T2 IRC Log: 2006-10-29

This is the log as captured by an IRC bot in the channel. The statements are those of the individual people and might not neccessarily reflect the policy and legal rules as set forth by the T2 SDE Project.

« prev | next »

--- Log opened Sun Oct 29 00:00:12 2006
00:31 < pamdirac_t2> I'm using/studing 6.0.0
00:43 * Stelz is away: Sleep is good. Zzzzz
01:10 < emte> who are you asking?
01:12 < emte> weird ... there must be magic tags somewhere ...
01:13 < emte> perhaps its the numbers ...
01:14 < emte> AHA!
01:17 < pamdirac_t2> excuse me
01:17 < pamdirac_t2> I ask to *]
01:17 < pamdirac_t2> :D
01:19 < CIA-9> ragnar * r20544 /trunk/package/x11-ng/luminocity/luminocity.conf: * luminocity: X11R6 -> X11
01:23 < emte> i am playing with the dev branch
01:25 < emte> lets see if what i have now does anything :)
01:25 < pamdirac_t2> ah
01:26 < emte> i "think" i finally figured out how to make pkgsel work :)
01:27 < pamdirac_t2> ahha
01:28 < emte> simular to some initrd types it uses a numerical priority
01:29 < emte> as for the contents ... i am still working on that
01:29 < pamdirac_t2> ok
01:30 < pamdirac_t2> mmm
01:32 < pamdirac_t2> I'm sleeping on the keyboad
01:32 < pamdirac_t2> night
01:32 < emte> waffle face :)
01:32 < emte> night
01:32 -!- pamdirac_t2 [n=pamdirac@88-149-140-37.f5.ngi.it] has left #t2 []
02:50 < emte> hmm odd error out
02:50 < emte> but not unexpected
02:03 -!- idealm [n=ideal@218.82.57.80] has joined #t2
02:35 < emte> hmm
02:35 < emte> is there a way to append or override the patch src dir by default?
02:44 < emte> or specify specific patches?
02:45 < emte> i'd like to use the ones i know work before i sort out the dups or specifics
03:25 < emte> i've no idea how glibc selects the right patches ....
03:26 < emte> unless the other ones are applied even when your not building that arch ...
03:53 -!- idealm [n=ideal@218.82.57.80] has quit [Remote closed the connection]
03:55 -!- idealm [n=ideal@218.82.57.80] has joined #t2
04:12 < emte> hmm
04:13 < emte> guess everyone is at hallows parties
06:50 -!- idealm [n=ideal@218.82.57.80] has quit [Remote closed the connection]
06:53 -!- idealm [n=ideal@218.82.57.80] has joined #t2
06:55 -!- mtr [n=Michael@pD9E11F7F.dip0.t-ipconnect.de] has joined #t2
08:09 -!- idealm_ [n=ideal@58.33.58.84] has joined #t2
08:21 < sepp> emte: all .patch files in a package dir are applied
08:27 < emte> that is odd then
08:28 < sepp> why?
08:28 < emte> why would you apply 64bit patches or sparc patches when your building for a 32 bit system
08:28 -!- idealm [n=ideal@218.82.57.80] has quit [Read error: 113 (No route to host)]
08:28 < emte> for example
08:29 < sepp> it does not break things, so it is no problem
08:30 < sepp> if you want to apply a patch only on some condition you can do that in the .conf
08:30 < emte> that will become very ugly very quick
08:31 < sepp> i don't mean glibc architecture patches
08:32 < emte> most of the ones i have will be arch
08:37 < sepp> what kind of patch?
08:41 < emte> i am doing an initial "port" of Familiar for the h36/37/38 series devices
08:45 < sepp> iPaq ?
