IRC Logs for #crux Thursday, 2006-08-24

*** pitillo has joined #crux00:09
*** mike_k has joined #crux01:24
*** predatorfreak has quit IRC02:14
pitillogood morning all, cptn, is there a way to get the validator script? (is a extension of checkdups?)02:33
cptnhi pitillo02:33
cptnjue is currently working on a validation framework02:34
cptnit should be ready within a few weeks02:34
cptnthe validator scripts I use are available from
cptnbut I don't recommend using them :-)02:35
pitillowel, I take note, but I?ll check it firts :) Thanks.02:36
Hancptn, is that for regression?02:42
cptnjue's scripts will probably used for regression checking02:45
cptnif that's what you mean :-)02:45
rxiHan: did you see my message in irc the other day about the eet port? cos if you didnt ill email it to you02:53
HanIndeed, good idea.02:56
*** _Viper_ has joined #crux02:56
rxiyeah only picked it up after installing crux-ppc 2.2 alpha02:57
rxiwhich doesnt have cvs installed by default02:57
*** Viper_ has quit IRC02:58
*** _Viper_ is now known as Viper_02:58
marowhy is it that local files are md5sum'ed in pkgmk?02:59
HanBecause it isn03:00
HanBecause it isn't useful.03:00
marocompelling argument03:00
cptnmaro: local files?03:00
cptnlike patches, which you get from someone's httpup?03:01
maroyeah, and configs, init scripts, ...03:01
aonthe web servers you get them from never get compromised!03:01
aonno, wait03:01
Han*g* they are.03:01
cptnaon: :-)03:02
maroaon: err?!03:02
HanI thought he meant the contents of tarballs.03:02
maroaon: if they're compromised, wouldn't they be able to change the .md5sums too?03:02
aonbut maybe they'd forget? :)03:02
cptnwe should md5sum the md5sum03:03
cptnand keep it hidden!03:03
maroand why isn't the Pkgfile md5sum'ed too, since it can contain malicious commands?!03:03
maroseriously though, security isn't really a valid argument in my POV03:03
aonalso just md5sum isn't clearly enough03:03
aonsince it can collide03:04
cptnevery hash collides03:04
aonbut not necessarily with the same contents03:04
marothe proper solution is to implement the checksums at the httpup level (if it isn't already) just in case of corrupted downloads03:04
HanLike with openbsd, they use _3_ hashes!03:04
aonand size too, right?03:04
copworkerpotheads :)03:04
cptnwe can easily switch to a stronger hash in the future03:05
cptnalthough I'd rather just have one03:05
marobut *please* get what I'm saying :)03:05
copworkercrux is for simplicity with the right poison of security03:05
cptnyou define a reason why md5sums are used, and then claim that they're not effective at it03:05
cptnwhich is probably true03:05
maroI'm not arguing about the security, other than that it isn't a valid argument to md5sum the local files03:05
*** bd2 has quit IRC03:06
copworkerthe md5sum basically is to test for corrupted downloads03:06
maroyes, that's how I see it03:06
HanWell then, lets use chksum, much lighter ;-)03:07
cptnto a packager, md5sum missmatches usually mean that you have to review the build03:07
aonyou should probably look at svn if the md5sum changes without any reason03:07
maromy point is that it's useless to check the local files, and it's annoying (and noisy in the commits) to update the md5sums every time you change a local file in the port03:07
cptneither because you changed a local file yourself (needs review), or because upstream silently changed the tarball03:07
HanI just wish I could force wget _not_ to download html files!03:07
aonsince it's unlikely that the compromiser will make an 'modified for rootkit' commit03:07
cptnaon: "add xgl support" :-)03:08
aon+ rm -rf /03:08
rxiHan: theres a switch for it03:08
HanI still think it's a bad idea to build ports as root.03:09
marocptn: a packager needs to review his own changes?03:09
*** bd2 has joined #crux03:09
marocptn: what I'm suggesting is that only remote files are md5sum'ed03:09
