**** BEGIN LOGGING AT Wed Nov 08 02:59:57 2006 Nov 08 03:11:55 03bzhou * r4403 10optware/trunk/make/py-selector.mk: py-selector: upstream upgrade to 0.8.10 Nov 08 07:24:34 03bzhou * r4404 10optware/trunk/make/python.mk: python: added PYTHON_MAJOR variable Nov 08 07:25:22 03bzhou * r4405 10optware/trunk/make/python24.mk: python24: added host-stage target Nov 08 07:25:36 03bzhou * r4406 10optware/trunk/make/python25.mk: python25: added host-stage target Nov 08 07:26:35 03bzhou * r4407 10optware/trunk/make/py-setuptools.mk: py-setuptools: added py25-setuptools Nov 08 07:27:03 03bzhou * r4408 10optware/trunk/make/py-cherrypy.mk: py-cherrypy: added py25 Nov 08 07:27:25 03bzhou * r4409 10optware/trunk/make/py-cheetah.mk: py-cheetah: added py25 Nov 08 07:30:32 03bzhou * r4410 10optware/trunk/make/py-zope-interface.mk: py-zope-interface: added py25 Nov 08 07:32:04 03bzhou * r4411 10optware/trunk/make/py-crypto.mk: py-crypto: added py25 Nov 08 08:11:38 eno-zzz Can you help me with package generation with staging/bin/ipkg-make-index written in python. Nov 08 08:15:44 the problem is that packages/uclibc_0.9.28-1_mipsel.ipk is selected in favor of packages/uclibc_0.9.28-4_mipsel.ipk when generating staging/Packages file Nov 08 08:16:11 The command for this is staging/bin/ipkg-make-index -v -v packages > Packages Nov 08 08:29:23 rwhitby can you remove old uclibc_0.9.28-1_mipsel.ipk from oleg and dd-wrt dirs? Nov 08 08:32:10 morning rwhitby Nov 08 08:32:37 It looks like control file of uclibc_0.9.28-1_mipsel.ipk is strangely ordered Nov 08 08:38:56 NAiL can do this mainenance job of removing old packages? Nov 08 08:39:10 s/can do/can you do/ Nov 08 08:39:11 oleo meant: NAiL can you do this mainenance job of removing old packages? Nov 08 08:44:20 oleo: not until after work, I don't have access from work. Nov 08 09:16:48 is it just me or is monotone.nslu2-linux.org down ? Nov 08 09:17:30 it's just you Nov 08 09:17:34 This is monotone.nslu2-linux.org, serving nslu2-linux related branches Nov 08 09:17:34 The best way to use these repositories are to use the Master Makefile Nov 08 09:17:34 View the repository (monotone.nslu2-linux.org/viewmtn) Nov 08 09:17:34 View recent activities (cia.navi.cx) Nov 08 09:17:34 Note that you need monotone 0.26 (or later) to connect to this server Nov 08 09:18:02 CieD: let me rephrase ... doing an anonymous mtn pull gives me connection refused from monotone.nslu2-linux.org Nov 08 09:18:13 mtn 0.29 Nov 08 09:18:18 "Unable to list branches: mtn: error: sqlite error: database is locked" Nov 08 09:18:27 the monotone database is temperamental at best Nov 08 09:18:42 the only advice I have ever been given is to wait till it fixes itself Nov 08 09:18:48 can take hours, unfortunately Nov 08 09:18:57 great ... Nov 08 09:19:03 it is shocking Nov 08 09:19:37 i am kinda needing it right now ... as i newed to put together a cross-compile environment Nov 08 09:19:46 Ahh, as I pasted I thought I was missing something Nov 08 09:22:42 NAiL, whenever. It looks like the problem ocurrend when ipkg-utils were upgraded. Nov 08 09:23:28 monotone is fixed Nov 08 09:23:58 thanks rwhitby :) Nov 08 09:24:38 * Caelian|work waits for the mtn server to settle down .... Nov 08 09:24:54 it looks like it doesn't like connections this soon after fixing :) Nov 08 09:25:49 Wheeeeeeeeeeeee Nov 08 09:26:10 just takes a while to check the db which was locked Nov 08 09:26:26 i figured ....connection is going now :) Nov 08 09:32:40 rwhitby, mazbe zou will be so handy and remove uclibc_0.9.28-1_mipsel.ipk from oleg and dd-wrt dirs? Nov 08 09:32:44 rwhitby, mazbe zou will be so handy and remove uclibc_0.9.28-1_mipsel.ipk from oleg and dd-wrt dirs? Nov 08 09:32:56 rwhitby, maybe zou will be so handy and remove uclibc_0.9.28-1_mipsel.ipk from oleg and dd-wrt dirs? Nov 08 09:33:03 rwhitby, maybe zou will be so handy and remove uclibc_0.9.28-1_mipsel.ipk from oleg and dd-wrt dirs Nov 08 09:33:14 oops sorry Nov 08 09:33:15 why remove them? Nov 08 09:33:34 oleo: very funny :) Nov 08 09:33:40 Packages refer them as latest. Nov 08 09:34:12 oleo: on holidays in qwertz contry? :) Nov 08 09:34:45 I Use caps lock to change from my national to english keyboard. Nov 08 09:35:13 aha :) Nov 08 09:35:18 But then i need to change a brain too Nov 08 09:35:24 oleo: what's the