08:45 < emte> familiar is mosty IPAQ, yes
08:45 < emte> mostly*
08:46 < sepp> :)
08:46 * Stelz is back (gone 09:02:58)
08:46 < Stelz> moin all
08:47 < sepp> moin Stelz
08:47 < Stelz> hi sepp
08:49 < emte> so i guess i'll have to invent a way to handle specific patches without "cluttering" the main packages up
08:50 < sepp> no
08:51 < sepp> if you have a h3? patch put it into the pkg dir
08:52 < emte> in a worst case if i have 5 packages per machine type
08:52 < emte> patches*
08:53 < emte> and the one i am currently working on means 3 different configurations
08:53 < emte> taht is 15 more patches
08:54 < sepp> 5 patches for 5 packages?
08:54 < sepp> i don't see the problem
08:54 < emte> no
08:54 < emte> 5 patches per machine for one package
08:54 < sepp> if you have 5 patches for h36 glibc put them in 1 file
08:55 < emte> most people do not appreciate monolithic patches
08:55 < emte> but that is an option
08:56 < sepp> well
09:00 < CIA-9> msluis * r20545 /trunk/package/audio/vorbis-tools/vorbis-tools.desc:
09:00 < CIA-9> * Changed vorbis-tools description to lossy instead of lossless.
09:00 < CIA-9> * Changed vorbis-tools url, the old one was a dead link.
09:02 -!- misl [n=chatzill@84-104-172-187.cable.quicknet.nl] has joined #t2
09:02 < misl> hi folks
09:02 < emte> oi
09:03 < sepp> hi misl :)
09:03 < misl> hi emte, seppi
09:10 < sepp> bbl
09:24 < rxr> re
09:24 < rxr> moin sepp
09:24 < rxr> hi misl and emte
09:24 [Users #t2]
09:24 [@ChanServ] [ CIA-9 ] [ LMJ ] [ rxr ] [ valentin]
09:24 [ _Ragnar_] [ emte ] [ misl] [ sepp ] [ zod ]
09:24 [ Baldzius] [ idealm_] [ mtr ] [ Stelz]
09:24 -!- Irssi: #t2: Total of 14 nicks [1 ops, 0 halfops, 0 voices, 13 normal]
09:25 < misl> hi rxr
09:25 < emte> oi
09:28 < rxr> emte: if you still have open questions it is a good time to ask them now .-)
09:28 [Users #t2]
09:28 [@ChanServ] [ CIA-9 ] [ LMJ ] [ rxr ] [ valentin]
09:28 [ _Ragnar_] [ emte ] [ misl] [ sepp ] [ zod ]
09:28 [ Baldzius] [ idealm_] [ mtr ] [ Stelz]
09:28 -!- Irssi: #t2: Total of 14 nicks [1 ops, 0 halfops, 0 voices, 13 normal]
09:29 < emte> they are all open, i am kinda wandering through them
09:29 < emte> at this moment i am trying to figure out how you specify specific versions of a package
09:39 < emte> hmm ... only vague ref i can find in the defaults is linux24 or linux26
09:39 < emte> i'll sleep on it is guess
10:05 < CIA-9> rene * r20546 /trunk/package/security/opencdk/opencdk.desc: * updated opencdk (0.5.10 -> 0.5.11)
10:08 < rxr> emte: we have not yet a way (for a target) to request a custom package version
10:08 < rxr> however I thought about that already for some time
10:08 < rxr> Maybe we should just allow override .desc files in the target directory like:
10:08 < rxr> target/ipaq/pkk_linux26.desc
10:08 < rxr> pkg_ even
10:09 < rxr> where you can define whatever you need tweaked from the base .desc
10:09 < rxr> such as the version and download location
10:09 < rxr> for the patches this is ewasy
10:09 < rxr> easy
10:09 < rxr> just drop them in your target directory
10:09 < rxr> if you want to auto apply them name them after the package, such as pkg_glibc.patch
10:09 < rxr> if you do want to keep single patches you need code
10:10 < rxr> in pkg_glibc_post.conf
10:10 < rxr> var_append patchfiles ' ' "`ls target/ipaq/whatever-pattern.patch`"
10:10 < rxr> if you want to wipe all our default patches you also code in this file:
10:10 < rxr> patchfiles=
10:10 < rxr> or if you want to remove just some:
10:11 < rxr> var_remove patchfiles ' ' 'the-patch-that-conflicts-with-something.patch'
10:11 < rxr> you can also alter and mangle the variables with normal bash ways, but the var_append, var_remove, var_insert and var_regex_remove are just convienience functions to work with lists seperated by some character
10:12 < Stelz> re misl, rxr :>
10:13 < rxr> h ist
10:13 < rxr> hi Stelz
10:15 < rxr> Stelz: regarding the checksum
10:15 < rxr> I think later we can change this to sha1
10:16 < rxr> but for now I think we should keep more eyes on new features and bug fixing than incompatible changes
10:16 < rxr> also additionally because sha1 is also already broken in the way md5 can be atacked
10:16 < rxr> so we can just confortable concentrate on new features, stability and backward compatibility
10:17 < rxr> and when there is a strong sha1 successor we change to that on in some years
10:17 < rxr> after all this is just the download checksum
10:17 < rxr> or what do you thinkg ?
