**** BEGIN LOGGING AT Thu May 19 23:59:56 2005 May 20 00:19:44 <__law__> i have tryed to build ah fresh opie-image today :-( (http://pastebin.com/286852) May 20 00:20:29 __law__: OE now requires current bitbake. either the 1.2.1 release, or current svn. May 20 00:22:08 <__law__> kergoth, i have done svn up bitbake before May 20 00:22:47 <__law__> BitBake Build Tool Core version 1.1, bitbake version 1.2.1 May 20 00:24:14 __law__: clear cache May 20 00:24:16 hi all May 20 00:25:20 1 hour hrw 1.3350 konqueror-embedded_20030705: fixed SRC_URI May 20 00:25:21 1 hour hrw 1.3349 uclibc moved to Subversion May 20 00:25:38 my system does not run cia triggers and I dont know why May 20 00:26:58 BK_USER ? May 20 00:27:28 guys May 20 00:27:31 BK_USER is set BK_SILENT is not set May 20 00:27:39 how do I specify a specific version of a package to build? May 20 00:28:00 I want to build xserver-xorg_6.8 not xserver-xorg_CVS May 20 00:28:02 Spyro: PREFERRED_VERSION_package="version" May 20 00:28:18 in conf May 20 00:28:24 cheers May 20 00:28:38 pfft, that one is even in the bitbake manual :P May 20 00:29:05 kergoth: :-~ May 20 00:29:30 I have to find some time and replace PSU fan with zalman... May 20 00:31:55 I will be back in 30 minutes - have to go to work May 20 00:59:52 <__law__> hrw|gone, i have remove the whole tmp dir, but i have got the same problem :-( May 20 01:07:46 what problem May 20 01:09:47 <__law__> robert__, (http://pastebin.com/286852) May 20 01:10:48 re May 20 01:11:55 __law__: update to bitbake trunk maybe? May 20 01:12:02 __law__: which version of python you used? May 20 01:12:41 PYTHONPATH=/usr/local/lib/python2.3/site-packages May 20 01:18:09 <__law__> python -V May 20 01:18:09 <__law__> Python 2.3.4 May 20 01:18:32 <__law__> i have exporte my PYTHONPATH already May 20 01:19:30 <__law__> hrw|work, i should install a newer bitbake version? how? i always say svn up bitbake May 20 01:19:30 try to update your bitbake to 1.2.1 May 20 01:19:51 <__law__> bitbake --version May 20 01:19:51 <__law__> BitBake Build Tool Core version 1.1, bitbake version 1.2.1 May 20 01:20:12 10:11 hrw@marcinj:bb$ bitbake --version May 20 01:20:12 BitBake Build Tool Core version 1.2.1, bitbake version 1.2.1 May 20 01:20:25 BitBake Build Tool Core version 1.2.1, bitbake version 1.2.1 May 20 01:20:28 Updated to revision 184. May 20 01:21:06 update again.maybe some error your bitbake have May 20 01:23:47 __law__: I always delete bitbake directory and refetch it... May 20 01:25:26 <__law__> hrw|work, how can i update to revision 184 ? May 20 01:25:45 <__law__> i will delete the whole bb dir and refetch May 20 01:26:08 __law__: I just do "svn up" as I use trunk May 20 01:26:18 __law__: I keep bitbake and openebmedded in different dirs ... May 20 01:26:24 <__law__> hrw|work, how can i use trunk ? May 20 01:30:18 get bitbake and install it May 20 01:30:45 __law__: fetch, keep in one dir and add that dir into PATH? May 20 01:32:06 <__law__> hrw|work, i have build a rpm for fedora core 3 May 20 01:33:02 what variable can I use to get the staging directory of a native package from a non-native build? May 20 01:47:22 how can I get the HOST ARCH (i686-linux in my case) from bitbake? May 20 01:48:41 is it ${HOST_SYS} ? May 20 01:49:37 b0ti|work: use bbread May 20 01:53:30 finally one user started to work on bt950_cs drived cards for 2.6 May 20 01:59:24 morning May 20 01:59:39 b0ti|work: how do you define 'host'? host is ambigious. try to use the terms 'build' and 'target' to make it clear. May 20 02:00:02 hi mickeyl May 20 02:01:16 fyi, i started on a series of emails to the bitbake-dev list outlining plans and the bitbake-ng design bits. still a bunch more info to send yet though May 20 02:01:37 hmm.. should sleep May 20 02:05:30 kergoth: how long you will be in TX? May 20 02:07:45 03mickeyl 07 * r1.3348.1.3 10openembedded/packages/gpsbabel/gpsbabel_1.2.5.bb: add gpsbabel, courtesy Patrick Ohly. Thanks May 20 02:07:49 03mickeyl 07 * r1.3348.1.2 10openembedded/packages/libsigcx/ (7 files in 2 dirs): add libsigcx 0.6.4, courtesy Steven Brown. Thanks May 20 02:07:52 03mickeyl 07 * r1.3348.1.1 10openembedded/packages/libeasysoap++/ (4 files in 2 dirs): add libeasysoap 0.6.1, courtesy Steven Brown. Thanks May 20 02:11:06 morning mickey|sick May 20 02:11:56 morning RP May 20 02:12:30 mickey|sick: What were the problems you were seeing with orinoco? May 20 02:12:43 RP: it goes like that: May 20 02:13:20 mickey|sick: doesn't work at all? May 20 02:14:10 mickey|sick: so far I've patched 0.13e to 0.14alpha2 and its still working. The 0.13e-SN6 driver seems to have bits of 0.15rc2 in it... May 20 02:14:24 0.13e works great May 20 02:14:36 0.15rc1 introduces lots of dropped packages May 20 02:14:52 right. I'm just wondering what to look for May 20 02:14:52 0.15rc2 not even registers the netdev and when it does it drops 99% of packages May 20 02:15:56 i never did tests with the 0.14 series, so i dunno May 20 02:16:34 the interesting stuff debuted in 0.15 May 20 02:16:40 uf.. good that I have hostap card May 20 02:16:46 monitor mode extensions, iw scan, native monitoring, etc. May 20 02:16:49 good morning all May 20 02:16:51 03koen 07 * r1.3352 10openembedded/packages/gpe-contacts/gpe-contacts_0.38.bb: Add gpe-contacts 0.38 May 20 02:16:53 morning koen May 20 02:16:58 mickey|sick: The 0.14 changes look quite sane. The 0.13e-SN6 we use is rather different from 0.13e though... May 20 02:16:59 get well soon mickey|sick May 20 02:17:04 morning koen May 20 02:17:11 hey RP May 20 02:17:18 thanks koen May 20 02:17:21 hi koen May 20 02:17:22 mickey|sick: Do we care abou 2.4 kernels? May 20 02:17:31 hey hrw|work May 20 02:17:33 RP: not for 0.15. May 20 02:17:34 mickey|sick: yes, GWS! May 20 02:17:46 hrw|work: will be here for another couple days. May 20 02:17:48 RP: 0.13 works fine on 2.4, we can stick with that for a long time. May 20 02:17:53 mickey|sick: Right. That makes my life easier :) May 20 02:18:24 the 0.13e-26 looks kind of different because of the monitor mode patches from shmoo are already in that tarball May 20 02:18:51 RP: thanks. just something with stomach. didn't keep my dinner last nite to where it belongs :/ May 20 02:19:15 mickey|sick: Right - that'll be what's confusing my diffs. It rules out some 0.15 code though which is good May 20 02:20:09 mickey|sick: Ugh :-/. Not nice :-( May 20 02:21:23 Now, with the 2.6.12-rc3 kernel, and a patch for module-init-tools, the May 20 02:21:24 USB hotplug program can be written with a simple one line shell script: May 20 02:21:24 modprobe $MODALIAS May 20 02:21:24 So, with a few more kernel patches for the other subsystems (hint, I'll May 20 02:21:24 gladly take them...) the other helper programs can too go away entirely. May 20 02:21:24 That will leave us with only the main /sbin/hotplug multiplexor program May 20 02:21:26 that any distro that uses udev is starting to abandon entirely. May 20 02:21:28 heh, nifty May 20 02:21:34 * kergoth catching up on email May 20 02:22:39 heh.. that reminds me that I had to test hotplug-perl or hotplug-ng on my destktop May 20 02:22:51 yeah, some exciting things on lkml recently May 20 02:22:54 mickey|sick: Is 18/115 "lots of dropped packets" ? May 20 02:23:25 RP: i've experienced more like 40% and a whole lot of dmesg entries with "BAP error" May 20 02:23:54 0.13e doesn't emit any BAP errors, 0.15rc1 continuesly and 0.15rc2 even more May 20 02:24:05 (when it registers netdev at all) May 20 02:24:28 mickey|sick: Ok. I'm not seeing any BAP errors with 0.14alpha2 May 20 02:24:37 hm, i forgot to mention 'scopes' as a required relationship between recipes, on the list of bitbake-ng requirements i sent to the list May 20 02:24:50 Was just getting a bit worried by the number of dropped packets... May 20 02:24:54 and the obvious consequence, that the api needs to have a means of doing scoped lookups May 20 02:25:20 RP: good. the amount of BAP errors in the log is the major indication if things go wrong or not May 20 02:26:20 kergoth: I read your emails and think we're in basic agreement :) May 20 02:26:28 good good May 20 02:27:08 if you think of any other requirements, be sure to note them. i want to make sure we dont forget about any key functionality May 20 02:27:10 kergoth: The one question in my mind was will bitbake-ng find any users outside of openembedded? May 20 02:27:24 kergoth: are you suggesting that I should use ${BUILD_SYS} instead of ${HOST_SYS}? I'll try that. May 20 02:27:58 RP: the portage guys are looking forward, long long term, at a possible rewrite. i've discussed my intentions with bitbake-ng, and there was some interest amongst developers. i think that if we get it rolling, we can sell it as a backend to a portage rewrite. May 20 02:28:07 thats one of the key other use cases May 20 02:28:24 Right, I wasn't aware of that but it makes sense May 20 02:28:30 and the real reason, other than marc singer's opinions, for the requirement involving imperative recipe formats May 20 02:28:52 consider the possibilities of a flexible library like this. it could easily underly any major build infrastructure tools May 20 02:28:56 apt-build, even May 20 02:29:54 I apprecite its possibilities, I just wondered if we had some actual potential target users. We have so that makes sense :) May 20 02:29:58 right :) May 20 02:30:19 only 2, but i think those 2 are diverse enough that they cover all the major requirements, and scalability concerns May 20 02:31:11 It might be an idea to spell that out somewhere as I wasn't aware of that and its something any ponential bitbake-ng developer should know about May 20 02:31:18 was planning on putting that in the next email :) May 20 02:31:26 right :) May 20 02:31:37 i probably shouldve sent those first, since use cases drive requirements, not the other way around May 20 02:31:38 oh well :) May 20 02:32:38 RP: oh, was thinking.. a good candidate for an incremental improvement that would benefit the rewrite as well is simply to start work on yanking bb.fetch in favor of do_fetch hooking into curl for most things. May 20 02:32:50 fetching really didnt belong in the core anyway. May 20 02:33:08 dunno if anyone wants to take that one on May 20 02:34:50 kergoth: Its a good idea. Personally, I want to try some more tweaks to memory usage and then see what can be done for "multi tasking" but someone might want to pick that up May 20 02:36:04 yeah, i agree. that one isnt as critical or user visible, but itd be a good task to hand off to someone that wants it, since itd help clean out the crap from the codebase May 20 02:36:27 i wonder if zecke has a TODO for the bitbake work yet May 20 02:41:53 If not, he probably should have :) May 20 02:42:58 * kergoth nods May 20 02:43:46 NOTE: [11:43] package autofs-4.1.4: completed May 20 02:43:48 ;) May 20 02:44:12 nice May 20 