**** BEGIN LOGGING AT Thu Aug 14 02:59:57 2008 Aug 14 08:10:37 bonjour Aug 14 08:55:47 good morning Aug 14 09:05:21 florian: good morning Aug 14 09:32:33 Hello Aug 14 09:33:53 The md5 and sha hashes of busybox-1.11.1-tar.patch are wrong in OE tree. Do I need to file a bug report or can someone update them please? Aug 14 09:34:54 kalev: that would be worth a bug report Aug 14 09:35:00 okey Aug 14 10:06:52 mickeyl: good morning Aug 14 10:07:02 morning pb___ Aug 14 10:10:50 mickeyl: what's the status of the gsm bits in openmoko nowadays? that is, what are my chances of being able to use my gta01 as some kind of functioning phone? Aug 14 10:11:13 I guess the supplementary question is, will the new firmware even run on gta01 anymore or is everything targetted solely at gta02? Aug 14 10:18:47 sudo su Aug 14 10:18:51 grr Aug 14 10:24:36 why is the site always down or slow as hell? Aug 14 10:24:55 make it work for gods sake Aug 14 10:27:50 He needs a faster internet connection... Aug 14 10:29:21 pb___: FSO seems to be pretty reliable and Im pretty sure they test it on GTA01 Aug 14 10:29:39 pb___: thats mickeys community effort Aug 14 10:34:50 XorA: ok, thanks. gsm? Aug 14 10:36:51 pb___: the GSM on FSO is so rock solid it even made a phone call on my phone with known broken gsm Aug 14 10:37:00 pb___: it handles errors :-) Aug 14 10:38:17 awesome Aug 14 10:55:22 libsdl bug fix with recent kernel headers: http://bugs.openembedded.net/show_bug.cgi?id=4401 Aug 14 10:55:34 can someone review please and commit to mtn? Aug 14 11:18:45 mickeyl: did you see my email about preferred-om-2008-versions.inc ? Aug 14 11:28:12 ya Aug 14 11:28:17 let me reply Aug 14 11:30:05 done Aug 14 11:30:42 pb___: FSO milestone2 should work ok for you. If you can wait a bit longer, we'll release milestone3 by the end of this month and it will be even better :) Aug 14 11:31:24 mickeyl: ah, awesome Aug 14 11:31:26 sure, I can wait a bit longer :-) Aug 14 11:32:07 good. I always test on 01 as well. Aug 14 11:33:44 XorA: Aren't you interested in sdl stuff? bug 4401 could be interesting for you, I think Aug 14 11:33:59 !oebug 4401 Aug 14 11:34:00 * * Bug 4401, Status: UNCONFIRMED, Created: 2008-07-02 00:30 Aug 14 11:34:00 * * gwossum(AT)acm.org: libsdl includes asm/page.h, uses deprecated getpagesize() Aug 14 11:34:01 Laibsch: Im hung over, give me a day to get brain in right order Aug 14 11:34:02 * * http://bugs.openembedded.net/show_bug.cgi?id=4401 Aug 14 11:34:13 3am in the pub is too late at my age Aug 14 11:34:18 XorA: Alright Aug 14 11:34:28 Now, don't you go again tonight, though ;-) Aug 14 11:34:33 I did look at bug, but it heart my head :-) Aug 14 11:34:39 hurt Aug 14 11:34:44 see I cant even use the right words Aug 14 11:34:56 yes, point proven and taken ;-) Aug 14 11:35:39 XorA, if you are going to bed at 3AM, take a nap b4 going out :) Aug 14 11:35:55 Crofton: how does the alleviate the 8 pints? Aug 14 11:36:27 you get to the pub later :) Aug 14 11:36:34 less time, so you drink less Aug 14 11:36:38 Crofton: it was 10pm before I got to pub Aug 14 11:36:44 hmmm Aug 14 11:36:51 Crofton: it was one of those, just one more pint nights :-) Aug 14 11:37:02 I understand Aug 14 11:37:14 pubs are open late at the moment for the festival Aug 14 11:37:18 "A round for all my friends" Aug 14 11:43:17 mickeyl: do you have a moment to lend me your python ski11z? I have a small problem with regular expression parsing. Aug 14 11:43:47 not exactly my strength, but let me see Aug 14 11:44:01 suppose I have: Aug 14 11:44:02 r=re.compile("^\s*(-L\s*(?P\S+))|(-l\s*(?P\S+))|(-I\s*(?P\S+))\s*(?P.