10:17 < rxr> -g
10:18 < rxr> valentin: regading reducing redundancy, my redundancy detector (*1) has detected we could happily remove the flist files since we have the filenames in the md5sum file already ...
10:19 < rxr> *1) http://www.openbsd.org/lyrics.html#35
10:20 < Stelz> hmm.. rxr. /me agreed
10:23 < valentin> rxr:
10:23 < valentin> $ wc -l /var/adm/{cksums,flists}/fuse
10:23 < valentin> 14 /var/adm/cksums/fuse
10:23 < valentin> 23 /var/adm/flists/fuse
10:23 < valentin> $ wc -l /var/adm/{cksums,flists}/firefox
10:23 < valentin> 3200 /var/adm/cksums/firefox
10:23 < valentin> 3358 /var/adm/flists/firefox
10:24 < valentin> so some items are missing in the cksums
10:26 < rxr> ogg123 ftp://ftp.openbsd.org/pub/OpenBSD/songs/songty.ogg
10:26 < valentin> the /var/adm stuff only apears in the flist
10:26 < rxr> valentin: that is bugy
10:26 < rxr> that are bugs
10:26 < rxr> e.g. also no symlinks are in the md5sum
10:26 < rxr> this is a bug I fix right now
10:27 < rxr> the only thing the checksum files had was the file size
10:27 < valentin> i'd vote to remove chksum and put the file list + hashs into flists
10:27 < rxr> and the installer used the chksum files because of the file size tag
10:27 < rxr> yep
10:27 < rxr> that is a good idea
10:27 < rxr> I though abaout taking the md5sums
10:28 < rxr> especially because the chksum file was format like this:
10:28 < rxr> checksum file-size file
10:28 < rxr> and the md5sum like:
10:28 < rxr> md5sum file
10:28 < rxr> note the two spaces
10:28 < rxr> now I thought maybe the md5sum command line program can still parse the file if we insert an number in the middle of the two spaces
10:28 < rxr> but it does not - sigh
10:29 < rxr> (written by Ullrick Drepper - what do you expect)
10:29 < rxr> it would have been a cool feature
10:29 < rxr> but when a such tweaked md5sum file is incompatible with standard tools anyway we are better of throwing all data into the flist file anyway ...
10:36 < rxr> maybe I should invest in such a openBSD 10 years of lyrics CD release :-)
10:36 < rxr> http://www.openbsd.org/lyrics.html
10:40 < rxr> valentin: stripping this unneeded package name from the flist file was also on my TODO
10:41 < rxr> so we completely reformat the flist into the format of the good old Unix checksum that is:
10:41 < rxr> sum size filename
10:45 < valentin> kay
10:46 < rxr> hm - despite the build system uses the md5sum for doing the backup in Emerge case of modified files
10:46 < rxr> but we can just plug in our own lua code there
10:46 < rxr> I think about using lua to generate the new-style md5sum file
10:46 < rxr> with some luck it might even be some % faster than
10:47 < rxr> (because md5sum is not exected for each file by xargs but some md5sum function called for each file from within the same program)
10:54 < rxr> real 0m0.435s
10:54 < rxr> user 0m0.256s
10:54 < rxr> sys 0m0.076s
10:54 < rxr> ^- gtk+ checksumming
10:54 < rxr> when the real vs. user overhead is due to process invocation we might be able to gain that bit by calling a md5sum C function thru a lua "for-each-file-loop"
10:58 < Stelz> rxr: why we don't have qt4 ?