02:45:54 hm, i should reply to this unionfs thread. the developers solicited use cases for direct writes to members of the union, since it fucks up unionfs cache coherency May 20 02:46:01 ipkg dests via unionfs is such a use case May 20 02:46:04 heh May 20 02:46:13 right May 20 02:46:52 i'm glad someone else pointed out the other obivous one, a setup with an ro flash / underlying the union, but letting the user write back certain files to that member on shutdown if they choose to, otherwise having their changes lost (ramdisk) May 20 02:47:02 thats a common one on lots of devices May 20 02:47:04 2heh May 20 02:47:08 s/^2// May 20 02:47:13 someone have idea why CIA trigger does not run on my machine? May 20 02:47:29 bk co Bitkeeper/triggers/* May 20 02:47:37 hrw|work: not offhand.. its python.. and requires the python xmlrpc.. May 20 02:47:38 <__law__> i have reinstalled bitbake from scratch, same error :-( (http://pastebin.com/286852) May 20 02:48:05 ah May 20 02:48:13 maybe I removed xmlrpc python May 20 02:48:20 __law__: have removed the cache? May 20 02:49:22 no.. it is installed May 20 02:50:11 koen: that wouldnt do anything :P. /* would only match files if they're already in the directory :) May 20 02:50:44 kergoth: that would explain why it doesn't always work May 20 02:50:57 bk co BitKeeper/triggers would do what you intended May 20 02:50:57 I guess I did a bk co triggers May 20 02:51:00 :) May 20 02:53:12 hi BigAl May 20 02:53:39 03mickeyl 07 * r1.3351.1.2 10openembedded/packages/python/ (5 files in 2 dirs): add dedicated python-dbus packages instead of polluting the dbus one May 20 02:53:42 03mickeyl 07 * r1.3351.1.1 10openembedded/packages/ (3 files in 2 dirs): add python-pyrex{-native}, a language specially designed for writing Python extension modules May 20 02:54:21 hi BigAl, how's the CXK 2.6 doing? May 20 02:54:30 s/X/3/ May 20 02:55:09 mickey|sick: boy you can tell i've been out of it. the first i heard about the distros merging was your latest newsletter :) May 20 02:55:12 mickey|sick: Haven't had much time to work on it over the last couple of days unfortunately, but I'm going to try to get the LCD going this weekend. May 20 02:55:34 BigAl: cool May 20 02:55:38 kergoth: any ideas for the new name? May 20 02:55:44 kergoth: heh, really? D'oh May 20 02:55:53 not really. looked over the wiki page but didnt come up with much May 20 02:56:02 ELD looked the best out of the ones there May 20 02:56:16 yeah, i also favour ELD May 20 02:57:11 <__law__> koen, i have remove the whole tmp/ dir May 20 02:57:21 __law__: just tmp/cache May 20 02:57:31 heh. does oe require 1.2.1 atm, or does it require svn? May 20 02:57:44 kergoth: btw., can you adjust the permissions of index.php in oe homedir please May 20 02:57:53 mickey|sick: k May 20 02:57:56 thanks May 20 02:58:00 mickey|sick: btw, you heard hh isnt hosting us anymore? May 20 02:58:05 I like ELD name May 20 02:58:08 d'oh. May 20 02:58:12 kergoth: why's that ? May 20 02:58:13 <__law__> koen, to late ;-) May 20 02:58:21 IT rules changing to the HP ones from the old ones May 20 02:58:28 they can host an hh subdomain, like oe.handhelds.org May 20 02:58:30 but not our domain May 20 02:58:34 hmm May 20 02:58:36 fine with me May 20 02:58:42 told them to do that for now, can forward to it May 20 02:58:47 decide what ot do later May 20 03:01:08 bbl, need to lay down May 20 03:01:52 guys May 20 03:02:09 the xorg 6.8 build is broken as well as the CVS build, in exactly the same way May 20 03:02:28 damn May 20 03:02:37 Spyro: I guess you have to pester pb about that May 20 03:03:02 k :) May 20 03:06:04 looks like the Makefiles arent getting created in the build subdirs May 20 03:11:07 should xc/config/Makefile contain nothing? (seems doubtful :-) May 20 03:16:39 03kergoth * r185 10bitbake-data/doc/test_scoped_lookups.sql: May 20 03:16:39 Set up the scopes for autotools.bbclass and base.bbclass in the May 20 03:16:39 test_scoped_lookups.sql, so that variable values can be looked up for those May 20 03:16:39 recipes in the sample. Also add an example select doing an additional JOIN so May 20 03:16:39 that you can specify the recipe you're looking up in by name rather than by the May 20 03:16:40 recipe_id in the table. May 20 03:18:43 SELECT val FROM vars May 20 03:18:44 JOIN scope ON vars.recipe_id = scope.scope_recipe_id May 20 03:18:44 JOIN recipes ON scope.recipe_id = recipes.id May 20 03:18:44 WHERE vars.var = 'CC' AND recipes.recipe = 'base.bbclass' May 20 03:18:44 ORDER BY scope.priority DESC May 20 03:18:44 limit 1; May 20 03:18:48 hehe May 20 03:18:48 getting awfully ugly May 20 03:18:49 morning May 20 03:18:52 morning May 20 03:19:09 hi XorA May 20 03:20:19 oh, i forgot to add another requirement to the list. for performance reasons, Should be able to manipulate scopes at runtime, so that changing, and thereby reparsing, base.bbclass, doesn't require a reparse of all files using it. May 20 03:26:06 re May 20 03:29:26 hmm. in my tmp/rootfs, there's a few devices in dev.. Shouldn't those come from my device_table, when doing the genext2fs? (i.e., why are those devices in tmp/rootfs/dev?) May 20 03:29:59 ehh, first of all, _why are you root_? May 20 03:30:04 :P May 20 03:30:15 If I re-do genext2fs manually, it clashes May 20 03:30:18 root? May 20 03:30:30 you cant create device nodes as a user, and you shouldnt be building as root May 20 03:30:45 i'm not running as root... May 20 03:31:03 once again, you cannot create device nodes as a user. May 20 03:31:09 are you sure they arent just symlinks or something? May 20 03:31:45 hm. they are just plain files.. how weird. May 20 03:31:52 -rw-r--r-- 1 rob rob 0 May 19 12:13 fb0 May 20 03:31:53 -rw-r--r-- 1 rob rob 0 May 19 12:13 hda May 20 03:31:58 odd May 20 03:32:01 * kergoth shrugs May 20 03:32:17 normally rootfs/dev should not exist? May 20 03:32:49 mickey|sick: opie-cardmon 20050520 will probably need PARALLEL_MAKE="" May 20 03:33:07 rootfs/dev willl exist. it gets created by base-files. May 20 03:33:22 but, be empty? :) May 20 03:33:47 i was just at base-files.. scanning some more.. May 20 03:33:53 thats the idea, it should be empty, yep May 20 03:34:31 ok, perhaps residue then from some previous screwup.. not the first time May 20 03:34:45 hell, i'm up late enough that i'm hungry for breakfast again. i hate that May 20 03:35:10 that can't be good :) May 20 04:04:09 bitbake does not notice changed recipe (r184 here) May 20 04:10:37 hrw|work: yes, I've also noted that problem May 20 04:20:39 hrw|work: bah, do you've time to look at the problem? May 20 04:20:49 (inserting one print statement) May 20 04:20:54 zecke: I can May 20 04:22:07 just say where May 20 04:23:42 * zecke updates his bitbake copy here May 20 04:24:16 hrw|work: lib/bb/make.py May 20 04:24:28 hrw|work: uncomment lines 79 and 82 May 20 04:24:47 ok May 20 04:25:33 hrw|work: and print "Cache times: old: %s new: %s" % (cache_mtime, file_mtime) on line 78 May 20 04:26:13 zecke: dont have that 78 line May 20 04:26:23 a.. add ;) May 20 04:27:16 hrw|work: must be a bug in your FileSystem driver, what do you use reiser? May 20 04:27:44 reiserfs3 May 20 04:27:52 * zecke actually likes Hans Reisers view on the VFS layer May 20 04:28:01 13:27 hrw@home:husky$ bitbake meta-opie|grep meta-opie May 20 04:28:01 Cache times: old: (33188, 2631060L, 773L, 1, 1000, 1000, 86974L, 1116588419, 1116587016, 1116587016) new: 1111145545 for /home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie-dynamic.bb May 20 04:28:05 : '/home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie-dynamic.bb' clean. loading from cache... May 20 04:28:08 Cache times: old: (33188, 16272303L, 773L, 1, 1000, 1000, 94574L, 1116588419, 1116587017, 1116587017) new: 1116588456 for /home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie.bb May 20 04:28:11 : '/home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie.bb' clean. loading from cache... May 20 04:28:20 78 print "Cache times: old: %s new: %s for %s" % (cache_mtime, file_mtime, bbfile) May 20 04:28:21 oops May 20 04:29:15 nice relation (int,long,long,int...)x int May 20 04:29:25 I wonder how < is defined on that relation... May 20 04:29:29 zecke: cache_mtime have a little more data ;) May 20 04:30:51 hrw|work: line 328 in data_smart.py May 20 04:30:55 hrw|work: add a [8] May 20 04:31:45 328 return 0 May 20 04:31:48 this? May 20 04:32:11 http://www.openbsd.org/zaurus.html May 20 04:32:12 wow May 20 04:32:22 326 return os.stat( "%s/%s" % (cache,cache_bbfile[8]) ) May 20 04:32:38 no May 20 04:32:43 behind the statement May 20 04:32:48 return os.stat()[8] May 20 04:32:56 aha May 20 04:33:09 hrw|work: sorry I can't remember more than 2 seconds :} (line numbers...) May 20 04:33:20 zecke: like goldfish? May 20 04:33:36 Cache times: old: 1116587016 new: 1111145545 for /home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie-dynamic.bb May 20 04:33:39 : '/home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie-dynamic.bb' clean. loading from cache... May 20 04:33:42 Cache times: old: 1116587017 new: 1116588785 for /home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie.bb May 20 04:33:45 : '/home/hrw/zaurus/oe/openembedded/packages/meta/meta-opie.bb' dirty. reparsing... May 20 04:33:48 works May 20 04:35:08 will you push that fix? May 20 04:40:27 hrw|work: yupp May 20 04:40:40 ok May 20 04:43:40 03zecke123 * r186 10bitbake/lib/bb/ (data_dict.py data_smart.py): May 20 04:43:40 bitbake/lib/bb/data_smart.py: May 20 04:43:40 Return only the mtime from the os.stat syscall May 20 04:43:40 bitbake/lib/bb/data_dict.py: May 20 04:43:40 Return only the mtime from the os.stat syscall May 20 04:44:52 thx zecke May 20 04:46:40 morning May 20 04:47:24 hi seb May 20 04:47:43 hi, it's been long time I went ther. May 20 04:47:52 yep May 20 04:48:20 A lot to do. I've seen bitbake is moving along. May 20 04:48:56 <_law_> has anyone the new bitbake running with fedora core 3? May 20 04:49:26 proti: 43M memory usage during build May 20 04:49:50 proti: Nice. with bbmasks ? May 20 04:50:01 hrw|work: Nice. with bbmasks ? May 20 04:50:20 with empty bbmask May 20 04:50:28 all 23xx packages in May 20 04:51:11 hrw|work: Really nice. I think it can be improved. May 20 04:51:28 I had this idea for a long time : May 20 04:51:29 works little slower but does not need extra ram May 20 04:51:49 Why does bitbake keeps the cache files ? May 20 04:52:07 proti: because it is secondary storage May 20 04:52:17 In fact to caculate the dependencies of the packages. May 20 04:52:47 zecke: nope, to improve the parsing of metadata. May 20 04:53:04 zecke: it's a shortcut. May 20 04:53:56 hrw|work: I was thinking, how about to make bitbake caculate itself the BBMASK. May 20 04:54:14 proti: the data format is not changed May 20 04:54:57 zecke: correct, but you restart at the stage where the file