problem with the packages? this is the first I've heard of this, so please explain. Nov 08 09:36:34 it is explained 1 hour ago. It looks like ipkg-make index wrongly catches first uclibc package Nov 08 09:37:20 I have revisited both control files and it seems that this is cause of ipkg-utils upgrade Nov 08 09:41:19 ipkg-make-index is run on the build host, not the feed directory. Nov 08 09:42:32 Yes. I know. But if you take both uclibc packages and run "ipkg-make-index -v tmp" you will end up with -1 and not with -4 Nov 08 09:43:00 any idea why? Nov 08 09:43:01 I could not figure out what is wrong with this python. Nov 08 09:43:23 also our autobuild script should remove both first, before building the correct one Nov 08 09:43:31 I have changed -1 to -5 and then -4 was selected Nov 08 09:44:10 build dir for ddwrt only has -4 Nov 08 09:44:25 No. You said that you are preserving old packages due to scash problems before. Nov 08 09:44:39 preserving in the feeds only, not on the build machine Nov 08 09:45:05 but for some reason there are others in the packages dir on the build machine. That shouldn't happen Nov 08 09:45:07 Realy? Nov 08 09:45:44 maybe autoclean needs revision Nov 08 09:46:08 the rsync from the build dir to the packages dir uses --delete Nov 08 09:46:19 but we don't use --delete when uploading to the feed Nov 08 09:47:19 bbiab Nov 08 09:47:31 maybe uclibc is taken out by http://trac.nslu2-linux.org/optware/changeset/3780 Nov 08 10:04:13 ls -1 packages |sed 's/\(.*\)-.*/\1/'|uniq -d packages shows duplicates Nov 08 10:05:15 optware-autoclean.pl failed to clean microperl_5.8.8 postfix-doc_2.3.0 and uclibc_0.9.28 on my machine Nov 08 10:16:09 Is this because there is no microperl.mk postfix-doc.mk and uclibc.mk Nov 08 11:03:54 YES. Nov 08 11:09:10 03blaster8 * r566 10kernel/trunk/patches/2.6.19/ (77-velocity-module.patch KERNEL series): Bump to 2.6.19-rc5, remove patch now in mainline Nov 08 11:10:01 03blaster8 * r567 10kernel/trunk/patches/2.6.19/defconfig: Update defconfig Nov 08 11:13:23 rwhitby, if I touch nonexistent make/uclibc.mk then autoclean finds uclibc as outdated Nov 08 11:47:20 I see no solution to the problem with changing optware-autoclean.pl as there is no connection between .mk and package name. Unless we use grep. Nov 08 11:48:16 or add dummy uclibc.mk microperl.mk postfix-doc.mk Nov 08 11:54:15 Will try to fix autoclean with grep if .mk nonexistent. Nov 08 11:56:45 The problem is rather huge. I've found the following problematic packages: apache-manual, busybox-base, busybox-links, cups-doc-de, cups-doc-es, cups-doc, cyrus-imapd-devel, cyrus-imapd-doc, cyrus-sasl-libs, dspam-mysql, dspam-pgsql, findutils-doc, freeradius-doc, imap-libs, microperl, openldap-libs, php-dev, php-embed, php-gd, php-imap, php-ldap, php-mbstring, php-mysql, php-pear, postfix-doc, sdl-dev, uclibc, wget-ssl Nov 08 13:42:09 03oleo * r4412 10optware/trunk/scripts/optware-autoclean.pl: optware-autoclean: grep for packages without explicit .mk file Nov 08 13:45:04 03oleo * r4413 10optware/trunk/make/findutils.mk: findutils: Package name should be explicit for autoclean to function Nov 08 15:00:45 03bzhou * r4414 10optware/trunk/make/py-setuptools.mk: py-setuptools: mv python-host-stage up Nov 08 16:53:06 oleo: the packages you mentioned are all sort of like slave packages, their version matches the master package Nov 08 16:53:48 so we don't need to check separately the version of apache-manual Nov 08 16:54:08 if apache is out-of-date, then rebuild Nov 08 16:54:26 what is the problem? Nov 08 17:04:27 hi Nov 08 17:04:47 rwhitby are you there ? Nov 08 17:04:58 i got an error trying to make unslung Nov 08 17:05:16 NOTE: Unpacking /home/ubuntu/slug/downloads/fakeroot_1.2.13.tar.gz to /home/ubuntu/slug/unslung/tmp/work/fakeroot-native-1.2.13-r1/ Nov 08 17:05:20 gzip: stdin: not in gzip format Nov 08 17:05:22 tar: Child returned status 1 Nov 08 17:05:25 tar: Error exit delayed from previous errors Nov 08 17:05:27 NOTE: Task failed: Nov 08 17:51:58 03bzhou * r4415 10optware/trunk/make/ (python24.mk python25.mk): python2?: disable bluetooth for now Nov 08 18:01:43 ah k got it the download was corrupt :-/ Nov 08 23:25:03 03fcarolo * r4416 10optware/trunk/ (5 files in 2 dirs): enhanced-ctorrent: upstream upgrade to dnh2.2 Nov 08 23:59:05 03fcarolo * r4417 10optware/trunk/make/enhanced-ctorrent.mk: enhanced-ctorrent: small typo Nov 09 00:23:12 eno, The problem with slave packages was that thy did not have .mk file and autoclean was not cleaning old packages. Nov 09 00:23:31 03bzhou * r4418 10optware/trunk/scripts/optware-autoclean.pl: revert to PREV version, false positive with the grep version Nov 09 00:24:08 oleo: but can you have a separate uclibc.mk ? Nov 09 00:25:13 no. because uclibc is just tiny part of buildroot package. Nov 09 00:26:03 oleo: i thought autoclean calls master-dirclean Nov 09 00:26:07 I have found many similar packages. Nov 09 00:26:31 autoclean is makefile independent. Nov 09 00:26:32 and master-dirclean should clean slave packages as well Nov 09 00:27:36 No. The problem occurs when you end up with two versions of the same package in packages/ Nov 09 00:28:16 autoclean will never clean microperl for example. Nov 09 00:29:18 and packages/Packages will get first version and not last. Nov 09 00:30:20 I do not understand what do you mean with false positives. Nov 09 00:31:14 autoclean is not very clever when extracting package versions. Nov 09 00:31:29 on nudi, we keep getting messages like: cups-doc-de is out of date. Feed=1.2.4-1 .mk=- Nov 09 00:32:28 as far as master.mk is concerned, i think autoclean is quite clever Nov 09 00:32:34 yes. I've saw this. Nov 09 00:33:50 i'll take a closer look at the packages/ clean issue Nov 09 00:33:52 the problem with Feed=1.2.4-1 .mk=- is that many such packages do not have strict naming Nov 09 00:34:39 cups-doc-de is ok, but cups_doc_de is not by the debian book. Nov 09 00:38:38 And line 160 in scripts/optware-autoclean.pl is obvious bug. Reverted :( Nov 09 00:40:55 i'm not good at perl, what does \L do? Nov 09 00:41:09 to lowercase Nov 09 00:42:12 ok, i'll add that - although not much a problem in real life Nov 09 00:43:46 reverting is no solution. If i raise IPK_VERISION in buildroot package, I will not get it in PAckages. Nov 09 00:44:08 03bzhou * r4419 10optware/trunk/scripts/optware-autoclean.pl: to lower case when matching as oleo suggested Nov 09 00:44:25 hi Nov 09 00:44:38 we need solution and not a fix. Nov 09 00:45:00 does anyone know how to compile apex used in the new debianslug? Nov 09 00:45:27 well, why it's a problem for uclibc and not a general problem in the first place? Nov 09 00:46:01 microperl always has the same IPK_VERSION as perl Nov 09 00:46:51 kinsa: debianslug or Debian/NSLU2 ? Nov 09 00:46:51 because autoclean searches .mk files. If .mk file does not exist it just coutinue with next package. Nov 09 00:46:56 when we bumped apache IPK_VERSION, i'm sure we got apache-docs too Nov 09 00:48:03 Debian/NSLU2 Nov 09 00:48:11 http://trac.nslu2-linux.org/optware/browser/trunk/scripts/optware-autoclean.pl#L128 is the line that skips the package Nov 09 00:48:19 i guess Nov 09 00:49:29 kinsa: there should be a .src.deb for it then Nov 09 00:49:55 sounds to me that if uclibc is a slave package, it should not be in optware/Makefile at all Nov 09 00:50:01 yes, Marc singer uploads Apex to debian unstable Nov 09 00:50:09 and it has a debian/nslu2 config file Nov 09 00:50:42 rwhitby: yup, I had a look there Nov 09 00:50:44 many packages are slave packages. they are listed in Makefile Nov 09 00:51:10 I tried compiling and I got the same result Nov 09 00:51:34 the resulting binary can't seem to switch from BE to LE mode Nov 09 00:51:50 kinsa: do you have the shim prepended to the kernel? Nov 09 00:51:58 or do I need to do some byte swapping Nov 09 00:52:22 no, I haven't touched the kernel yet Nov 09 00:52:48 eno, when you bumped apache IPK_VERSION youv got two apache-docs. And by chance last was listed in Packages. Nov 09 00:52:57 i was trying to bypass the secondary bootloader so I can enter my own kernel flags Nov 09 00:53:44 but I can't create Apex that switches to LE successfully Nov 09 00:55:50 oleo: i'll look at it tonite Nov 09 00:56:35 kinsa: see trac.nslu2-linux.org/kernel Makefile for the devio stuff we apply to the kernel Nov 09 00:57:44 ok, thanks. will have a look Nov 09 00:59:22 eno. Please. It bugged me fo a while. I've needed 3 hours to point at solution. Although incomplete. But I think ths versions can be extracted cleverly wit addons to mine work. **** ENDING LOGGING AT Thu Nov 09 02:59:57 2006