10:59 < rxr> because nothing needed it so far
10:59 < Stelz> w00t?
10:59 < rxr> and kde 4 (for qt4) will ealiest ship at the end od this year
10:59 < Stelz> there a lot of nice apps for it
10:59 < rxr> but you are free to add it alongside qt as qt4 package
10:59 < Stelz> nono.. i'm not about kde
10:59 < rxr> Stelz: go go go - add it :-=
10:59 < Stelz> :)
11:03 < Stelz> rxr: kde4 will be released soon?
11:04 < rxr> nope earliest end of the year
11:04 < rxr> but I have feelings that it might be q1 2007 due to the enourmous rewrite
11:04 < rxr> Stelz: what qt4 applications do you have in mind ?
11:05 < Stelz> i'll list them. gimme some secs :-)
11:16 < Stelz> rxr: qbittorrent, qdevelop, speedcrunch, qtwine, linq, qsvn
11:17 < rxr> nice
11:17 < Stelz> and ofcourse psi-beta and hydrogen-svn already ported to q4
11:17 < rxr> Stelz: maybe you really wanna package them ?
11:17 < Stelz> *qt4
11:17 < Stelz> yeah, rxr. i'll try
12:24 < rxr> ok - /me off for lunch and for seeding my home ppc64 with the 6.0.1 build for release tomorrow
12:24 < rxr> cu then
12:30 -!- sepp_ [n=sepp@p83.129.204.18.tisdip.tiscali.de] has joined #t2
12:30 < sepp_> hello :)
12:31 < sepp_> freenode sucks!
12:32 < Stelz> hi sepp_
12:32 < sepp_> :)
12:39 -!- sepp [n=sepp@p83.129.175.64.tisdip.tiscali.de] has quit [Read error: 110 (Connection timed out)]
12:39 -!- sepp_ is now known as sepp
13:34 -!- idealm_ [n=ideal@58.33.58.84] has quit ["Leaving"]
14:52 -!- idealm [n=ideal@58.33.58.84] has joined #t2
17:08 < CIA-9> aldas * r20547 /trunk/package/database/mysql/mysql.desc: * updated mysql (5.0.26 -> 5.0.27)
18:11 < CIA-9> aldas * r20548 /trunk/package/emulators/wine/wine.desc: * updated wine (0.9.23 -> 0.9.24)
18:28 < rxr> moin Baldzius
18:28 < rxr> rehi
18:28 < rxr> sepp: ack
18:36 < rxr> http://freshmeat.net/releases/239722/
18:36 < rxr> ^- valentin wanna package ?
18:49 < CIA-9> rene * r20549 /trunk/misc/luabash/md5/ (. Makefile md5.c md5.h md5.lua md5lib.c test.lua): * added a enhanced md5 Lua module
18:50 < _Ragnar_> btw Config is still happily throwing away all settings occasionally
18:52 < rxr> _Ragnar_: yes, a) when you C-c (Ctrl-C) or b) run it in parallel
18:52 < rxr> _Ragnar_: we are rewriting Config in Lua these days in some weeks all this shortcommings will be gone and it will be bleedingly fast (20-50 times faster at least)
18:52 < _Ragnar_> yes I encountered a)
18:52 < _Ragnar_> I didn't notice any speedup up to now
18:53 < rxr> only one part was speed up ...
