IRC Logs for #crux-devel Thursday, 2007-08-09

tilmanprologic: mmh, no. but since git is distributed, it doesn't matter much02:33
tilmanprologic: so i suggest you set up your own repo, and if there's anything cool in it, i'll pull it02:33
clbUpdate from core: 9 Aug 08:05 - hdparm: updated to 7.7. || 9 Aug 08:00 - libarchive: updated to
clbUpdate from core: 9 Aug 08:32 - xfsprogs: update to 2.9.3 || 9 Aug 08:32 - strace: update to 4.5.16 || 9 Aug 08:30 - dhcpcd: update to 3.1.303:55
clbUpdate from opt: 9 Aug 08:38 - mpg123: update to 0.67 || 9 Aug 08:37 - mathopd: fixed md5 sums || 9 Aug 08:36 - dropbear: update to 0.5004:05
*** jjpk has joined #crux-devel04:10
*** treach has joined #crux-devel04:16
*** sepen has quit IRC04:22
clbUpdate from opt: 9 Aug 10:26 - evilwm: improved maximise behaviour05:35
tilmanthe ports tree is busy today05:43
*** sepen has joined #crux-devel05:57
treachpeople coming back from vacations.05:58
tilmanyeah, that was my thought, too05:58
sepenI've vacations on september06:00
tilman(12:43) <@   tilman> the ports tree is busy today06:00
tilmansepen: ^ treach responded to that06:01
tilmanpoppler's is fucking retarded06:35
tilmanChecking for automake >= 1.7...06:35
tilmanTesting automake-1.7... not found. Testing automake-1.8... not found. Testing automake-1.9... not found.06:35
tilmanError: You must have automake >= 1.706:36
tilmanhow about testing 'automake', you idiots?06:36
jjpkPoppler's Saga continues.06:36
tilmanit comes with a "builtin zlib decoder"06:37
*** jaeger has joined #crux-devel07:55
tilmanjaeger: are you aware of poppler's --enable-zlib option?08:28
j^2morning all08:28
tilmanhello j^208:29
jaegertilman: only peripherally. does it need to be added?08:29
jaegermorning, j^208:30
tilmanjaeger: no, it's not *needed*08:30
tilmanthe poppler people ship with their own zlib decoder08:30
tilmani'm not sure what their point is08:30
tilmani doubt theirs is faster than zlib08:31
tilmani'll do some size comparisons08:31
tilmanjaeger: the size benefit is negligable, but some googling shows that their own code is slower than zlib08:44
tilmanother distros use --enable-zlib, too btw08:45
jaegerok, I'll flip it on08:46
jaegerIf that's the case, I wonder why they bothered08:52
tilmani'll ask08:52
Romstermaybe compatability for something?08:53
tilmanto what?08:54
* Romster shrugs08:57
jaegerI suppose at some point I should add SCSI support to the updated ISO08:57
tilmanyou already have scsi for sata anyway08:58
tilmanso it shouldn't hurt the image size much, i guess?08:58
jaegerright, but no low-level scsi drivers08:58
jaegeryeah, it should be much of a big deal09:01
jaegerjust need to add enough drivers to get cdrom support going, then raid controllers or whatever can be modprobed later09:01
jaegerhrmmm... anyone have a patch for wxpython? :)09:18
Romsteri got wxpyton 2.8 working but i haven't messed with the API for buttons and gif09:47
Romsteramule has a wxpython 2.8 patch that waraps the new gif api to the old api, and buttons is a slight change, but thats as far as i got to looking at it.09:48
Romsterjaeger, ^^09:49
jaegerI guess maybe the problem is that I've got wxgtk 2.6 installed but was trying to install wxpython 2.8 :) hehe09:54
jaegerdidn't notice that at first09:54
jaegerwas out of my office for a few mins, sorry for the slow response09:54
Romsterah yes you need both the same..09:56
Romsterah no problem..09:56
Romsteri was refering to API diferences in the built wxgtk button and gif modules.09:59
jaegeraye, understood10:04
Romsterjaeger, i got wxgtk and wxpython in my repo if oyur interested.10:04
Romsterjust added, i had them compiled a few days ago.10:05