marosince they're the ones you can't really be sure about the integrity of03:10
cptnmaro: and your main argument is that it'll make commit diffs easier to read?03:10
marocptn: and package maintenance/derivation easier and more fun03:11
marothe chicks will like it too03:11
maroand all the cool guys do it03:11
maroetc. :)03:11
rxiHan: -R.hmtl03:12
cptnrxi: that's just the suffix03:12
rxicptn: that will reject any files with the .html extension?03:12
marocptn: in my view, a package needs to review the build when *any* file changes03:13
marochanging more files unnecessarily for each commit doesn't make it better03:13
cptnany file -> md5sum everything in source=()03:14
cptnI'll have to think about it03:14
rxiHan: hehe03:14
marono, any file = any file in the package dir03:14
maroif it's the remote source, that file is .md5sums03:14
marocptn: thanks :)03:15
HanWhat concrete problem would that solve?03:16
maroit's called "improvement"03:17
maroyou should get yourself familiar with that term03:17
HanHmmm whimsical. :-)03:17
rxifight! fight! fight!03:17
cptnmaro: it's a trade of03:17
cptnspecial case handling in pkgmk03:18
cptnvs. less commit noise03:18
cptnhaven't understood the "easier derivation" part yet03:18
marocptn: I had imagined to completely drop the inclusion of local files in ${source[@]}03:18
cptnthat would be more consistent03:19
cptnbut doesn't feel natural03:19
cptnwhy would a file you use as source to build a binary package not be in source=()?03:20
marobecause it's local03:20
maroand usually the files included aren't really "source code"03:20
cptnremote files aren't "source code" per se either03:21
marono, but in practice they usually are03:21
cptnsome for local files03:21
maro(and it's irrelevant)03:21
cptnthey're usually patches -> source code03:21
maronow we're discussing renaming the source array to something different?03:22
Hanmaro, is that `improvement'?03:22
maronot in my book03:22
marowhich is why I call it irrelevant03:22
cptnit sounds like changing something for the sake of change03:23
marowhat I'm suggesting is that source is the upstream sources (possibly other remote files, e.g. for the linux headers)03:24
maroand the files included in the port dir are "local files"03:24
cptnokay, here's an example of something I'm doing just now:03:25
maro(and not md5sum'ed since it's useless)03:25
cptnI had made a patch, tested it as local file03:25
cptnthen I uploaded it to a webserver to exchange it, and adjusted the Pkgfile03:25
cptnwhy would it be a "source file" in one case, but not the other?03:25
cptnwhy should it be treated differently?03:26
marocptn: because it's now on a remote server03:26
maroso you can't be sure of the integriry03:26
marowhereas when your local files start to corrupt, you're screwed anyway03:27
cptnso I'll end up with an md5sum missmatch03:27
cptnbecause it's not listed there anymore03:27
cptnerr, it's listed now03:27
cptnit wasn't before, when it was local03:27
maroyou'll end up with a corrupted /bin/cp or whatever03:27
maromd5sum all the files on your system if you want, I'm just saying that checking integrity for local files in pkgmk is solving a problem at the wrong level03:28
cptnin the moment I change source=() to list the very same patch a remote file, I'd get a md5sum missmatch03:28
cptn*as a remote file03:28
marohow often do this happen in practice?03:29
maroand how is it harder to solve than it is currently?03:29
cptnit's not about solving, it's about understanding the missmatch03:29
cptncurrently, there's a missmatch if you change any file from sources03:30
maroyes, and that's the way it should be03:30
maromy argument is that local files shouldn't be part of "sources"03:30
cptnwe've reached an egreement03:30
cptnyeah, I don't agree with that03:30