18:53 < rxr> intead of 40s it takes 32s now ... :-)
18:53 < rxr> +/- depending on your CPU horsepower
18:53 < rxr> the pkgsel parsing was rediculous slow
18:53 < rxr> now the biggest CPU cycle waster is the continous restart of the whole script
18:53 < rxr> gradually rewriting all this madness in Lua will solve this easily
18:54 < _Ragnar_> ok hopefully
18:54 < rxr> definetly
18:55 < rxr> the only thing I'm not sure about if we finish this before Linux World Expo in mid November
18:55 < rxr> it would be good (tm) as showing T2 usually is not perfectly impressive with this overslow Config ...
18:55 < _Ragnar_> ;D
18:55 < _Ragnar_> indeed
18:55 < _Ragnar_> I tend to avoid config cuz of that and edit the files manually
18:56 < rxr> yeah, ...
19:02 < CIA-9> rene * r20550 /branches/6.0/package/ (7 files in 7 dirs): * merged r18288 from trunk: repriorized mkinitrd include packages for cross build
19:34 < _Ragnar_> btw rxr: While writing key 'cksums': No such file or directory
19:34 < _Ragnar_> ^ happens for every package now
19:43 < rxr> hm
19:46 < rxr> guess from mine, hm ?
19:48 < CIA-9> rene * r20551 /trunk/scripts/Build-Target: * removed chksum reference from Build-Target
20:00 < rxr> we are in the middle of a chksum removal and migration to md5sum only mirgration
20:00 < rxr> migration even
20:04 < Stelz> :)
20:07 < CIA-9> rene * r20552 /trunk/misc/luabash/posix/ (. Makefile lposix.c posix.lua): * added the 5.1 aware flavour of lua-posix
20:07 < rxr> historic relict cleanup process that is
20:07 < rxr> valentin: the two needed lua modules are imported, now
20:07 < rxr> valentin: if you could do a exact-build migration and statically link those all together into our luabash bridge ...
20:08 < rxr> I prepare the new /var/adm/md5sums layout generation via lua in the meantime
20:08 < Stelz> :)
20:08 < rxr> and patch mine for the md5sums -> cksums layout
20:08 < rxr> (and hope we do not need mine that much longer)
20:09 < Stelz> rxr: our pkg-sys will use /var/adm/md5sums ?
20:13 < CIA-9> rene * r20553 /trunk/misc/luabash/posix/modemuncher.c: * added lua/posix/modemuncher.c, indeed needed
20:13 < rxr> sure, /var/adm is our installed package database
20:14 < _Ragnar_> rxr: uhm we should always use something in addition to md5
20:14 < _Ragnar_> md5 has been broken
20:14 < rxr> sha1 also ...
20:14 < rxr> well - the unix cksum was also never used in the past
20:14 < _Ragnar_> and that do you have against mine? o_O
20:14 < rxr> the only thing the file was used for was the file size
20:14 < rxr> well - mine was for the properitary .gem - we use vanilla tarballs now
20:15 < rxr> also mine is codeed without a GUI in mind and the text gui of gasgui is also far from optimal
20:15 < _Ragnar_> yea but loosing lots of the features of mine
20:15 < _Ragnar_> like backuping modified files :P
20:15 < rxr> we do not use anything
20:15 < rxr> for all this the md5sum was used
20:15 < _Ragnar_> I don't WANT a gui installer :P
20:16 < rxr> you do not get forced to a gui instasller :-)
20:16 < _Ragnar_> I like the cmdline mine
20:16 < rxr> you will not notice the migration aside that the text gui will be more useable and look better and Joe users actually have the choice of a GUI version
20:17 < rxr> well - and you will see less segfaults and strange behaviour like read errors before the next CD is requested and the like
20:17 < rxr> and less segfaults
20:17 < rxr> well - that is I patched all known segfaults away in mine ...
20:19 < _Ragnar_> .~.~.~.
20:19 < rxr> _Ragnar_: ?
20:24 < rxr> _Ragnar_: all fine ?
20:24 < rxr> _Ragnar_: or major doubts about the .plan ?