Romsterhow long does it take the portdb page to update?10:06
jaegerI wonder if there's a good reason for wxgtk to be at 2.610:06
* jaeger pokes aon10:06
treachhe's probably still in the woods. :P10:08
jaegercould be10:12
Romsteri've wondered that myself other than it breaks things at 2.8 without patching scorched earth amule and xmule.. hmm one other program i can't remember too.10:13
treachjaeger: it's just thursday, I doubt they let him leave until friday anyway.10:14
treach"oh, oh, your in the army now". :p10:14
treachs your/you're/10:14
treachdamn headache10:14
Romsterone of my errors treach sure i'm not infecting you :P10:16
treachnah, I can barely see what I type.10:17
Romsterthat's a bad headache then.10:19
treachyup. going afk for a while.10:20
Romsterfor the upcumming CRUX-2.4 i'd love to see pkgutils use sqlite database and a API for all programs like prt-get to use.10:22
Romstererr upcoming10:22
Romsterinstead of that slow flatfile it uses now.10:23
clbUpdate from opt: 9 Aug 15:18 - poppler: switched to external zlib (--enable-zlib)10:36
jjpk2.4 for 200910:59
jjpkI dare say it is more probable to see pkgutils move over to c, than re-engineering an sqlite db backend + api.11:04
jjpkWhich is unlikely in itself. :s11:05
jaegerhrmm, wxgtk 2.8 didn't make a difference11:05
*** sepen_ has joined #crux-devel13:23
*** sepen has quit IRC13:23
*** sepen_ has quit IRC13:23
*** sepen has joined #crux-devel13:23
sepenanother silly port on my repo: 'tuxpaint' (only for kids and wildboys) xD13:26
sepenreally nicest to have a paint program with sounds13:27
*** Romster has quit IRC13:34
tilmanslow flat file?13:51
tilmanoh, he's gone13:52
tilmananyone who can profile pkgutils to identify the slow code is welcome to do so13:52
tilmanand i'll happily help fixing that stuff13:52
tilmani'm not entirely sure we need a binary package database13:52
jaegerits speed has never bothered me much13:53
tilmanwe can always improve speed13:54
tilmanone obvious idea is to try to mmap the file13:54
tilmanchanging the format is a last resort ;)13:54
jaegerI remember discussion in the past about this, flat file vs. dbm or something13:54
tilmandbm, sqlite, postgres db :D13:54
jaegeroracle :)13:55
tilmanerr, our own simple binary format would probably the first thing to try13:55
*** Romster has joined #crux-devel13:59
treachwell, sqlite is small, and a lot of users probably needs it anyway.13:59
tilmankeep in mind that we'd have to link it statically14:00
treachalso, it would mean that design and maintainence of that part becomes "SEP" to qoute ford prefect. :p14:00
tilmani think it's nice to have a static binary that will work almost anywhere14:00
jaegerRomster: where's your repo? I'd like to take a look at your wx* stuff14:00
tilmanRomster: i replied to your "flat file is slow" claim above, read the logs please14:01
Romsterhmm i can't reach work for you?14:01
jaegerdoesn't seem to be reachable here14:02
Romsterwas offline well is on but was blocked till prologic fixed a rotuer problem after his move but the portdb hasn't updated it yet.14:02
Romsterdidn't think so, i've noticed a few sites being unreachable..14:03
Romster<treach> well, sqlite is small, and a lot of users probably needs it anyway. < first thing i got after pingout and connecting.14:04
tilmanthat's why i pointed you at the logs14:05
Romsteri was expecting staticly linked so it works everywhere, not sure what the binary size overhead would be, although your not staticly linking libarchive?14:05
Romstereh *goes to look*14:05
Romstertilman, right.. well prt-get fsearch is quite slow, i haven't done profilling but i should learn how too..14:10
tilmanott, the guy who did the initial libarchive work opened another bug14:11