marowhich has the additional benefit that pkgmk won't have to guess whether a file is local or remote03:31
*** predatorfreak has joined #crux03:36
MukundaHow do I actually install packages with the pkgutils stuff?03:40
cptnthere's a section in the crux handbook03:40
cptnit's recommended reading around here03:41
MukundaGood idea, not thinking.03:46
pitillocptn, who made the validator script? I?m thinking in add comments to the scripts inside the package. Nice to see what them do before reading them all. (Only a opinion)04:18
pitilloummmm the name of the script gives some info about it. May be it isn?t a good idea.04:19
cptn wrote them04:21
cptnbut as I said, Jürgen Daubert is working on a new version04:21
cptnhe also announced it to crux-devel some time ago04:21
cptnso my prt-validator is likely gonna die04:22
pitillowell, hope I can see the newer version. I?m a bit new to the devel-ml. I wait to that version and see if I can do something to help04:22
pitilloadvance level in shell scripting and c here. Nice jobs.04:25
pitilloit?s true... the temple of knowledge. :)04:26
pitillo:O first crash of E.... brbrbrbr04:28
*** treach has joined #crux04:33
*** treach has quit IRC04:45
*** treach has joined #crux04:46
rugekanyone here know an alternative location of the libstdc++ files?04:50
cptnrugek: ports -u?05:14
cptnif it's libstdc++-comat05:14
cptnlibstdc++ itself comes from gcc05:14
rugekI meant libstdc++-compat05:18
rugekbut pers repo is down05:19
rugekthe same problem comes with pkgutils05:19
* rugek is afk for 30min05:21
*** jue has joined #crux05:23
*** ChanServ sets mode: +o jue05:23
cptnrugek: the post name changed05:27
cptnrugek: that's why I suggested running ports -u05:27
marohi jue :)05:27
cptnhey jue05:27
mike_khi, jue. Have you ever considered to follow FHS in your ports? There was some talk on FHS compliancy recently (regarding man pages)...05:29
mike_kAny cons?05:30
*** kingruedi has joined #crux05:32
juemike_k: to be honest I don't really care about FHS, ATM I see no point to change our filessystem layout05:47
mike_kI mean using /srv/www or even /srv/www/apache to allow other web server to have it's files in /srv/www/otherhttpd. Same with mysql.05:51
jueinstead of /var/www ?05:53
mike_kI know, it's been a good old tradition though05:54
jjpkThen again, there really is nothing stopping you from manually changing the www root.05:54
juemike_k: sorry, what's the point in using /src/www instead of /var/www ?05:54
treachgrr, why bother with *another* dir at / level?05:55
mike_kjjpk: I'm using that way jue's modifyed ports for quite some time05:55
aonthere's also /media etc. :)05:55
* treach takes out aon and shots him05:55
* treach remembers that aon is a nice guy, and revives him.05:56
marojue: all the cool guys do it!05:56
aonhello again05:56
mike_kI am not insisting, but standarts _sometimes_ make life easier... maybe not this time, as I see05:56
marohi aon :)05:56
aonit'd probably be good if the S in FHS was for real05:58
aonbut even now you have to always find out whether the stuff is somewhere under var, or perhaps in /usr/local/apache2, or even in opt05:58
aon/var is probably the most common and i don't see why it needs to be changed05:59
maroaon: arch uses /home/httpd x)05:59
mike_kjue: the point is 1) this what FHS 'recommends'  2) that way we can honor other web/db servers with dedicated place for their data06:00
aonwhat kind of a maniac runs multiple general-purpose web servers on one machine? ;)06:00
treachyou mean you can't dedicate a /var/www partition?06:00
maroaon: don't say that while jue's around ;)06:01
mike_kaon: if everyone is going to do it in their own way - we'll end up with such a mess you've described06:01
maromike_k: which is why we just use /var/www06:01
marowooha! magic!06:01
mike_kmaro: if it's a 'tradition vs standart' question...06:02
treachsome "standards" are just fads. tradition *is* the standard (way of doing things). ;)06:07