20:30 < CIA-9> rene * r20554 /trunk/package/archiver/mine/x-no-cksums.patch:
20:30 < CIA-9> * patched mine for our now gone cksum file (use the md5sums for the
20:30 < CIA-9> size tag, now)
20:30 < rxr> ^- untested but should work
20:30 < rxr> note that oru md5sums file has no size tag yet
20:30 < rxr> that code fragment will be added tomorrow (I hope)
20:37 < CIA-9> rene * r20555 /branches/6.0/misc/ (GenDepGraph.sh archive/GenDepGraph.sh): * moved misc/GenDepGraph.sh into misc/archive/
20:39 < CIA-9> rene * r20556 /trunk/misc/ (GenDepGraph.sh archive/GenDepGraph.sh): * merged r20555 from branches/6.0: moved misc/GenDepGraph.sh into misc/archive/
20:49 < rxr> oh _Ragnar_ one more new feature of our improvements: symlinks will be md5sum'ed as well
20:50 < rxr> currently you can rm or alter your symlinks without mine -y (check mode) noticing that
20:51 < CIA-9> rene * r20557 /trunk/misc/luabash/posix/test.lua:
20:51 < CIA-9> * added lua/posix/test.lua code fragments we need to inject the filesize
20:51 < CIA-9> and md5sums symlinks
20:53 < rxr> so - me soon leaving the office
20:53 < rxr> valentin: the lua work is done - if you have the free time to setup injecting the md5 and posix lua modules into our luabash we can make this code go live ...
20:56 < rxr> http://osiris.tfh-berlin.de/mirror/t2/incoming/
20:56 < rxr> ^- have fun - should be online in an hour
20:56 < rxr> but untested - testing tomorrow
20:57 < rxr> (testing and the ppc* isos)
21:00 < rxr> so - n8 all - /me off
21:01 < Stelz> 'night rxr
21:05 < emte> night
21:05 < emte> reading your comments now
21:06 < emte> rxr, why not add a new tag for pkgsel that specifies version with a default of edge? this would maintain compat
21:07 < emte> and be in a place you would expect it ...
21:12 < emte> altho the .desc file solution would work as well ... drawback would be (in my case anyway) are the maintained versions for each device would be different. semi-forcing them in sync would mean keeping them relativly maintained
21:13 < emte> which is along the lines of how T2 seems to work now
21:14 < emte> just need the ability to add things in a known working state and then move them towards working with everything else
21:14 < emte> to avoid reinvention syndrom
22:05 -!- kensai [n=kensai@206.248.81.149] has joined #t2
22:07 < kensai> hello
22:07 < Stelz> hi kensai
22:07 < kensai> :)
22:49 -!- misl [n=chatzill@84-104-172-187.cable.quicknet.nl] has quit [Read error: 110 (Connection timed out)]
22:52 < emte> moo
22:53 < Stelz> meow
22:56 < sepp> wuuuf
23:03 < kensai> laatee
23:04 < kensai> glibc has a problem if the buiold host used, has a tar package that uses wildcard
23:04 < kensai> taht is why it fails if I try to build on Debian
23:06 < kensai> is an easy fix though,... change build host :D
23:11 < Stelz> hehe
23:12 < Stelz> made a scr-shot of my current box
23:12 < Stelz> :P
23:12 < Stelz> http://slz.sytes.net/desktop.png
23:21 -!- Netsplit adams.freenode.net <-> irc.freenode.net quits: Stelz, emte, idealm, sepp, mtr
23:22 -!- Netsplit over, joins: idealm, sepp, mtr, Stelz, emte
23:29 < kensai> Stelz, ncie looking desktop
23:29 < kensai> f117
23:30 < Stelz> thanks kensai ;-)
23:42 -!- mtr_ [n=Michael@pD9E104B6.dip0.t-ipconnect.de] has joined #t2
23:56 -!- mtr [n=Michael@pD9E11F7F.dip0.t-ipconnect.de] has quit [Read error: 110 (Connection timed out)]
23:56 -!- mtr_ is now known as mtr
--- Log closed Mon Oct 30 00:00:14 2006