is parsed. And just after is the dependency calculation. May 20 04:55:30 I build my package twice as native and cross, but the include files in staging do not get installed into arm-linux, only i686-linux. What should I change? May 20 04:56:56 If you parsed all the bb files. Then you calculate the dependencies and then you store those dependencies in the cache files. May 20 04:57:31 b0ti|work: thats correct May 20 04:57:54 hrw|work: what is correct? May 20 04:58:07 b0ti|work: native stage in native, cross in cross May 20 04:58:40 hrw|work: but I have no staging includes for cross, that's not correct. May 20 04:58:45 Next time you use that cache to build the BBMASK and parse only the need bb files (which IMHO should not be more than 100 files max). May 20 04:59:08 b0ti|work: s/cross/arm-linux May 20 04:59:52 Those dependencies will not change a lot along dev once they are correct. May 20 04:59:52 hrw|work: my problem is that I build a package that depends on the cross compiled package's includes, and they are not in staging/arm May 20 05:00:09 ah.. May 20 05:00:28 includes are generated during -native build? May 20 05:00:59 proti: no inner deps are calculated already May 20 05:01:10 -native build is exactly the same as the cross May 20 05:01:10 proti: but yes we could cache all deps as well May 20 05:01:23 proti: or do not calculate all deps at once but only the needed ones May 20 05:01:37 proti: or example look at each bbfile not three times but only once May 20 05:02:19 hrw|work: I even get an -dev package with the includes from the cross build. May 20 05:02:37 b0ti|work: and do_stage() in cross build is wrong? May 20 05:02:44 zecke: bb file are parsed only once if I remember. But the update_data and update_keys are killing performances. There is 3 invokation of thoses functions and they are really expensive. May 20 05:03:45 hrw|work: I didn't touch do_stage() May 20 05:04:10 zecke: The bb files are designed in a way that you need to have all the metadata online to be able to caclulate dependencies. Not shortcut is possible at this stage. May 20 05:04:29 b0ti|work: outside of OE packages? May 20 05:04:34 hrw|work: but indeed the log file for do_stage() is empty May 20 05:04:45 b0ti|work: default do_stage() is empty May 20 05:04:52 proti: if you've any profiling hints they're welcome May 20 05:05:55 zecke: Yes I do. update_data is a prime target. Invoked twice after each parsing of bb file. May 20 05:05:56 hrw|work: so I need to explicitly tell in the cross.bb what do_stage() should do? May 20 05:06:04 yes May 20 05:06:47 zecke: The second culprit is of course the update_keys. May 20 05:07:50 But my main point is : cut the crap. I want to build that cross_gcc packages, I don't need you to read the whole 2300 packages. May 20 05:08:38 But to build the BBMASK automatically, I need the dependencies. May 20 05:08:39 proti: dont you need to read them until you got the one which gives gcc-cross? May 20 05:09:36 hrw|work: Hey, the name of the bb file is not the name of the package. Then you could have several candidate. How do know that without parsing the bb files ? May 20 05:10:17 proti: exactly. I was answering to "cut the crap. I want to build that cross_gcc packages, I don't need you to read the whole 2300" May 20 05:10:17 hrw|work: A good example are the linux kernels. May 20 05:11:05 hrw|work: Yes. Me too. look. You want to build gcc-cross. What file do you need to read to have the recipes for gcc-cross ? May 20 05:11:49 The one who says gcc-cross.bb is wrong. May 20 05:12:12 proti: dont know, so I read all until got correct one - then I think that I can stop reading - but how can I know does it is best one... May 20 05:12:54 hrw|work: The good answer is all the bb files you can get. That is a design spec. May 20 05:13:12 proti: so I read all from BBFILES anyway May 20 05:14:26 hrw|work: No, you read once all the files, write down the packages name they provides (use the caches files for this). Then you don't have to reparse them because you already have this information in the cache files, you use it. May 20 05:14:46 Write down the dependencies they need too. May 20 05:15:17 ok - I will go to lunch and then will try to follow discussion May 20 05:15:33 hrw|work: Bon appetit. May 20 05:18:07 ~curse mozilla cvs fetcher May 20 05:18:10 May the fleas of a thousand camels infest your most sensitive regions, mozilla cvs fetcher ! May 20 05:18:59 re May 20 05:26:52 proti: what you want is to not build the world tree May 20 05:29:35 zecke: What I want is to have the information when I need it, not all the information at the beginning which I don't need at all. May 20 05:32:18 proti: Basically what you're saying is we should only cache the provider and dependency data? Whilst that's a good idea, the design of bitbake makes that difficult. May 20 05:45:11 RP: Correct. Once you know the files you need to read and the order you need to read them, it leaves you with a few files to read which is bearable with the current implementation. May 20 05:47:15 anyone remember how long mozilla/firefox/minimo fetch from cvs? May 20 05:47:48 hrw|work: long May 20 05:47:55 > 20 minutes May 20 05:48:10 NOTE: [14:13] package minimo-0.0cvs20050520-r8: task do_fetch: started May 20 05:48:18 and still fetching on 1Mbps May 20 05:48:37 minimo cvs is < 1Mbit/s anyway :) May 20 05:49:06 Heres a thought for bitbake... May 20 05:49:06 rather than use a stamp to indicate successful compilation May 20 05:49:28 allow packages to nominate a file that will be produced and depend on that instead May 20 06:02:36 argh... screen was frozen... May 20 06:03:09 minimo do_fetch was 10minutes.. May 20 06:03:51 proti: with the cache your almost there, parse data once, next time check if it is present (see for depends you need to parse the file anyway as you could include or inherit deps) May 20 06:04:08 proti: and stop creating the world dep tree May 20 06:12:32 zecke: The big picture of bitbake is : 1- parse bb files, 2- compute the deps, 3- find the impacted files, 4- on those files, follows the recipes. May 20 06:13:02 hey May 20 06:13:11 hey VoodooZ_Work May 20 06:13:30 another "noobs" question... May 20 06:13:32 zecke: I propose : 1- read the already calculated deps. 2- compute impacted files. 3- parse the needed files. 4- follows the recipes. May 20 06:13:53 I'm adding a package for the more recent OV511 webcam driver... May 20 06:14:22 I added a new patch to my .bb file but when I bitbake it doesn't seem to skip it. May 20 06:14:25 zecke: And for the dep calculation (which could be done offline), 1- read all the files, 2- compute deps 3- write down deps. May 20 06:14:34 I did a -c clean and it still doesn't pick it up. May 20 06:14:52 try -b file.bb May 20 06:14:57 I had a similar problem with previous patches but they eventually worked after I wiped out tmp. May 20 06:15:04 the new bitbake doesn't notice changed files May 20 06:15:04 I'd rather not have to do that everytime though. May 20 06:15:13 And voila. No more cache files involved. Just one full parse to update the dependencies from time to time. May 20 06:15:16 i see. May 20 06:15:34 so bitbake -b package? May 20 06:17:07 bitbake -b path/to/file.bb May 20 06:17:29 that only builds the .bb, it doesn't do dependencies May 20 06:17:35 that's why it's so fast May 20 06:18:23 perfect. thanks. works May 20 06:19:57 proti: what you miss 1. look at the deps May 20 06:20:10 proti: they might have changed as conf/local.conf is different, a .bb file was changed May 20 06:20:14 proti: or a bbclass May 20 06:21:04 proti: sure we can write down the deps as well but we do not gain much May 20 06:21:42 proti: You could start profiling if there is a difference in unpickling a dict/list with some entries May 20 06:21:50 proti: and a dict with 1000 entries May 20 06:22:01 hi zecke May 20 06:22:41 zecke: No. I propose that the user should decide when to update the deps. That is his choice to rebuild it. May 20 06:22:44 pb_: Sorry, for my ignorance at #gpe, obviously GUADEC is taking place in Stutgart May 20 06:22:56 zecke: hehe, right May 20 06:23:06 proti: dep build toolchain May 20 06:23:22 proti: but what if he added an ASSUME_PROVIDED, see a dep cache can be outdatet as well May 20 06:23:45 zecke: I guess GUADEC is not of much interest to you. May 20 06:24:03 pb_: I could see the great Miguel :} May 20 06:24:08 heh May 20 06:24:47 zecke: ASSUME_PROVIDED is shortcut. It does not really change the dependencies calculated. It just drops a part of the dep subtree. May 20 06:25:40 proti: it can drop huge parts, anyway by looking at the cache first you do not know if it is current May 20 06:26:32 zecke: Well, up to the user to update the deps. A manual action from time to time instead of a full parse each time bitbake is invoked. May 20 06:26:57 kergoth: long time no c! May 20 06:27:06 proti: you're proposing that the user should update bitbake's dependencies by hand? That sounds rather unsatisfactory. May 20 06:28:40 No, I suggest that the user provoke an update of the deps calculation. First time (no deps calculated) could be automatic, but not the updates. May 20 06:29:07 koen: I was wondering if It's worthwhile for me to push my new package? The 2.6 kernel comes with OV511 built-in but it's relatively old and doesn't support the newer cards/V4L2. May 20 06:29:15 proti: will you respond to users wondering why bitbake does not 'work'? May 20 06:29:17 See. do bk pull. bitbake --dep-updates. Then you are ready to do anything to build. May 20 06:30:36 zecke: FAQ. remove cache files and reinvoke the bitbake to rebuild. May 20 06:30:56 I think that would be a significant retrograde step. May 20 06:31:02 proti: not before I die May 20 06:31:34 oops I shouldn't say stuff like this with my disease... May 20 06:32:05 pb_: I'm currently considering going to Stuttgart with my microwave and bags of popcorn May 20 06:32:17 pb_: I predict the funny language war to take place May 20 06:32:45 Well, just a hint. One could find a smarter way to find that deps should be updated. But, the only way I can think of finding deps have changed is to rebuild the deps themselves and compare to the cache. May 20 06:33:19 Zero win. back to start. May 20 06:33:37 proti: sure one could 'diff' the pickled and freshly parsed data but remember you want to save time May 20 06:34:04 zecke: ah, good idea. You could start a C++ faction. May 20 06:34:23 zecke: Or maybe, keep the old behaviour. Add an option that says : I know there is no modification in dependencies, I know what I do, don't re-compute them. May 20 06:34:29 make Gtk based on Arthur (Qts paint engine) May 20 06:35:05 proti: yeah, you could do that if you want. May 20 06:35:08 proti: help me profiling the actual bottlenecks May 20 06:36:11 zecke: I will have a look at the bottlenecks at home (in 4 hours now). May 