rugek-> Internal Server Error06:18
marocptn: weird sources ;)06:24
treachhmm. what happended to the "audio" group in the later udev ports?07:27
cptnsound is overrated07:27
rxiyeah i heard that too07:30
treach"25-lfs.rules:KERNEL=="controlC[0-9]*",   GROUP="audio",  NAME="snd/%k" etc, etc.07:30
treachbut effectively the permissions are root:root07:31
cptnand the group exists?07:31
treachhmm, you might be on to something there.. I 've forgotten to restore it. :D07:31
treach(fresh install)07:32
treachok, now the group exists, and udevd is restarted, but the permissions are still the same..07:35
*** treach has quit IRC07:36
*** Klavier has joined #crux07:37
*** sen has quit IRC07:56
*** sen has joined #crux07:57
*** kingruedi has quit IRC07:58
*** kingruedi has joined #crux08:00
deus_exCan qt3 and qt4 coexist on a same system?08:02
deus_excptn: Thank you.08:03
cptnwe don't have a qt4 port right now though08:03
deus_exYou had it in your repo at some point, right?08:04
cptnyes, I can tar it up if you want to08:04
deus_exThat would be great :)08:05
cptnI used it in my previous job, but not anymore...08:05
*** treach has joined #crux08:07
cptnwow, one ugly Pkgfile08:07
cptnit's adapted from qt3 which is a quite complex Pkgfile already08:07
cptnand I didn't take the time to make it look clean08:07
deus_excptn: Are you from Switzerland?If yes, I apologize for Ana Ivanovic kicking Martina Hingis ass at Montreal :)08:08
deus_exThanks for the port.08:08
cptnheh, I am08:09
cptnbut I don't really follow what Martina does...08:09
deus_exIf you are interested:
aonis there a reason why j2re is at 1.5.0_07 instead of _08?08:21
nipuLhmm, guess my mail server rules are too strict, all mail from crux is bouncing08:21
aonah, apparently there already is a ticket08:21
*** Klavier has quit IRC08:25
cptnI guess there's an arch convert submitting bugs :-)08:27
cptn"I've attached a Pkgbuild..."08:27
treachyay. :)08:27
*** Klavier has joined #crux08:32
*** clare has joined #crux08:55
*** sen has quit IRC09:06
*** treach has quit IRC09:07
*** treach^ has joined #crux09:08
*** Brzi has joined #crux09:18
*** kingruedi has quit IRC09:23
*** Klavier has quit IRC09:24
*** morlenxus has quit IRC09:48
*** morlenxus has joined #crux09:49
*** kingruedi has joined #crux10:09
*** predatorfreak has quit IRC10:17
*** j^2 has joined #crux10:39
j^2hey all10:39
jjpkhi j^210:43
j^2how you doing?10:44
jjpkA bit tired, and trying to figure out a project topic.10:48
tilmanwhat about "how to solve cancer"10:51
cptnjjpk: have you considered writing a window manager?10:56
tilmanoh,that reminds me10:56
cptnI heard there's some demand there :-)10:56
*** Brzi has quit IRC10:57
tilmanour gtk bug has been fixed in the latest release11:05
cptnare we credited ?11:06
tilmannope, no other replies to my thread either11:07
aonhm, who needs another wm? :)11:07
*** laod_ has quit IRC11:09
*** laod_ has joined #crux11:12
*** sen has joined #crux12:08
*** Han has quit IRC12:09
*** Han has joined #crux12:11
*** jue has quit IRC12:20
*** pinkpussy has joined #crux12:38
pinkpussyin wich packet is the xdd command?12:39
tilmanXpert Design and Diagnostics?12:41
tilmanKnowledge Representation Laboratory?12:41
tilmanX-ray Development and Demonstration beam line?12:41
pinkpussyyes xdd12:41
tilmanno idea what that is12:42
pinkpussyvim needs this for hex12:42
tilmanis the hex converter12:42
tilmannot xdd.........12:42
pinkpussyshame on me12:43
treach^Han: /usr/share/mimelnk/application/x-mplayer2.desktop in kaffeine collides with kdelibs.12:54
*** kingruedi has quit IRC13:03
*** mjeq has joined #crux13:13
*** brointhemix has joined #crux13:16
j^2crazy ass question....ever used a tablet pc?13:23
blizzis there a script which will find files under / which do not belong to any port?13:30
blizzexcepting /home and stuff..13:30
