, seems ok so far15:29
Romsteryeah i saw it after but i lagged15:29
frinnsti havent looked at those bugs yet. hopefully this weekend15:29
Romsteri've run those changes for a couple of days well since i made the bug report.15:29
Romsterand using thunderbird and emailed etc.15:30
jaegerfrinnst: that's fine, just providing some anecdata15:30
ryuoanecdote + data = anecdata?15:30
Romsterlooks like i need to poke docbook some more to15:30
frinnstaye ack jaeger15:30
jaegerryuo: yeah15:30
jaegerryuo: anecdotal data if you prefer15:30
ryuojaeger, do you find a faster revdep more useful than the original?15:32
jaegerIn that it's faster, sure. Did you mean something else?15:32
ryuonot really.15:32
jaegerok :)15:32
ryuoit has been a bit more precise in some cases.15:32
ryuoit found issues on Romster's machine that the original missed.15:32
jaegerDifferent invocation will take some muscle memory retraining but the speed is nice15:32
Romstereven found issues old revdep missed.15:33
jaegerwell, the previous revdep was not super thorough15:33
jaegerit only checked certain files in the port and only recently was support added for checking libs that the port itself contained15:33
ryuowhich is supplied an equal form of.15:33
ryuoi supplied15:34
jaegerwell, if your version found stuff the original didn't, something was different about its algorithm. what was it?15:34
Romsterrevdep -vv syslinux15:34
Romsterthat is a few files hmm15:35
ryuojaeger, it relied on file, and that isn't a very precise way to examine files. i'm not entirely sure.15:35
Romsterlibcom32.c32 => not found15:35
ryuoi know why it does that.15:35
RomsterFound in /usr/ports/opt/syslinux:15:35
Romster  /usr/share/syslinux/efi64/libcom32.c3215:35
Romster  /usr/share/syslinux/libcom32.c3215:35
Romsterthis again...15:35
ryuoRomster, syslinux is a special case.15:35
ryuothose are only used a boot time.15:36
ryuoi've worked with it before.15:36
Romsteri know.15:36
ryuobut it now uses ELF modules for its boot time environment.15:36
Romsterits not resolving because it's hitting the first path15:36
ryuodo you forget? i sort the list of files by their hash.15:36
Romsteryes i remember you said that.15:37
ryuousing zlib's crc32 algorith15:37
ryuousing zlib's crc32 algorithm15:37
Romsterright that's why zlib is linked15:37
ryuonot the best choice, but it works fine for this basic program.15:38
jaegereither way it seems a significant improvement from the shell version15:38
ryuojaeger, well i've come to see what shell is the worst language for performance. many tasks require excessive fork and exec. these are slow.15:39
Romsterits a ton faster. guess ryuo and i can see if we can solve the other few ports that keep popping up15:39
jaegerryuo: jue's tests of dash/bash and mesa3d are perfect indicators of that15:39
ryuoslow performance?15:40
jaegermesa3d's build apparently opens a boatload of shells15:40
jaegernot specifically fork/exec, I guess, but lots of shell starts15:40
ryuoshell is okay for basic tasks, but trying to do anything major is best done in a real language that doesn't have to fork so much.15:41
Romsterlibreoffice doesn't even have a dependency on gtk2 yet it needs it.15:41
ryuoi think even python would be faster.15:41
Romsterand as we found ldd is a shell script15:42
ryuoprobably for portability reasons.15:42
Romstermost likely and locale15:42
jaegeralan wrote a perl version that was much faster than the shell one but didn't do the internal checks I added to the shell version15:42
Romsterwhy didn't that get used?15:43
ryuoi went with C for maximum speed improvement.15:43
jaegerRomster: I can't speak for everyone but my personal opinion was that it needed to be at least on parity with the shell version, feature-wise15:43
jaegerI would prefer C for it as well but didn't have the required knowledge15:44
ryuoNo interpreter overhead and no excessive forking. :P15:44
Romsteri guess features ruled over slight speed.15:44
jaegera slower bash revdep was more useful to me personally than a faster perl one that might misreport more ports as broken than expected15:45
Romsteralso if you want to blame anyone over no RD_VERBOSE=2 i got sick of typing all that and went with lvm2 approach of revdep -vv for level 215:45
jaegerI don't care much about the invocation changing, just have to remember it changed and all's well15:46