tilmanwhich makes prt-get and pkgutils use 'tre' for regex14:12
tilmanit's apparently faster14:12
tilmanso that would speed up fsearch, too14:12
tilmanRomster: if you can do speed comparisons for prt-get fsearch with and without that patch, it would be much appreciated!14:12
Romsterhttpup driver for 'ports -u' also is quite slow it hashes each file, why can't it store the hash and timestamp and only rehash the file if the timestamp is newer than the last time it got hashed, being me5sum be slower than reading the time stamp and comparing that to a cache file.14:13
tilmanbut that isn't really a problem of the flat file db!14:13
tilmanor maybe it is14:13
tilmanprofiling would tell you :-)14:13
tilmanRomster: hint: use oprofile, it rocks14:13
Romstertilman, gladly, if you got some pointers on using profiling, or i should just google.14:13
Romsterk thanks14:13
tilmanand rebuild the app that you profile with CFLAGS="-g"14:15
tilmantry CFLAGS="-O2 -g" first14:16
tilmanif that gives stupid results, use -O0 -g14:16
Romsteri don't need any spefic profile config settings i should be using other than defaults?14:24
tilmanfor gcc or prt-get?14:25
tilmanor what? %)14:25
Romsteroprofile settings.14:26
Romsteropcontrol --start-deamon, then mess with prt-get then stop opcontrol then opreport right, looks simple enougth todo, but first i'll rebuild prt-get with bebug symbools.14:30
tilmanopreport -l /usr/bin/prt-get14:31
tilmancause you don't really care what X11 when oprofile ran14:31
tilmaneg :)14:31
*** cptn has joined #crux-devel14:32
cptnhi there14:32
*** ChanServ sets mode: +o cptn14:32
Romsterhi cptn14:33
Romstertilman, ah good don't wanna see X11 stuff when i'm not looking at that.14:34
tilmanjjpk: did you see arnuld's new postings?14:49
jjpkarnuld ze artiste14:52
cptnRomster: fsearch reads the .footprint files14:52
cptnso it doesn't touch pkgutils db14:52
cptnbecause usually when you look for a file, you don't have it installed yet, so it wouldn't be in the pkg DB14:53
Romsterno bloody wonder i thought it would use the cache file from prt-get cache, to update it...14:54
Romsterwell i'll test profiling anyways see what i get.14:54
tilmanprologic: i honestly hope that doesn't offend you. and really git makes this so easy :)15:02
prologictilman, about access to the pkgutils upstream repo ?15:05
prologicnot at all :)15:06
prologicI'll do some hacking here and you're welcome to pull in15:06
tilmanif you want to any big changes (api overhaul or somesuch), let's talk about them before15:07
jaegerRomster: your wxpython port breaks in fakeroot15:08
prologicnah no need to chnage anything :15:08
jaegercreating /usr/include/wx-2.8/wx/wxPython15:08
jaegererror: could not create '/usr/include/wx-2.8/wx/wxPython': Permission denied15:08
prologicas long as libpkgutils works well enough with something15:09
prologicI'd like to (finally) create a graphical pkg mgt tool15:09
treachziomg, heresy15:09
tilmani hope you'll create python bindings eventually :)15:09
Romsterwhat ... isn' it using the --prefix...15:09
prologicdefinately - first thing I'll do :)15:10
tilmanwhen the api is stable-ish15:10
prologicpython bindings to libpkgutils15:10
Romsteroh wxpython hmm that whould have --root15:10
prologiclibpkgutils is designed to do everything pkg(add,rm,mk) does right ?15:10
tilmannot pkgmk15:11
prologichow would an app go about insturctiong a port to be built then ?15:11
tilmanthe approach is identical to bd2's pkgutils-c15:11
tilmanbut with a usable library15:11
tilmanoh, i'm being mean.15:11
tilmanprologic: same way prt-get does it15:12
prologicwhich is ?15:12
prologiccalling the pkgmk script ?15:13