20 06:36:44 proti: we can write the world cache to a cache May 20 06:36:52 My point is : To keep bitbake usable, it need to keep the working set of bb files in a reasonable size. May 20 06:36:55 proti: and use it if when parsing nothing was changed May 20 06:37:09 proti: size in terms of what? May 20 06:37:50 zecke: Number of parsed files. Don't parse the qt files when you are build the cross compiler. It's a waste. May 20 06:37:58 /build/building/ May 20 06:38:34 proti: bitbake does not know anything about OpenEmbedded packages May 20 06:39:54 zecke: regarding the "world" thing, we should make bitbake only compute the cache if the user actually specified "--world" on the command line. There is no point generating that data for the 99.9% of builds that do not involve world. May 20 06:40:06 yupp May 20 06:40:24 and, if you _are_ building world, the ten seconds that it takes to calculate the dependencies are negligible compared to the ten weeks that it takes to actually compile all the packages. May 20 06:40:40 Yes. It was just an example. The point is to be able to decide something, bitbake must have parsed all the metadata because of possible side-effects. May 20 06:40:55 indeed. May 20 06:41:59 So give it a hint to ease its work. (deps are a precalculated hint for instance) May 20 06:42:58 If you can propose a way of doing that without causing a significant functional regression, please go ahead. May 20 06:45:00 03pb 07 * r1.3355 10openembedded/packages/howl/ (3 files in 2 dirs): update howl to 1.0.0 May 20 06:45:04 pb_: I will try. It was the point of the whole discussion. The main difficulty is to be able to decide when a dependency has changed without reparsing the whole world. May 20 06:46:17 Personally, I think that it would be better to direct effort towards making the parser run faster, rather than trying to find ways to avoid running it. May 20 06:48:02 2 different ways. I know the parser well now. To accelerate it, you need to make the update_data and update_keys functions faster, or to call them less often (3 times per bb file). May 20 06:48:45 Or modify the rules for variable expansions. May 20 06:51:48 VoodooZ_Work: we do the same for hostap May 20 06:52:18 hostap? May 20 06:52:48 we build them outside the kernel, since the kernel has old ones May 20 06:53:25 * koen adds motion and vloopback to his TODO list May 20 07:02:27 NOTE: [16:01] package openttd-0.4.0: completed May 20 07:02:31 ;) May 20 07:06:36 it builds, not yet package and who know does it even works.. May 20 07:08:30 all open bugs are assigned to oe@handhelds.org now May 20 07:11:51 koen: ok. May 20 07:15:40 Morning May 20 07:19:37 hey CosmicPenguin May 20 07:20:30 hrw|work: does it have a Qt backend or is it SDL based? May 20 07:21:22 zecke: sdl May 20 07:26:34 zecke have you thought about my patches anymore? May 20 07:32:33 Crofton: did you send them to me? May 20 07:32:55 the ones for the Osk board? May 20 07:33:04 I think I was talking to you about them ... May 20 07:33:40 Crofton: yes and I want to apply your patches, either mail them to me May 20 07:33:49 Crofton: or send the URI again :} May 20 07:33:50 http://pastebin.ca/12257 May 20 07:33:53 OK May 20 07:34:22 http://omap-dev.balister.org/~balister/p1.patch May 20 07:34:28 through p5.patch May 20 07:34:45 Crofton: see I'm not home, I will not remember the uri when I'm home :} May 20 07:34:50 then I can send new patches per your comments May 20 07:35:01 what is your email? May 20 07:35:30 Crofton: zecke@handhelds.org May 20 07:36:06 eventually those patches can be submitted via bugtracker.. May 20 07:36:19 ok May 20 07:36:34 hrw|work: is there any reason they can't be submitted via bugzilla right now? May 20 07:36:39 afaik, bugzilla is working fine. May 20 07:37:17 pb_: there is none May 20 07:37:59 Crofton: linux-omap_bk.bb get kernel from file:// - how other devs can get it? May 20 07:38:20 I need to delete it :( May 20 07:38:35 I screwed up and put mixed stuff in one changeset May 20 07:38:51 the only way I know to proceed is apply the patch, then delete May 20 07:38:59 why can't you just fix the patch? May 20 07:39:14 * zecke hands quilt to Crofton May 20 07:39:37 if there is junk in those patches that shouldn't be there, please remove it. May 20 07:39:42 anyway in p4.patch I see linux-omap1_2.6.12-rc2.bb May 20 07:39:52 yep that is the good one May 20 07:40:33 I made the patches via bk, I didn;t want to hand edit and screw them up May 20 07:44:03 I do promise to clean the cruft out as I move forward, and get better at defining changesets May 20 07:44:18 of course then we get to learn a new system in another month or so :( May 20 07:45:20 p1.patch does not apply to site/arm-linux May 20 07:46:48 mmmm... quilt May 20 07:47:14 p1 adds a few lines May 20 07:49:55 fsck.. I forgot that at work bk refuse to do commit ;( May 20 07:52:05 from p1, the important changes are the arm-linux, i686-linux, omap5912osk.conf, and openomap.conf May 20 08:00:13 what's the DISTRO variable that I should put in local.conf (is it openzaurus)? May 20 08:02:50 depends which version you want to build May 20 08:03:04 offroadgeek: use openzaurus-3.5.4 May 20 08:03:18 hrw|work: thanks May 20 08:10:21 should it possibly be 3.5.3? May 20 08:10:24 michael@deb-serv-1:~/stuff/build$ bk clone -ropenzaurus-3.5.4 bk://openembedded.bkbits.net/openembedded May 20 08:10:24 remote: ERROR-rev openzaurus-3.5.4 doesn't exist May 20 08:12:17 Crofton: http://home.hrw.one.pl/oe/diffs/ - files 1.patch to 7.patch - please take a look at them and compare to your patches May 20 08:12:35 offroadgeek: grab latest OE May 20 08:14:21 Crofton: maybe linux-omap1 should be just linux-omap May 20 08:14:48 well, there will be omap2 kernels in the future May 20 08:15:49 I see... May 20 08:16:02 http://focus.ti.com/general/docs/wtbu/wtbuproductcontent.tsp?templateId=6123&navigationId=11990&path=templatedata/cm/product/data/omap_2420 May 20 08:16:05 it's rather linux-2.6.12-rc2-omap1 May 20 08:16:12 hrw|work: if I just did a bk clone this morning, shouldn't that be the latest OE? May 20 08:16:25 offroadgeek: exactly May 20 08:17:12 In the machine file I call for linux-omap1 May 20 08:17:38 this way I get the latest omap1 kernel without calling for a psecific version May 20 08:17:54 hi there. May 20 08:18:06 is there any known issue in the current bitbake? May 20 08:18:18 if i try to bitbake something it fails with the following error May 20 08:18:20 bedboi: update to atleast 1.2.1 May 20 08:18:33 File "/dvdrip/stuff/openembedded/packages/gnu-config/gnu-config-native_cvs.bb:base_do_fetch", line 4, in base_do_fetch May 20 08:18:34 AttributeError: 'module' object has no attribute 'createCopy' May 20 08:18:41 hrw|work: i updated today (with svn) May 20 08:19:13 Crofton: I would also suggest using linux-2.6.11 + patch-2.6.12-rc2 - less downloading when you switch to 2.6.12-rc4 ;) May 20 08:19:20 bedboi: bitbake --version gives? May 20 08:19:35 BitBake Build Tool Core version 1.1, bitbake version 1.2 May 20 08:19:38 old May 20 08:19:41 cool May 20 08:19:43 BitBake Build Tool Core version 1.2.1, bitbake version 1.2.1 May 20 08:19:44 how to update? May 20 08:20:15 svn update or grab 1.2.1 from berlios or checkout trunk via svn May 20 08:20:38 hrw|work, too late I already have the rc4 kernel downloaded :) May 20 08:20:54 1.patch looks good May 20 08:21:08 i svn updated today iirc. May 20 08:21:10 let me see May 20 08:21:26 bedboi: and remove TMPDIR/cache May 20 08:21:42 there are some minot nits I will fix once I get the base stuff in May 20 08:21:45 Crofton: basically I applied p1-p5 and made separate patches for it May 20 08:21:57 I understand May 20 08:22:14 When I fix small stuff, I'll keep the changesets better organized May 20 08:22:40 larger changes are easier to look when split to small ones May 20 08:22:47 yes May 20 08:23:41 2 and 3 look fine May 20 08:24:39 ok that looks good May 20 08:24:44 all of them May 20 08:24:46 Crofton: consider rename linux-omap1_2.6.12-rc2.bb into linux-omap_2.6.12-rc2-omap1.bb - then omap5912osk.conf can have "linux-omap" as kernel and when you switch to newer patches only version will change (but maybe I'm wrong - maybe omap1 and omap2 are not related in the same way as rc2 rc3) May 20 08:25:16 omap1 and omap2 won't run on the same machine May 20 08:25:22 aha ok May 20 08:25:34 NOTE: [17:21] package gpe-image-1.0: completed May 20 08:25:47 Crofton: omap1 and omap2 are like pxa255 and pxa270? May 20 08:25:51 those arm-linux changes are ok too? May 20 08:25:55 different archs? May 20 08:25:55 hmmm, but configs come from seperate directories May 20 08:26:09 I'm not sure about the pxa May 20 08:26:22 Crofton: its like i486 and i586? May 20 08:26:41 BitBake Build Tool Core version 1.1, bitbake version 1.2.1 <- is it good? May 20 08:26:45 anyway - I go home and from there I can push those changes May 20 08:26:50 bedboi: 1.2.1+1.2.1 May 20 08:26:51 yes May 20 08:27:02 cu May 20 08:27:05 There are a few people interested is using OR with OSK May 20 08:27:09 thanks alot May 20 08:27:15 how to update build tool core? May 20 08:27:21 getting the exposure will help us make things better May 20 08:27:27 ooh, evil - Perforce/Emacs integration May 20 08:28:23 The patch from muru.com might change between omap1 and omap2 May 20 08:53:25 re May 20 08:53:32 Crofton: I'm pushing those changes now May 20 08:55:16 pushed May 20 08:56:39 1 hour hrw 1.3354.1.7 Add flags to help build omniorb into site files (patch by Philip Balister) May 20 08:56:42 1 hour hrw 1.3354.1.6 kernel changes for omap5912osk dev board (patch by Philip Balister) May 20 08:56:45 1 hour hrw 1.3354.1.5 switched distro openomap to glibc (patch by Philip Balister) May 20 08:56:48 1 hour hrw 1.3354.1.4 added xerces-c 2.6.0 (patch by Philip Balister) May 20 08:56:51 1 hour hrw 1.3354.1.3 added linux-omap1 2.6.12-rc2 (patch by Philip Balister) May 20 08:56:53 1 hour hrw 1.3354.1.2 added omnievents 2.6.1 (patch by Philip Balister) May 20 08:56:56 1 hour hrw 1.3354.1.1 added omniorb 4.0.5 (patch by Philip Balister) May 20 09:01:57 cu May 20 09:04:34 meh. can makedevs make directories at all, or is that part completely broken.. May 20 09:08:26 heh, now I can work on fixing them :) May 20 09:08:30 Thanks alot May 20 09:28:11 03jcrouse 07 * r1.3350 10openembedded/ (2 files in 2 dirs): May 20 09:28:12 libxfce4util_4.2.2.bb: May 20 09:28:12 Stage .m4s needed by other XFCE apps May 20 09:28:12 delete unneeded 4.0.5 files May 20 09:28:14 03jcrouse 07 * r1.3338.1.1 10openembedded/ (41 files in 14 dirs): Remove xfce 4.0.5 etal, and add 4.2.2 (I hope) May 20 09:28:48 Ok - If I fucked that up - remmeber, its a miracle my dumb ass got that far May 20 09:30:44 proti, pb_: methinks i should add a specific requirement to the list for scalability, both with regard to performance and memory usage. Both