Auge^write on :)13:32
qidblizz: I was working on such a script13:33
blizzAuge^?? du hier? :-P13:36
qidit mostly works, the problem is there are a fair amount of things that don't belong to any port for fairly obvious reasons13:36
qide.g. everything under /dev, /proc, /sys, /lib/modules, /usr/src, etc.13:36
blizze.g. stuff generated by post install scripts?13:36
Auge^blizz: natürlich ;P13:36
blizzcool ;-)13:37
Auge^nuja, crux ist halt so schön bsd-like... das muß mal probiert werden =)13:37
blizzoh ja, habs shcon seit 2 jahren glaub ich jetzt.13:38
Auge^<- newbie13:38
Auge^2 days13:38
blizzviel spass und erfolg dann :-)13:38
Auge^danke :)13:39
qidI wrote a ruby script that basically loads the pkg db into memory, then runs `find /bin /etc /lib /sbin /var -type f 2>/dev/null` and checks everything against the db13:39
qidso dev, proc, sys, boot, opt, and so on are filtered out, but it'll still report all of /lib/modules and /usr/src which is probably not intended13:40
qideasiest way to deal with those will probably be processing in ruby13:40
tilmantakes two seconds to fix13:41
tilmanmaybe 313:41
qidI know, I'm not saying it's hard13:42
blizztilman, scriptkiddie! :-)13:42
tilmanblizz: i run yarv!! i'm a bytecode kiddie now!13:42
qidI was hoping to make find do as much of the work as possible because I figured it would be faster, but I suppose ruby has to do processing on everything anyways13:42
tilmanif it's a ruby script, you should use Find.find13:43
tilmanand not call find(1)13:43
blizztilman, lol, nice13:43
blizztilman, for how long is there a working yarv release? didn't follow its progress..13:48
blizzin fact im such a ruby noob13:48
tilmanwell, i lied, i don't run yarv13:50
tilmanthere's no stable release either13:50
Hantreach^, odd, not over here.13:50
treach^ok, I guess it could be related to that I'm using 3.5.4 here. Didn't think of checking that.13:51
qidtilman: didn't know there was a Find module13:52
blizztilman, grr :-)13:56
*** clare has quit IRC14:05
qidheh... I called the list of directories to ignore 'exclude'... so to check if a dir is ignored, I have to call 'exclude.include?'14:06
tilmani'd make exclude an array of regular expressions perhaps14:07
tilmanexclude.each do |excl| if file.match(excl) ... }14:07
qidI considered that but figured it would be too complicated14:08
tilmanyour approach will work, too, i think14:09
qidrunning multiple regex's against every file name on the system would take a while14:09
tilmani didn't think of find passing each directory to the block as it traverses the tree14:09
qidif and exclude.include?(path) { Find.prune }14:10
qidthat's semi-pseudocode14:10
tilmanyes, that looks good14:10
*** clare has joined #crux14:21
qidcrud, the find module considers a symlink to a directory to be a directory14:27
blizzis there a possibility to upgrade a reiserfs (v3) to reiser4?14:30
bd2qid, tilman:  - fast enough. ;-)14:30
blizzAuge^, i always miss ffs (the filesystem)..14:30
bd2and with refactored pkgutils I've rewritten the patch to use std::set_difference, which is little faster.14:31
Auge^blizz: hehe ;)14:33
*** mjeq has quit IRC14:45
qid <-- why wasn't the same change made for cdrom.rules?14:48
qid/var/lib/pkg/db not in db14:52
tilmananyone familiar with libtar?14:54
bd2I'm (smattering)14:54
bd2yeah, BASICS knowledge -> not a guru :-)14:55
tilmanbd2: i'm messing with the th_read() loop in pkgutil::pkg_install14:55
tilmanwhen i don't do anything with the tar handle in the loop body, the next th_read() call fails14:55
tilman!!! tar header checksum error14:55
tilmanany idea what's causing that?14:56
tilmanor rather, how to pacify libtar? :D14:56
bd2a moment...14:56
bd2hm... maybe it's a bug in libtar, as I don't see anything like th_skip (like archive_read_data_skip in libarchive)15:01
tilmansucks to call tar_extract_file() to /tmp15:03