*)") Aug 14 11:44:07 m=r.match('-L/home/js/proj/oe-reciva/svn/openembedded/tmp/staging/arm-linux/lib -lglib-2.0') Aug 14 11:44:19 now, m.group("libdir") contains the -L stuff, as I would expect Aug 14 11:44:40 but m.group("rest") is None, whereas I was hoping to find the "-lglib-2.0" here. Aug 14 11:45:39 let me start kodos Aug 14 11:45:51 mickeyl: is someone working on gllin and fso? Aug 14 11:46:02 rwhitby: yeah. alphaone knows the details Aug 14 11:48:01 morning all Aug 14 12:09:11 pb___: hmm... tough one Aug 14 12:09:19 something with the greedieness Aug 14 12:09:24 i had some luck using Aug 14 12:09:24 ^\s*(?:(-L(?P\S+))?)\s*(?:( -l(?P\S+))?)(?:( -I(?P\S+)))\s*(?P.*) Aug 14 12:10:01 but that will probably be depending on order Aug 14 12:10:48 mickeyl: yeah, it's tricky. the funny thing is that if I put a $ at the end, it still matches successfully but I can't figure out what part of the regexp is accepting the "-lglib-2.0" string. Aug 14 12:11:19 using the one I pasted I can get at least -lglib-2.0 into libdir Aug 14 12:11:22 (where it belongs) Aug 14 12:11:43 i found I need the space before the -l Aug 14 12:11:43 mickeyl: no, I want that string in libname, not libdir Aug 14 12:11:51 ya, libname i meant Aug 14 12:11:55 ah right, yeah Aug 14 12:11:57 * mickeyl tired Aug 14 12:11:59 ;) Aug 14 12:12:04 :-) Aug 14 12:12:49 without the space before the -L (likewise -l and -I), the parsing matches too early, namely for arm-linux (-l!) Aug 14 12:13:00 hence gets confused Aug 14 12:13:03 ah Aug 14 12:13:04 right Aug 14 12:13:37 so you should better change your arch ;) Aug 14 12:13:40 heh Aug 14 12:13:45 yeah, I should use the hurd Aug 14 12:13:48 haha Aug 14 12:13:55 nah, that will take too long to match... Aug 14 12:14:51 I'll try your thing of requiring spaces before the -s and see how I get on Aug 14 12:15:02 I guess I can just stick an extra space on the front of the string to help it match the very first item in the line Aug 14 12:15:12 * XorA wants to start a lib call inux now just to really confuse people Aug 14 12:15:14 yep Aug 14 12:15:23 XorA: heh Aug 14 12:15:47 * XorA imagines BSD people frothing at -linux commands Aug 14 12:15:49 in the same tradition as libowfat and stuff Aug 14 12:20:28 op odesan Aug 14 12:20:38 op odesan Aug 14 12:21:07 oops zapomniale / sorry Aug 14 12:25:44 mickeyl: would you be interesting in what happens when running an FSO built from angstrom-2008.1 distro, or would you really prefer that I just always use openmoko distro? Aug 14 12:26:12 and should I expect the results to be much the same, or widely different? Aug 14 12:28:30 aargh Aug 14 12:28:32 ~lart speedbone Aug 14 12:28:32 * ibot drops a humongous exploding nuke on speedbone Aug 14 12:28:47 mickey: doh Aug 14 12:29:07 down again? Aug 14 12:29:10 we're down again and i did not finish my rsync to the VM yet Aug 14 12:29:22 drat Aug 14 12:30:13 sounds like a case of please patch us into a different