use cases will require it, and it's extremely obvious, but it really should be on the list. May 20 09:31:01 (regarding proti's comments on the bitbake-dev list) May 20 09:31:45 pb_: Im told I should talk to you re: xorg 6.8 build failure? May 20 09:34:44 kergoth: ah, I'm not on that list May 20 09:34:54 so, unfortunately, I have no idea what you are talking about. May 20 09:34:54 heh May 20 09:35:41 kergoth: yeah May 20 09:35:48 pb_: https://lists.berlios.de/pipermail/bitbake-dev/2005-May/thread.html May 20 09:35:50 kergoth: I think parallalization should be a key goal too May 20 09:35:53 kergoth: thanks May 20 09:38:29 CosmicPenguin: good point. that's one of those things we always talk about, but which isnt on the list May 20 09:38:45 yeah May 20 09:38:53 Anyone else having issues with flex not downloading May 20 09:40:10 Ken|JLime: change the sourceforgemirror in bitbake.conf May 20 09:40:57 I strongly suspect aleron.sf.net has a faulty flex May 20 09:41:10 Ah, thx. Where can i find a list of mirrors? May 20 09:41:30 download something from sourceforge :) May 20 09:41:45 I have heanet.sf.net (ireland based) May 20 09:42:21 I use easynews in Arizona, USA May 20 09:42:32 can't remember the server name right off the bat May 20 09:42:32 thx man May 20 09:42:34 pretty amazing downloading from ireland is faster than a mirror from my country May 20 09:43:04 koen: heanet is well fast :) May 20 09:43:15 I use heanet too May 20 09:43:38 and so we see that mirror based systems really work. LOL May 20 09:43:40 I used to get 3MB/s on my old connection with heanet May 20 09:44:20 Personally, I blame the mirrors more then I blame Sourceforge May 20 09:44:26 Not to say that Sourceforge isn't blameless May 20 09:45:31 Heh - I sort of like that missing variables cause warnings in the new bitbake - its like use strict; in per May 20 09:45:32 l May 20 09:51:30 proti: i think your email on the bitbake-dev list comes down to just one requirement. the one i just sent regarding scalability. then comes how exactly to make that happen. the next emails i'm sending will be the actual current bitbake-ng design, and i'd welcome your comments and criticisms on the approaches. May 20 09:52:11 * kergoth heads off to get clean and stuff May 20 09:52:28 kergoth must have had a visit from sweet lady motivation May 20 09:52:40 hey zecke May 20 09:52:57 morning May 20 09:53:02 CosmicPenguin: or bad women leave...... May 20 09:53:07 hi treke May 20 09:53:08 home :1 May 20 10:01:10 kergoth: Correct. I was seeing it as do what I asked, and not more. You can however spell it like this : keep the workset as small as possible in memory. May 20 10:07:52 I still think fixchicken takes the cake as the best patch name May 20 10:09:41 proti: the current implementation of the bitbake-data library uses sqlite3 to hold the parsed data, and is capable of attaching scopes at runtime, so that changing base.bbclass will not result in a reparse of the other files. so when a given file changes, the parser reloads it and stuffs it into the db, adjusts any scopes, if necessary, and everyone else is free to poke at it from there. May 20 10:10:24 bye May 20 10:10:29 if we find bottlenecks in sqlite3 performance due to sillyness like requesting the same variable over and over, we can always implement a running cache in ram if need be. wont know whether thats necessary until we profile May 20 10:10:44 I go shopping May 20 10:11:06 kergoth: sqlite3 seems to perform badly on writing to ext3 May 20 10:11:47 zecke: note that sqlite3 is -much- faster if you do all your operations in a single run. so do all your writes of the changes from a parsed file in a single BEGIN TRANSACTION / COMMIT and that may help matters, due to the way sqlite3 manages its cache. May 20 10:12:11 thats the impression i got form reading performance information adn benchmarks on their site May 20 10:13:31 course i'm not exactly sure how to implement that in the api, but we'll figure out something :) May 20 10:13:51 kergoth: I guess even with BEGIN TRANSACTION :} May 20 10:14:06 what i was thinking about is a couple functions to supply _hints_ to the datastore. "hey, i'm going to be operating on this file a _lot_ over the next few minutes, so yeah, you might want to consider holding it in ram or something" May 20 10:14:13 hwh May 20 10:14:15 heh May 20 10:14:46 kergoth: hinting is always good May 20 10:15:28 anyway, lets cross that bridge when we come to it. i figured we might hit something like that, so it shouldnt be difficult to implement if the numbers justify it May 20 10:16:00 It might be that SQL is way over complicated for what we need May 20 10:16:14 heh, i've already benefited from it hugely May 20 10:16:20 scoped lookups are nearly free May 20 10:16:26 as is reattaching them on the fly May 20 10:16:31 kergoth: do you plan to generate the 'trees' on the fly? May 20 10:16:43 kergoth: or do you want to store them as well? May 20 10:17:25 zecke: havent yet decided. lots of the design is still open to discussion, like the exact semantics of the builders and representations of inter-recipe relationships May 20 10:17:32 what do you think? May 20 10:17:59 CosmicPenguin, zecke: May 20 10:18:01 SELECT val FROM vars May 20 10:18:01 JOIN scope ON vars.recipe_id = scope.scope_recipe_id May 20 10:18:01 JOIN recipes ON scope.recipe_id = recipes.id May 20 10:18:01 WHERE vars.var = 'CC' AND recipes.recipe = 'base.bbclass' May 20 10:18:01 ORDER BY scope.priority DESC May 20 10:18:02 limit 1; May 20 10:18:08 heh :) May 20 10:18:53 heh May 20 10:19:03 Now thats some tasty SQL May 20 10:19:30 so all the variables for all the files are kept in the db, independent from one another, and when you do the query, it only hands you the highest scope of each variable May 20 10:19:40 given CC in multiple scopes, the highest priority scope's value is what you'll get May 20 10:20:37 so the sql is doing what COW was doing, the chained lookups, automatically. May 20 10:20:44 without having to have it all in ram May 20 10:21:57 <_law_> i have updated to revision 186 but building opie-image always doenst work :-( (http://pastebin.com/287044) May 20 10:22:04 <_law_> tmp is also clean May 20 10:23:50 OT: is there a way to get mails for each commit to openembedded? May 20 10:24:08 drop a new trigger in BitKeeper/triggers May 20 10:24:10 heh May 20 10:24:33 though I suspect it isnt worth the effort May 20 10:24:42 since there's only what? another month to go? :p May 20 10:24:45 hehe May 20 10:27:25 eh, whats the comment delimiter in a .sql? May 20 10:27:49 dunno May 20 10:33:13 hmmm May 20 10:33:53 * CosmicPenguin taps his foot waiting for glibc.... May 20 10:41:34 kergoth: does base.bbclass in oe doing do_unpack? May 20 10:41:48 yes May 20 10:42:24 ok one TODO is placed wrongly there May 20 10:42:45 03zecke123 * r187 10bitbake/TODO: May 20 10:42:45 bitbake/TODO: May 20 10:42:45 Add basics tasks I want to see finished May 20 10:43:18 basic even... god damn May 20 10:46:48 hehe May 20 10:47:36 treke: I can not query all bugs :} May 20 10:47:43 ? May 20 10:48:29 anyone can do that May 20 10:48:49 hell I dont think you can turn that off in bugzilla May 20 10:49:11 treke: I get an SQL error May 20 10:49:21 * CosmicPenguin just did it May 20 10:49:24 be more specific. May 20 10:49:24 treke: 'Search' -> 'Open' -> 'OpenEmbedded' -> 'Search' May 20 10:49:25 werks for me May 20 10:49:38 treke: when I use 'All' instead of 'Open' it works May 20 10:50:06 treke: I would paste the error message but it is on another X display May 20 10:50:13 zecke: this url? May 20 10:50:15 http://bugs.openembedded.org/buglist.cgi?query_format=specific&order=relevance+desc&bug_status=__open__&product=Openembedded&content= May 20 10:50:25 works for me May 20 10:50:56 me too May 20 10:51:12 find me a url that breaks May 20 10:53:31 treke: ok will do May 20 10:53:37 www.undefined.void May 20 10:54:49 treke: doesn't happen now, I had it today and yesterday... May 20 11:00:37 hmm, i really need to learn sql better May 20 11:00:38 lets see May 20 11:05:52 Stupid frickin' corporate expense report system May 20 11:06:02 yours sucks also? May 20 11:06:05 ~lart concur May 20 11:06:05 * ibot takes a rusty axe and swings it violently, taking concur's head off May 20 11:09:10 lol the mcci usb driver looks funy May 20 11:09:47 hehe yeah I saw that source May 20 11:09:57 I dont see how it's been obfuscated :p May 20 11:10:10 thought it was just written in perl May 20 11:11:12 03zecke123 * r188 10bitbake/lib/bb/data_dict.py: May 20 11:11:13 bitbake/lib/bb/data_dict.py: May 20 11:11:13 -Fix one bug in the unused data_dict.py. There is still one May 20 11:11:13 left... again cookies for the one finding it May 20 11:37:16 treke: could you change 'OpenZaurus' to 'Derived Distros' or something in bugzilla? May 20 11:37:29 sure May 20 11:37:37 well May 20 11:37:38 thanks May 20 11:37:38 maybe May 20 11:38:03 03zecke123 * r189 10bitbake/lib/bb/data.py: May 20 11:38:03 bitbake/lib/bb/data.py: May 20 11:38:03 -We've killed the default arguments in the methods May 20 11:38:03 now the tests were updatet to initialize a data May 20 11:38:03 instance to work again. May 20 11:38:06 mickey|sick's comment was "Yes, change it" May 20 11:38:38 and the answer is no.It can't be changed May 20 11:38:48 oh wait wrong window May 20 11:38:49 it can May 20 11:39:23 Are you sure we want it changed? It was intended to be bugs strictly for openzaurus and not for derived distributions May 20 11:39:56 Is there a difference? May 20 11:40:05 yes May 20 11:41:32 ipkg-1.4.21 not working isn't a bug in openzaurus if we don't ship ipkg1.4.21. It may be a bug in familiar if you ship it May 20 11:42:38 I can still change it if everyone wants it changed May 20 11:45:07 treke: the problem is that the ipkg-1.4.21 bug would be caused by OE, but currently has no place in bugzilla May 20 11:45:14 at least no logical one May 20 11:46:28 unfortunatly bugzilla sucks as well for us :p May 20 11:47:12 a distinction between 'metadata' and 'runtime' bugs might be usefull in bugzilla May 20 11:47:32 the problem here is that versions are product specific and not component specific May 20 11:47:55 So I guess OpenZaurus could become a Version Number May 20 11:47:58 and not a Component May 20 11:48:33 and no one is going to have a clue what runtime is May 20 11:48:51 that's also true May 20 11:49:45 maybe the versions method is the more sane way to go May 20 11:50:21 Do we really have that many packages that are specific enough that it wouldn't be interesting to other distros / platforms that might use those packages? May 20 11:50:46 heh, we had a distros queue in the oe bug tracker May 20 11:50:55 and everyone bitched and moaned left and right about OZ bugs being opened in it May 20 11:51:03 "oh my, oz bugs on the oe list! eep!" May 20 11:51:06 exactly