tilmanor something15:03
bd2you can try using TAR_IGNORE_CRC in tar_open, but I doubt that it's good idea15:03
bd2it's last argument to the tar_open call15:04
bd2tilman, are you trying to implement INSTALL rule to pkgadd?15:04
tilmani'm done, except this bug15:05
tilmanoh well15:06
tilmantrying to extract to /dev/null now15:06
tilmanseems to work15:08
* treach^ sends an angry thought to people whos code doesn't compile if you leave the frame-pointer out.15:08
tilmanit fucked up /dev/nulls permissions :D15:09
bd2haha :-)15:09
treach^tilman: libtar is the one responsible for turning /dev/null into a file every now and then? :p15:10
tilmani'm messing with pkgutils15:10
tilmanc++'s crappyness pushed me to the edge of sanity15:10
tilmanand libtar's retardedness pushed me over it15:10
bd2tilman, - the tool I'm using for ARM builds. it's using libarchive (without any temporary files). It's using /etc/pkgadd.conf's INSTALL rule (my pkgadd.conf syntax is little different, it's using ACTION "regex", which premits using spaces inside regex). Take a look if you're interested "how libarchive usage looks like". :-)15:11
*** mike_k has quit IRC15:14
bd2tilman, hey, look: int tar_skip_regfile(TAR *t);15:25
tilmani've seen that. the _regfile_ bit worries me, but i didn't try it yet15:27
bd2        if (!TH_ISREG(t)) { errno = EINVAL; return -1; }15:28
bd2just don't check the returing value15:28
bd2the above code from libtar/lib/extract.c:tar_skip_regfile15:29
tilmanwill try that tomorrow15:29
tilmannow, sleep.15:29
bd2good night15:30
*** hackedhe1d has joined #crux15:50
*** hackedhe2d has joined #crux15:55
*** lordalbert has joined #crux15:57
qidis there a reason why packages e.g. sshd syslogd klogd don't include pid files and log files as part of the package?15:58
lordalbertcrux it's really difficult? Now i'm using Archlinux, but a friend tells me that crux is more difficult than archlinux. it's true?15:59
treach^crux is really, really hard.16:00
bd2qid, maybe bacause of /etc/rc:(cd /var/run && /usr/bin/find . -name "*.pid" -delete) ?16:00
bd2thus, they are just runtime files, not a package really16:00
treach^only real 1337 people find themselves able to use it, so we usually recomend people who haven't used slack or something like that for a decade to get ubuntu.16:01
lordalberttreach^, i'm using Arch16:06
treach^I read that.16:06
lordalbertarch is crux based16:06
*** hackedhead has quit IRC16:07
pitillolordalbert, have you taken a look to the CRUX web page? read a bit and take conlusions by yourself16:07
treach^iirc judd claims it never was, though it seems to have a lot of concepts in common.16:07
j^2if i have a dual proc machine, how can i tell that linux sees it?16:07
treach^lordalbert: I'm not really serious, because asking if something is "hard" is a bit of a dumb thing to ask.16:09
bd2 there are some note about CRUX16:09
*** hackedhe1d has quit IRC16:10
*** pinkpussy has quit IRC16:23
*** gaqzi has quit IRC16:56
*** gaqzi has joined #crux16:57
*** gaqzi has left #crux16:58
*** gaqzi has joined #crux16:58
*** treach^ has left #crux17:07
*** treach has joined #crux17:12
*** treach has left #crux17:32
*** kingruedi has joined #crux17:59
*** lordalbert has quit IRC18:06
*** pitillo has quit IRC18:48
*** Plazma has left #crux19:11
*** kingruedi has quit IRC19:12
*** sen has quit IRC19:22
*** thrice` has joined #crux19:22
*** sen has joined #crux19:22
*** Romster has joined #crux20:09
*** Romster has quit IRC20:20
*** thrice` has quit IRC20:38
*** clare has quit IRC21:07
qidtilman: is your RoR port being maintained?22:00
*** mrks has joined #crux22:21
*** predatorfreak has joined #crux22:28
*** mrks_ has quit IRC22:35
*** morlenxus has quit IRC23:05
*** morlenxus has joined #crux23:06
*** predatorfreak has quit IRC23:47

Generated by 2.11.0 by Marius Gedminas - find it at!