tilmanexec ("pkgmk") basically15:13
prologicwell then pkgmk needs to have enough return codes to be useful to a gui15:13
tilmanthough you could popen it and display the output in a GtkTextView15:13
prologicor a curses app or anything15:13
tilmantbh i'm not sure whether that's a problem right now15:14
prologicit would also be useful if we could store build times of ports somewhere15:14
prologictilman, what's that, a gui pkg mgt tool ?15:14
prologicI never said it was a problem - just nice to have :)15:14
tilmani mean:15:14
tilmani'm not sure we have much to improve on pkgmk's exit codes15:15
prologicsepen, xterm is over there ------>15:15
tilmani just don't know :)15:15
tilmanprt-get seems to do the right thing most of the time15:15
cptntilman: different exit codes for footprint missmatches or md5 would be handy15:15
cptnthen the frontend could directly suggest the proper action15:15
prologicyes this is what I meant15:15
prologicthank you cptn15:15
prologicI don't believe it currently does so15:16
cptnalso, if there was a 'footprint missmatch with only added files' users could define a policy to always ignore those15:16
prologic*nods* nice idea!15:16
prologicalso does ports have decent return codes for it's errors ?15:16
Romsterjaeger, it has --prefix=$PKG/usr i'll have to look into it..15:17
jaegermaybe it needs --root as well?15:17
Romstermaybe i'll look in the file in a bit..15:18
prologicmost python apps with you only need15:18
prologicpython install --root=$PKG15:18
prologicyou generally don't even need the prefix (in most cases)15:18
prologicbut afaik you _do_ need to specify an installation root with --root or --home15:18
jaegeryeah, gonna test it with --root15:21
Romsterit appeared it was going into the work/pkg/...15:37
Romsterjaeger, when you find it can ya paste the corection in here i'm gonna nap i'm tired as.15:45
jaegerI imagine --root will fix it but it's still building so not sure15:45
*** cptn has quit IRC16:11
tilmanRomster: got any results yet?16:21
prologictilman, isn't doing a pkgutils in C a waste of your time16:50
prologicand waste of bd2's time16:50
prologicsince he has already done significant work on a C version16:51
prologicjust a curious question is all16:51
prologicseems to me that you're both doing the same thing16:51
tilmanshort answer: i think it's worthwhile16:51
tilmanlong answer: sorry, maybe another time16:51
tilmanprologic: one thing. bd2 and me seem to have different opinions on a number of critical issues.16:52
prologicoh well16:53
prologicnot anything I can do about it16:53
prologicjust seems like a waste of code and development time :/16:53
prologicon both of you16:53
prologicor just him16:53
prologicsince we will probably never use his version16:53
treachnobody stops you from using it.16:54
treachif you think it's a good idea.16:54
prologicI don't care really16:55
prologicI tend to use officla stuff only16:55
prologicand rhe related xorg/xfce/gnome ports16:55
*** jaeger has quit IRC17:00
tilmanthis touchpad is freaking me out17:42
*** treach has quit IRC18:05
prologic+1 real names18:49
prologicI'm James Mills18:49
prologicI cannot still post to crux lists :/18:49
prologichas this been fix ?18:49
prologicbut +1 on real names anyway18:50
prologicRomster will +1 to that too18:50
* Romster yawns19:12
Romsterso defently i've given up on replying to anything that goes to @crux.nu19:13
prologicwell I can't yet19:15
prologicstill get timeout issues19:15
prologicmaverick needs to fox this :/19:15
prologicneways I gotta go19:15
*** jjpk has quit IRC19:17
*** jaeger has joined #crux-devel22:39
*** jaeger has quit IRC23:25

Generated by 2.11.0 by Marius Gedminas - find it at!