May 20 11:51:15 so I gave oz bugs their own place to go to shut them up May 20 11:51:18 doing that again is just going to hit the same complaints May 20 11:51:18 * kergoth nods May 20 11:51:47 CosmicPenguin: afaik there aren't any distro specific packages in OE May 20 11:51:47 The option is I set OZ 3.5.3 and Familiar 0.8.2 as Versions for the Distros component May 20 11:52:05 CosmicPenguin: maybe distro specific overrides in some packages May 20 11:52:11 koen: but I don't necessarilly care about every bug in OE May 20 11:52:24 I want to know what bugs have been reported against OZ when it comes time to do updates May 20 11:53:01 treke: suppose I file a bug about a package not working in $random_distro May 20 11:53:12 and users wont know if a bug is oz specific or not May 20 11:53:16 that would also apply to the other distros May 20 11:53:23 not necessarily May 20 11:53:23 seperating them is going to cause confusion for them in an attempt to make us happier May 20 11:53:32 heh May 20 11:53:41 so you still need to manually poke bugzilla for a complete overview May 20 11:53:49 kergoth: they know they use OZ May 20 11:53:52 * kergoth wades through sql and tries his best not to go mad May 20 11:53:59 versions make sense to me May 20 11:54:17 versions might make the most sense here May 20 11:54:44 having a bug tracker that can't tell us what bugs are present in a particular release puts us in the position opie is in May 20 11:54:50 anyone really good at sql? May 20 11:55:13 define really good :p May 20 11:55:43 03zecke123 * r190 10bitbake/lib/bb/data.py: May 20 11:55:43 bitbake/lib/bb/data.py: May 20 11:55:43 -Add two test cases to the expand method. The behaviour May 20 11:55:43 was not tested before. May 20 11:55:54 'select * from users' got my my boss' password and SS number :) May 20 11:56:02 So what I'll do is just create a Distros component, and use the different ${DISTRO} values as versions and move everything over May 20 11:56:15 that makes sense May 20 11:56:40 which gets around the problem that bugzilla can't do per component versions May 20 11:57:06 koen: hah May 20 11:57:20 kergoth: the data implementation is correct dunno where the TARGET_FPU problem comes from May 20 11:57:47 weird May 20 11:58:55 kergoth: or my test case is broken, incomplete May 20 12:05:23 kergoth: yeah the fetcher does not call expand but getVar May 20 12:05:46 oh wait May 20 12:05:59 getVar calls expand :P May 20 12:06:56 going home May 20 12:07:52 kergoth: This shows a need for a baseclass for the data modules May 20 12:10:23 kergoth: you should probably be playing with scms and not bb-ng :p May 20 12:10:29 heh, yeah, probably May 20 12:10:37 heh May 20 12:11:13 03zecke123 * r191 10bitbake/lib/bb/data.py: May 20 12:11:14 bitbake/lib/bb/data.py: May 20 12:11:14 -Write an expand test case that actually shows a bug I need to fix May 20 12:11:29 very good May 20 12:11:36 treke: maybe we can use bb-ng as a SCM? May 20 12:11:49 treke: using git as recipe handler May 20 12:11:56 hehe May 20 12:12:00 good idea, we can store all the .bb files in a big sql database May 20 12:12:17 i'm already doing that part :P May 20 12:12:32 fear May 20 12:12:45 SELECT val FROM vars May 20 12:12:45 JOIN scope ON vars.recipe_id = scope.scope_recipe_id May 20 12:12:45 JOIN recipes ON scope.recipe_id = recipes.id May 20 12:12:45 WHERE vars.var = 'CC' AND recipes.recipe = 'base.bbclass' May 20 12:12:45 ORDER BY scope.priority DESC May 20 12:12:46 limit 1; May 20 12:12:48 * kergoth grins May 20 12:13:01 pb_: I prefer to store the .bb files as Gödel Number in a file May 20 12:14:48 zecke: good idea. I suppose we could also reimplement bitbake as a turing machine. May 20 12:15:18 then we could use kergoth's large sql database to simulate an infinite tape. May 20 12:15:46 or we could all just buy palms and try to forget this conversation ever happened May 20 12:15:49 haha May 20 12:16:04 treke: no, that'd suck May 20 12:16:42 hehe May 20 12:17:31 bleh, time to learn about subselects May 20 12:19:33 enjoy May 20 12:21:40 hmm May 20 12:21:58 dont suppose any of you know sql well? May 20 12:22:06 heh May 20 12:22:15 used too May 20 12:22:20 in another universe May 20 12:22:37 heh, me too. May 20 12:22:57 I used to hack sql at futuretv, but that was half a decade ago and I have mostly managed to suppress those memories. May 20 12:23:05 to simplify, we've got a table with variable name, variable value, and priority (scope). i want a single query to get the highest priority value of all variables. May 20 12:23:14 and cant see to find a wayt do it May 20 12:23:34 I sometimes pretend to know sql May 20 12:23:44 seems that now would be a good time. May 20 12:24:03 i cant just arbitrarily use MAX(), because thatd get the max priority of the entire table. if i GROUP BY the variable name so that the MAX ends up correct, then the value is hosed from the GROUP BY. May 20 12:24:37 i can look up a single variable easily, because i can ORDER BY priority LIMIT 1 May 20 12:25:19 lol the test case is wrong :} May 20 12:25:24 hah May 20 12:25:27 ~herring zecke May 20 12:25:28 * ibot whacks zecke on the side of the head with a large red herring named alfred May 20 12:25:46 kergoth: you can use group by and then the 'having' clause May 20 12:27:46 zmiq: already tried. doesnt work. if i say HAVING priority = MAX(priority) with a group by on the variable name, it returns nothing for any variables that have more than one value/priority. you see, when the gruop by happens, it seems to blow away the value and priority fields, giving us the values for those columns from the first instance of the variable. so that by the time we get to HAVING, _none_ match the clause May 20 12:27:54 at least, thats how i'm seeing it. i'm pretty new to this May 20 12:28:07 why does bitbake have so many default parameters? May 20 12:28:12 * kergoth shrugs May 20 12:28:12 source archive is down while I put the new one in place May 20 12:28:23 should just take 5 or so minutes May 20 12:28:34 * zecke starts filing bug reports May 20 12:28:57 zecke: usually for backwards compatibility. May 20 12:29:12 the defaulted arguments are ones that didn't originally exist in the function signature. May 20 12:29:26 ah yeah, good call May 20 12:29:48 pb_: ok that would make sense if it wouldn't be the data you're operating on May 20 12:30:01 * kergoth tries to wrap his brain around this sql madness May 20 12:30:27 perhaps this is a sign that you don't want to do this with sql :p May 20 12:30:49 treke: heh, well, sql has given us enough benefits that its definately what i'm using in general, but you're probably right for htis particular operation May 20 12:31:01 i can just query for all the variables for a given recipe and its scopes, and sort out the highest priority of each in C May 20 12:31:03 what are you trying to do? May 20 12:31:09 i just want to see all the vars for a damn recipe May 20 12:31:11 like bbread May 20 12:31:13 dead simple May 20 12:31:18 zecke: heh. that does sound rather funky. May 20 12:31:24 i want the highest priority (most local scope) of each variable May 20 12:31:27 kergoth: so what is making it complicated? May 20 12:31:48 figuring out an sql query to get me only the highest priority value for each variable in the table. May 20 12:31:54 hehe May 20 12:32:08 this is so fucking simple, its trying me nuts that i cant find a way to do it in sql May 20 12:32:11 :\ May 20 12:32:16 tell your cursor to order by the priority field? May 20 12:32:31 thats how i do a lookup for one variable. i order by priority, then limit to 1 May 20 12:32:40 should be the same thing May 20 12:32:46 but i want a query to show _all_ variables, and then only get the highest priority of each] May 20 12:33:01 kergoth: group by May 20 12:33:11 group by variable name using MAX() blows away the value. May 20 12:33:21 ends up having a value of the first variable it found, not the _value_ of the max May 20 12:33:30 already tried May 20 12:33:38 what database/ May 20 12:33:38 ? May 20 12:33:40 sqlite3 May 20 12:34:32 kergoth: email me your db greg@treke.net. Maybe I can fiddle about with it May 20 12:34:49 treke: http://svn.berlios.de/viewcvs/bitbake/trunk/bitbake-data/doc/test_scoped_lookups.sql?rev=185&view=markup May 20 12:35:04 treke: thats it. the example queries are to get the highest priority value of a single variable May 20 12:35:49 who knows, maybe this functionality (scoped lookup of all variables for a given recipe) wont be needed, but i suspect it will, for debugging if nothing else May 20 12:35:54 heh May 20 12:37:08 03zecke123 * r192 10bitbake/lib/bb/data.py: May 20 12:37:09 bitbake/lib/bb/data.py: May 20 12:37:09 -Don't be too keen finding a bug in expand when there is None... May 20 12:37:09 Fix the testcase to delete {TARGET_MOO} first. May 20 12:37:27 Moo! May 20 12:38:16 kergoth: we want in order of priority? May 20 12:38:29 treke: we want a query to get each variable's value, but only the highest priority value. May 20 12:38:51 so if CC has two values defined, one in bitbake.conf and one in base.bbclass, the highest priority (base.bbclass) one is the value we see May 20 12:39:48 treke: http://kergoth.com/files/test_scoped_lookups.sql with a couple more queries added, and comments showing what each is doing May 20 12:40:02 03zecke123 * r193 10bitbake/lib/bb/ (fetch.py parse/BBHandler.py): May 20 12:40:02 bitbake/lib/bb/fetch.py: May 20 12:40:02 -Warn when init is called without a data instance, we're not able to reorder May 20 12:40:02 them as it would break bitbake <= 1.2.1 and OpenEmbedded May 20 12:40:02 bitbake/lib/bb/parse/BBHandler.py: May 20 12:40:03 -Do not call fetch.init without a data instance. This changes the AtiCore May 20 12:40:05 problem the OE folks are seeing a bit... May 20 12:41:22 ok so the value is in scope? May 20 12:42:30 03zecke 07 * r1.3352 10openembedded/classes/base.bbclass: May 20 12:42:30 classes/base.bbclass: May 20 12:42:30 Call the fetcher methods with a data-instance instead of None May 20 12:43:36 kergoth: would something like this work? : select vars.val, scope.priority, max(scope.priority) May 20 12:43:37 from vars, scope, recipes May 20 12:43:39 where vars.recipe_id=scope.scope_recipe_id and May 20 12:43:40 scope.recipe_id=recipes.is and May 20 12:43:42 vars.var='CC' and recipes.recipe='base.bbclass' May 20 12:43:44 group by vars.val, scope.priority May 20 12:43:45 having scope.priority=max(scope.priority) May 20 12:46:25 sqlite> .read doc/test_scoped_lookups.sql May 20 12:46:26 base.bbclass_cc|2|2 May 20 12:46:26 bitbake.conf_cc|1|1 May 20 12:46:26 bitbake.conf_cxx|1|1 May 20 12:46:38 both values of CC show up with that query May 20 12:46:39 :\ May 20 12:46:48 and man, that query went way over my ability to comprehend May 20 12:46:52 * kergoth reads it again May 20 12:48:13 kergoth: yeah, both last