Romsterryuo, did all the programming i jsut gave some input and wrote the majority of the man page. and generated it.15:46
Romsterless typing but it is a small change to get used too.15:47
ryuoif not for my existing C experience, this would have taken much longer.15:47
ryuoi've written manual parsers before, etc.15:47
ryuoand gotten used to counting in C.15:47
Romster wasn't expecting this done so fast.15:47
ryuoit's not as major as your other project :P15:48
jaegerI noticed it doesn't process ports in alphabetical order but that's cosmetic, not important15:48
Romsterthat will be big ryuo15:48
ryuothat's due to that fact that i sorted the database by hash.15:48
jaegeryeah, I saw that comment15:48
ryuoits somewhat of a crude hash table.15:48
Romsterok libreoffice sof ar missing gtk and glu15:49
Romsterso far*15:49
Romsterand it sues jre O_O15:50
ryuoRomster, fyi you can use its 2nd mode to just check a subset of the ports.15:50
ryuorevdep libreoffice15:50
Romsterrevdep -vv libreoffice15:50
Romsterknew that15:50
Romsterlibreoffice: jre glu gtk gstreamer-compat ... and it wont ever find the jre library as its not in the ld path15:57
Romsteron top of that libreoffice links to kde stuff
Romsterit wont ever fully resolve unless you ran a kde desktop...15:58
Romsterand it also uses qt3 when it also wants gtk215:58
Romsteraka wtf libreoffice?15:58
jaegerkitchen sink office15:59
ryuopulls out*15:59
Romsternot wrong when someone joked about that.15:59
Romsterargh qt3 appears to be missing freetype on its dependency list.16:00
Romsteri mean i keep finding issues.16:00
ryuowait, you guys still have qt3? :)16:01
ryuowe dropped that ages ago. O_o16:01
Romsteri'll have to do a full sweep of ports one of these days16:01
jaegerryuo: 3, 4, and 5 are all available. I don't know how much 3 is still used, though16:01
ryuoin time you should be able to drop it.16:01
Romsterlibreoffice and hydrogen apparently.16:02
Romsteroen of my ports probably so old it uses qt4 now16:02
ryuoRomster, if i had to guess its used for UI modules. perhaps those should be removed and see if the port still works?16:02
Romsterwow maybe not 0.9.3 -> 0.9.616:03
Romsterguess it supports qt and gtk16:03
Romsterqt3 is broken16:05
Romsterwont compile as it can't find freetype probably needs a patch16:05
Romsteri'm done digging at libreoffice16:05
Romsterone word, mess.16:06
ryuoyes, we dropped qt3 because it was no longer worth the maintenance hassle.16:06
Romsterwe should do the same16:06
ryuosee what if anything still uses it.16:06
Romsteronly hydrogen that i have but i haven't used that in awhile16:07
ryuosometimes you just have to put down an old library, like gtk+1 :P16:07
Romsterlibreoffice works on gtk so.16:07
Romsteroh gawd yeah that16:07
ryuobut some toolkit like NEWT will never become obsolete...16:08
ryuoi mean really. how do you improve a text console? :)16:09
Romsterpro be into that he loves CLI and hates GUI yet he uses xfce16:16
ryuowell you have to use a GUI to use today's web even if you don't like it.16:16
ryuosome things can't be done in text.16:16
Romsteri'll bbl after 3am, zzz.16:18
*** leo-unglaub has quit IRC16:42
*** pitillo has quit IRC17:41
*** ryuo has quit IRC17:47
*** pitillo has joined #crux-devel17:59
*** deus_ex has joined #crux-devel18:40
*** crash_ has joined #crux-devel18:48
*** erdic has quit IRC19:38
*** erdic has joined #crux-devel19:39
*** crash_ has quit IRC20:12
*** crash_ has joined #crux-devel20:12
*** heroux has quit IRC20:44
*** heroux has joined #crux-devel20:47
teK_new revdep says: loader cannot load itself21:33
teK_woot? :)21:33
frinnstworks for me (tm)21:39
teK_I hit "make" and sudo ./revdep21:40
frinnstfrom what tarball?21:40
teK_it finished in <2 minutes, btw21:40
frinnstnot sure if its the "final" version or not, but y u no use opt/prt-utils ?21:40
teK_was that already pushed?21:41
teK_didnt  know and on this box I'm still at 3.0 :-)21:41
frinnstffs dude21:41
teK_spare time is luxury :-)21:42
frinnstyou wont break stuff by upgrading, promise21:43
frinnstwell, nothing major anyways21:43
teK_very funny :P21:44
teK_sudo revdep  1185.77s user 676.16s system 164% cpu 18:52.94 total22:05
teK_the old revdep22:05
*** Feksclaus has quit IRC22:39
*** ryuo has joined #crux-devel22:58

Generated by 2.11.0 by Marius Gedminas - find it at!