values are supposed to be equal, thus the having clause; apart from that, are you getting the rows you expect? May 20 12:48:37 you missed my point. the first two results of that are both CC May 20 12:48:44 we want only the highest priority value of each variable May 20 12:48:48 that returned both May 20 12:48:59 * kergoth adds vars.var to that SELECT to make it clearer May 20 12:49:21 sqlite> .read doc/test_scoped_lookups.sql May 20 12:49:21 CC|base.bbclass_cc|2|2 May 20 12:49:21 CC|bitbake.conf_cc|1|1 May 20 12:50:04 so you see from that that the MAX() isnt the MAX for a given variable May 20 12:50:08 due to the dual group by May 20 12:50:15 the MAX() for CC is always 2 May 20 12:50:18 03zecke 07 * r1.3353 10openembedded/packages/opie-lockapplet/opie-lockapplet.inc: May 20 12:50:18 opie-lockapplet/opie-security: May 20 12:50:18 lock.png is installed by both opie-lockapplet and opie-security. May 20 12:50:18 As opie-lockapplet depends on opie-security we will leave it May 20 12:50:18 to opie-security to install lock.png May 20 12:50:41 and zecke closes another bug! May 20 12:51:55 then what about this: select vars.var, scope.priority, max(scope.priority) May 20 12:51:57 from vars, scope, recipes May 20 12:51:58 where vars.recipe_id=scope.scope_recipe_id and May 20 12:52:00 scope.recipe_id=recipes.is and May 20 12:52:01 vars.var='CC' and recipes.recipe='base.bbclass' May 20 12:52:03 group by vars.var, scope.priority May 20 12:52:04 having scope.priority=max(scope.priority) May 20 12:52:29 just changes vars.val by vars.var May 20 12:52:46 koen: if # do not grow over 10 we will be able to fix all of them :} May 20 12:52:54 sqlite> .read doc/test_scoped_lookups.sql May 20 12:52:54 CC|bitbake.conf_cc|1|1 May 20 12:52:54 CC|base.bbclass_cc|2|2 May 20 12:52:56 koen: I know we won't May 20 12:52:59 reversed the order, same behavior May 20 12:53:18 s/^/it / May 20 12:54:10 zecke: a lot of bugs will be: can't fetch source May 20 12:56:02 koen: well, I hope to be able to find that out without bug reports May 20 12:56:54 as soon as the lab has been moved I'll work at bb_qa a bit more May 20 12:57:03 ok May 20 12:57:06 new oemirror is live May 20 12:57:13 this one sucks a bit less :p May 20 12:57:26 treke: how do you mirror source? May 20 12:57:34 with oemirror.py May 20 12:57:55 treke: ran with a cron job? May 20 12:57:58 yes May 20 12:58:11 uhm... I really don't understand what you are trying to do; my last try: select vars.var, vars.val, scope.priority, max(scope.priority) May 20 12:58:12 from vars, scope, recipes May 20 12:58:14 where vars.recipe_id=scope.scope_recipe_id and May 20 12:58:15 scope.recipe_id=recipes.is and May 20 12:58:17 vars.var='CC' and recipes.recipe='base.bbclass' May 20 12:58:18 group by vars.var May 20 12:58:20 having scope.priority=max(scope.priority) May 20 12:58:45 i want _only_ the _highest priority_ _value_ of a given variable. May 20 12:59:11 for CC, in a base.bbclass lookup, i only want to get one row back, the CC|base.bbclass_cc|2 May 20 13:00:48 to make things simpler, ignore the joins and everything. consider a table that contains variable name, value, and priority, with CC|gcc|1, CC|gcc2|2, and CXX|g++|1 May 20 13:00:53 kergoth: have you tried my last try? May 20 13:00:55 a query of that shoudl return: May 20 13:00:59 CC|gcc2|2 May 20 13:01:02 CXX|g++|1 May 20 13:01:14 yes, and it returns: May 20 13:01:15 CC|bitbake.conf_cc|1|1 May 20 13:01:18 CC|base.bbclass_cc|2|2 May 20 13:01:26 which is wrong. the max priority for CC isnt 1, its 2. May 20 13:01:37 so we should only be seeing the base.bbclass one due to the HAVING clause May 20 13:01:42 that isnt whats happening because the MAX() is wrong May 20 13:02:35 * kergoth gets more caffeine May 20 13:03:22 can sqlite do subqueries? May 20 13:03:49 zmiq: i think so, yeah. do you have a query using them thatd work? iv'e been trying, and failing, to come up with one May 20 13:03:53 it seems the most likely possibility though May 20 13:08:05 uhm... what about this: select vars.var, vars.val, scope.priority, max(scope.priority) May 20 13:08:06 from vars, scope, recipes May 20 13:08:08 where vars.recipe_id=scope.scope_recipe_id and May 20 13:08:09 scope.recipe_id=recipes.is and May 20 13:08:11 vars.var='CC' and recipes.recipe='base.bbclass' May 20 13:08:12 and scope.priority=( May 20 13:08:14 select max(scope.priority) from May 20 13:08:15 from vars, scope, recipes May 20 13:08:17 where vars.recipe_id=scope.scope_recipe_id and May 20 13:08:18 scope.recipe_id=recipes.is and May 20 13:08:20 vars.var='CC' and recipes.recipe='base.bbclass' May 20 13:08:21 group by vars.var) May 20 13:09:10 okay this is the big dillemma. even if that works, what i _really_ want to do is do a query for all the variables, and get only the highest priority of _each_, which means that subselect would have to be on a per variable basis May 20 13:09:22 i think i'm just gonna say fuck it and do it in C :) May 20 13:12:10 ok. i'm pretty sure we were close, but you know, friday night here already .... good luck. May 20 13:12:16 hehe May 20 13:12:20 thanks for the help man May 20 13:15:15 kergoth: just iterate over all the variables and et the highest of that one :p May 20 13:15:50 hehe May 20 13:15:50 heh - "fuck it and do it in C" May 20 13:15:51 yep May 20 13:15:55 I should have a sign made and put it over my desk May 20 13:16:04 heh May 20 13:16:08 heh May 20 13:16:08 good idea May 20 13:16:16 it seems that no matter your schema there is always some odd thing that's fucking annoying to code May 20 13:25:58 hrw|husky: is houseoffish.org down or behind a broken router? May 20 13:30:02 zmiq: tested your last query, and it does indeed behave the way it should, returns just CC|base.bbclass_cc|2|2 May 20 13:32:01 wb _alwin_ May 20 13:41:31 zmiq: May 20 13:41:31 SELECT vars.var,vars.val,scope.priority FROM vars May 20 13:41:32 JOIN scope ON vars.recipe_id = scope.scope_recipe_id May 20 13:41:32 JOIN recipes ON scope.recipe_id = recipes.id May 20 13:41:32 JOIN (SELECT vars.var var,MAX(scope.priority) max May 20 13:41:33 FROM vars, scope, recipes May 20 13:41:35 WHERE vars.recipe_id=scope.scope_recipe_id AND May 20 13:41:37 scope.recipe_id=recipes.id AND May 20 13:41:39 recipes.recipe='b.bb' May 20 13:41:41 GROUP BY vars.var) m ON vars.var=m.var May 20 13:41:43 WHERE recipes.recipe='b.bb' AND May 20 13:41:45 scope.priority=m.max; May 20 13:41:47 there's probably a more optimal way, but that works May 20 13:42:00 joining the results of a subselect, named, on the variable names matching :) May 20 13:42:17 jesus is that ugly May 20 13:42:33 kergoth: store these as functions? May 20 13:43:43 kergoth: glad you made it; as ugly as a sith? May 20 13:43:53 hehe May 20 13:45:55 so now it's my turn: using oe-oz3.5.3 and the python packages, what packages are required to get pickle working? I get the error 'self.append(rep.decode("string-escape")) LookupError: no codec search functions registered: can't find encoding' when doing pickle.load May 20 13:46:19 zmiq: rm tmp/cache? May 20 13:46:40 it's a python question, not oe May 20 13:46:53 zmiq: import pickle May 20 13:47:12 zmiq: or for a better pickler use cPickle May 20 13:47:17 import pickle is already there ;-) May 20 13:47:37 add more vinegar for perfect pickle May 20 13:48:01 the problem is that even i have ipkg installed python-pickle, I get this error ... May 20 13:48:07 mmmm pickles May 20 13:48:25 zmiq: ok maybe some module is missing? May 20 13:48:42 which one? May 20 13:48:59 i wouldn't like to install them all ... May 20 13:49:39 zmiq: dunno, some picker format implementation :} May 20 13:49:46 zmiq: you would need to look into the code May 20 13:49:50 the script actually works on my gentoo; it's on the zaurus that I get this error, without info on what is missing ... May 20 13:50:31 and pickle.dump works fine; it pickle.load that brings the error ... May 20 13:50:51 zmiq: 'no codec search functions registered' May 20 13:51:06 zmiq: sounds pretty obvious it can't ask/create the factory May 20 13:52:34 zecke: i'm kind of newbie to python; what do you mean? Actually I think i need to install some other python-* package; just wondering which one. Anything I can try? May 20 13:53:07 zmiq: I don't know how pickle works internally but I know the concept of a factory as it is pretty language neutral May 20 13:53:18 zmiq: did you look into pickle.py yet? May 20 13:53:33 zmiq: it might be a packaging bug May 20 13:53:34 no May 20 13:54:07 should I? oh.. I hate to spend time on this type of debugging, while it works in the desktop ... May 20 13:55:36 zmiq: so now I should spend time to look into pickle.py for you? May 20 13:55:53 zmiq: what is rep? May 20 13:56:23 kergoth: you have a larger version of img.jpg? May 20 13:56:28 uhmm... I'm not asking for that; I was just asking for some help some of you guys could just give right away ... May 20 13:56:28 er logo.png May 20 13:56:48 03kergoth * r194 10bitbake-data/doc/test_scoped_lookups.sql: Update the scoped db example queries to show a query that displays the highest priority values of all variables in a given recipe and its scopes (not like a bbread on the recipe). May 20 13:56:50 treke: we have the svg :P May 20 13:56:53 make whatever size you want May 20 13:56:55 whehe cool May 20 14:06:35 evening May 20 14:08:56 anyone got a non-kdrive xserver binary handy ? May 20 14:09:48 03kergoth * r195 10bitbake-data/doc/test_scoped_lookups.sql: Make the sql queries in the .sql a bit more readable. May 20 14:22:54 hi xora May 20 14:24:53 hi woglinde May 20 14:26:13 he tmbinc May 20 14:26:43 tmbinc I saw that you tested gcc4.0 for the dreambox May 20 14:28:09 http://users.bigpond.net.au/Peter-Williams/ May 20 14:28:09 heh May 20 14:28:34 zecke: http://www.sqlite.org/cvstrac/wiki?p=PerformanceConsiderations May 20 14:28:44 I did a timed test with inserting 1000 records into a table in various ways to compare performance: May 20 14:28:44 * main database table without transaction - 5 seconds May 20 14:28:44 * main database table with transaction - 0.1 seconds May 20 14:28:44 * temporary table without transaction - 2 seconds May 20 14:28:44 * temporary table with transaction - 0.1 seconds May 20 14:28:46 heh May 20 14:28:52 May 20 14:29:32 kergoth: that is sqlite2 and no fs specified May 20 14:30:33 zecke: yeah, i know, but you get the idea. http://www.sqlite.org/speed.html has info as well May 20 14:33:25 kergoth: Don't take this complaining too seriously May 20 14:33:32 * kergoth nods May 20 14:33:41 just pointing out what i read before May 20 14:34:23 hmm, the sqlite sources includes a .tcl for speed testing May 20 14:34:26 going to try it May 20 14:35:22 woglinde: hu? no i'm sticking with gcc 3.4.3. I'm not good at testing experimental gccs :) i'm happy if i have a stable environment, and right now i wouldn't consider gcc4 to be that much stable... May 20 14:36:33 hm I meant ihr May 20 14:36:39 hm or only ghostrider May 20 14:47:29 woglinde: gcc4 is still misscompiling a lot of code... i would not want to use it right now May 20 14:48:16 roh i saw commits for engima May 20 14:48:17 roh: no pain no gain May 20 14:48:38 fixing code for gcc-4 *g* May 20 14:49:26 woglinde: You can compile code with g++-4.0 it is pickier/better May 20 14:51:05 03zecke123 * r196 10bitbake/lib/bb/fetch.py: May 20 14:51:06 bitbake/lib/bb/fetch.py: May 20 14:51:06 -be less anoying and only print debug messages when May 20 14:51:06 d is none May 20 14:51:29 hrw|husky: where do I get knights from? May 20 14:52:41 woglinde: i would wait till the bugs in gcc4 are fixed... and only fix code if its really the code and not gcc thats buggy. May 20 15:18:35 hey Hertog May 20 15:19:07 Hey Koen :) May 20 15:19:44 koen: and, what's your estimate, how many Zauri in Enschede? May 20 15:19:57 3, maybe 4 May 20 15:20:22 I've seen someone with a 5600 on the uni May 20 15:21:22 ah, prolly bought em in Germany, they where officially released there a while ago I think May 20 15:21:53 a few years back I almost bought one in the german mediamarkt May 20 15:22:21 :) I know the feeling ;) May 20 15:22:57 I got an ipaq 2210 instead May 20 15:23:07 no keyboard, but it does have bluetooth May 20 15:23:22 hehe, I go even further back, I think.. Philips Nino May 20 15:23:33 Massive brick May 20 15:23:41 I *almost* ordered a c1000 last week May 20 15:23:52 till I realized it still hasn't bt and wifi May 20 15:24:01 (and I have too much gadgets allready) May 20 15:24:15 hehe May 20 15:24:31 koen: you never have too much until the wife moans May 20 15:24:47 nope, the perfect handheld still doesn't excist.. clamshell, keyboard, HD a plenty, buildin wifi&bluetooth May 20 15:24:53 XorA|gone: girlfriend complained a while ago, so I gave her the h2210 May 20 15:25:25 showed her virtualdub, and now it's her mobile friends+ER watching device May 20 15:25:26 Girlfriend can complain all she wants, I just bought three Samsung 173P 17'' LCD's :) May 20 15:25:30 mine didnt want a 5500 May 20 15:25:52 Makes for one hell of a desktop,, not really mobile tho May 20 15:26:47 Hertog: I once had a sparc5 + 3x 20" working May 20 15:26:56 I had to give it up when I moved May 20 15:27:16 eh, sucks May 20 15:27:34 I thought I read you had to give your girlfriend up when you moved ;) May 20 15:28:46 heh, no May 20 15:32:22 hehe May 20 15:45:55 'night all May 20 15:46:27 hmmpf, smash me to pieces if this is the wrong place, but, err, is mtd-utils-native supposed not to be available on the downloadsites? May 20 15:47:56 it wants to fetch mtd-snapshot-20050414.tar.bz2, but there is no 200504XX at the various sites May 20 15:48:58 the template system in bugzilla is pretty ugly :p May 20 15:51:29 and theres a line on oe.org I litterally see no reason for May 20 15:51:40 * koen laughs at http://www.oesf.org/forums/index.php?showtopic=12718&pid=80723&st=0&#entry80723 May 20 15:52:38 koen oh my god May 20 15:53:59 init=emerge gcc-2.95.3 May 20 15:55:55 ~lart tinderbox2 May 20 15:55:55 * ibot beats tinderbox2 into protomatter with the andromeda galaxy May 20 15:56:57 wtf May 20 15:57:10 the same html renders differently on two different servers May 20 15:57:43 yeah May 20 15:57:52 apache server much better html as IIS May 20 15:58:07 s/server/serves/ May 20 15:58:13 koen: LOL :) May 20 16:00:40 http://bugs.openembedded.org/oz.html May 20 16:00:44 http://openembedded.org/ May 20 16:00:54 do those look the same to you? May 20 16:01:53 extra line on the right May 20 16:01:58 argh May 20 16:01:59 I dont get it May 20 16:02:00 the latter May 20 16:02:03 it's the same HTML May 20 16:02:24 i've seen reports of hh.org messing up css by setting hte wrong mime-type May 20 16:02:57 the css is in the file May 20 16:05:54 there is no logical reason they should render exactly the same way May 20 16:13:11 tinderbox outsmarts me May 20 16:14:15 hehe May 20 16:14:19 what's going wrong May 20 16:14:36 zecke: take a look at http://buildbot.sourceforge.net/ May 20 16:16:37 treke: is that python based? May 20 16:16:47 it must be better than tinderbox then :} May 20 16:17:30 zecke: yes May 20 16:17:47 zecke: Was talking to someone who used it before for a fairly large app. He was mostly happy May 20 16:17:53 http://bugs.openembedded.org/ is less ugly now May 20 16:17:56 BuildBot like BitBake... May 20 16:18:08 at least the main page May 20 16:20:05 'night all May 20 16:23:19 treke: can I create a new category in the bug tracker? May 20 16:23:22 treke: missing source? May 20 16:23:33 you can't May 20 16:23:44 those belong in the build section May 20 16:28:40 treke: I would like to send automatically generated bug reports with tinderbox... May 20 16:28:51 treke: but not spam the bug tracker, hence different sections May 20 16:31:57 they belong in the proper section May 20 16:32:07 if the source isn't accessible it's a build problem May 20 16:32:17 so spam the fucking tracker if there are that many of them May 20 16:33:26 I guess that means I should file a bug for mtd-snapshot then May 20 16:54:01 zecke: you fiddled with buildbot at all year? May 20 16:54:10 yet May 20 16:54:21 just saw Collateral. pretty good movie May 20 16:54:28 yeah May 20 16:54:30 I liked it May 20 16:55:11 can never go wrong with the classic 'ordinary guy gets thrown into extraordinary circumstances" plot lines May 20 16:55:12 heh May 20 16:59:08 treke: no, I guess I'll get tinderbox running first May 20 16:59:39 zecke: if you need hardware, you can install it on treke.net May 20 17:01:11 hehe thank you for your offer May 20 17:01:24 until my setup gets secure I'll keep it local May 20 17:01:27 it's a fairly speedy machine May 20 17:01:28 cool May 20 17:02:11 treke: I think I've understood 40% on how tinderbox May 20 17:02:26 but I didn't set up an apache for ~3 years May 20 17:04:36 03kergoth * r197 10bitbake-data/TODO: Add note about usage hint functions as a TODO item. May 20 17:06:45 treke: I'm quite excited about QA, we've some good offers up... May 20 17:07:14 I think Im onto something re: Xorg 6.8 May 20 17:07:20 (Im running the build manually May 20 17:07:22 ) May 20 17:07:39 what are BUILD_CC under bitbake ? May 20 17:07:44 the native tools ? May 20 17:08:02 yep. BUILD is your build machine May 20 17:08:40 curiously, when I run it manually and simply omit the BUILD_ bits so that it defaults to my native compiler, it seems much happier. May 20 17:11:25 what is CC under bitbake? the cross tools ? May 20 17:24:06 good nite May 20 17:24:14 Spyro: TARGET_CC? May 20 17:24:34 hm. May 20 17:24:43 Im not sure whats going on in this damn Xorg build May 20 17:25:14 looks like a single failure renders the entire tree unbuildable anyway... May 20 17:25:25 since ti delete the makefiles and fould up the mackup... May 20 17:27:19 s/ti/it s/fould/fouls/ s/delete/deleted/ s/mackup/backup/ May 20 17:27:23 :-/ May 20 17:27:27 it must be late May 20 17:28:50 zecke, proti: sent out the next of my bitbake-ng emails, and cc'd you two and phil directly so you see it sooner :) please comment May 20 17:29:32 ah good, hit the list right away, no hour delay this time May 20 17:31:30 <[g2]> hey kergoth ! May 20 17:31:34 hey May 20 17:31:51 <[g2]> all moved ? May 20 17:32:13 naw. movers came and got all my stuff a couple days ago though May 20 17:32:18 just a matter of cleaning my apartment and taking off May 20 17:32:22 <[g2]> cool May 20 17:32:34 <[g2]> code or clean... tough choice :) May 20 17:32:41 hehe May 20 17:32:49 <[g2]> I'll take comet and steel wool for 15 May 20 17:33:11 <[g2]> so what the latest with bitbake ? May 20 17:33:22 i've got a lot to do. thankfully most of it is just a matter of throwing out shit i dont need or want or cant fit in the car May 20 17:33:33 I'll reply tomorrow May 20 17:33:35 good nite May 20 18:26:42 [g2]: zecke's work has bitbake memory consumption down to like 20-30 megs or something with all the files loaded. more improvemnts to the current codebase are in the todo May 20 18:26:53 [g2]: i'll be picking up bitbake-ng and running with it when i get to MN May 20 18:49:21 <_ggilbert> until then he isn't allowed to work on it May 20 18:49:26 <_ggilbert> he needs to find us a new scm :p May 20 18:49:36 <_ggilbert> or about 50 grand May 20 18:49:45 heh, yeah, need to look at cogito/git again, and current monotone May 20 19:51:04 * france is back (gone 63:39:04) May 20 19:57:52 <_ggilbert> this guy is confusing me May 20 19:58:00 <_ggilbert> now he might actually have insurance May 20 20:12:22 <[g2]> kergoth, THX for the update. Safe travels and GL in MN May 20 20:12:27 np May 20 20:12:28 thanks May 20 20:36:27 <[g2]> is bkbits.net down or is it just me ? May 20 20:36:42 here too (singapore) May 20 20:36:48 yep, its toast May 20 20:37:01 oh well.. May 20 20:37:09 <[g2]> is there a snapshot of the openembedded repo around ? May 20 20:37:22 check oesources.org May 20 20:37:35 <[g2]> there's one from 5/4 there May 20 20:37:53 <[g2]> that's pre-reorg of the metadata right ? May 20 20:38:01 probably May 20 20:38:12 i have a tree from yesterday i can export if you'd like May 20 20:38:14 <[g2]> I wanted to test the latest May 20 20:38:26 <[g2]> that would be great May 20 20:38:33 <[g2]> can put put it on oesources ? May 20 20:38:51 no, i dont maintain that, nor do i generate its content May 20 20:39:01 i dont want just random people throwing shit up there anyway May 20 20:39:07 that content shoudl be at least somewhat controlled May 20 20:39:30 <[g2]> hey I'm trying to get the sw controlled :) May 20 20:39:50 <[g2]> ok where would be a good place to get it ? May 20 20:40:13 ? May 20 20:40:19 i just told you, i have one i'll export if you want May 20 20:40:28 "good place" is whatever machien i feel like putting it on May 20 20:40:52 oesources.org is the official correct place to grab such things, but it sounds like the export process is behind May 20 20:41:07 <[g2]> nod. nod. nod. May 20 20:42:45 <[g2]> kergoth, THX for the offer, I'm about to head off to bed. May 20 20:42:53 <[g2]> I'll try the repo again tomorrow May 20 20:42:58 k May 20 20:43:12 <[g2]> I'm sorry if I'm acting impatient I'm just a little tired May 20 20:43:18 np May 20 20:43:19 night May 20 20:43:33 <[g2]> cool I look forward to trying the latest tomorrow May 20 20:44:16 <[g2]> I probably told you I rolled a custom Knoppix that could build Openslug/Unslung and actually probably any OE build May 20 20:45:45 <[g2]> Well off to sleep May 20 20:45:48 <[g2]> nite all **** BEGIN LOGGING AT Fri May 20 21:22:35 2005 May 20 21:52:32 @familyguy May 20 21:52:33 kergoth: [writing letter] Dear MacGuyver, Enclosed is a rubber band, a paper clip, and a drinking straw. Please save my dog. **** ENDING LOGGING AT Fri May 20 23:59:56 2005