router Aug 14 12:30:24 mickey: Do you have backups of fs.o? Aug 14 12:30:27 no Aug 14 12:30:36 oh, okay Aug 14 12:30:38 let me check Aug 14 12:30:45 i rsync once and then Aug 14 12:30:58 anything with fso? Aug 14 12:31:07 (was talking bout oe.net atm.) Aug 14 12:31:08 no Aug 14 12:31:12 *phew* good Aug 14 12:31:15 Yeah I know Aug 14 12:31:19 Was just thinking Aug 14 12:31:26 mickey: That's why we decided to rent a machine at a big computing centre. Aug 14 12:32:33 I could add an rdiff backup rule to my setup here Aug 14 12:32:43 fs.o doesn't seem that huge Aug 14 12:33:20 mickey: on another topic, could I persuade you to adjust the mailing list setup for oe-devel to stop these duplicate postings? Aug 14 12:33:21 mickeyl. looks like a Thursday problem to me Aug 14 12:34:11 indeed Aug 14 12:34:12 *sigh* Aug 14 12:34:23 pb___: The postings will stop once we switch to git as our main repo which will happen really soon now (tm) :-) Aug 14 12:34:26 it's not the server though Aug 14 12:34:35 some broken router or switch Aug 14 12:34:36 alphaone: er, I think we are talking about different postinst Aug 14 12:34:40 postings, even Aug 14 12:34:51 ah, okay Aug 14 12:35:12 right, om lists seem to have that as well Aug 14 12:35:26 I don't mean the commit messages, these are the ones that get duplicated by being sent to oe-devel@lists.oe.net, cc: oe-devel@oe.net Aug 14 12:35:48 is there a nice way to filter out offending compiler flags in .bb files? Aug 14 12:36:04 the problem seems to be that both addresses work, but the mail server always adds a reply-to: pointing to the latter. Aug 14 12:36:18 so, if someone posts to @oe.net, then someone else hits "reply to all", the list gets two copies Aug 14 12:36:58 kalev: what do you mean by "offending"? Aug 14 12:37:20 and up again Aug 14 12:37:29 heh Aug 14 12:37:31 well done :-) Aug 14 12:37:35 i'll stop web services and launch an rsync Aug 14 12:37:45 NOW Aug 14 12:37:52 mickey: :) Aug 14 12:38:00 pb___: some packages don't compile with -Os, which is my default Aug 14 12:38:20 kalev: you can set SELECTED_OPTIMIZATION to something else for those packages Aug 14 12:38:41 or file bugs against gcc as appropriate :-} Aug 14 12:39:58 ok Aug 14 12:40:23 * pb___ lunchtime now Aug 14 12:40:47 in Gentoo build system it is possible to do something like "filter-flags -fomit-frame-pointer -fforce-addr" Aug 14 12:41:34 this way we could filter "bad" flags out in generic packages Aug 14 12:50:46 rsync underway... Aug 14 12:56:19 mickey: You do have some kind of connection to the machine? how is that possible when a router in-between is dead? How can mtn still be functional in this case? Aug 14 12:56:25 * Laibsch is really confused Aug 14 12:56:34 online again Aug 14 12:56:39 great Aug 14 12:56:39 it was just down for 5 minutes Aug 14 12:57:03 but i'm doing some rsync now Aug 14 12:57:09 and shut down web for that Aug 14 13:13:59 Oh, I see Aug 14 13:28:43 helllo to all Aug 14 13:29:35 I have a problem maybe you can help me Aug 14 13:29:38 when I try to crosscompile I recibe this answer Aug 14 13:30:32 File "/home/moko/bitbake/bin/bitbake", line 189, in Aug 14 13:30:32 main() Aug 14 13:30:55 File "/home/moko/bitbake/bin/bitbake", line 189, in main() Aug 14 13:31:33 File "/home/moko/bitbake/bin/bitbake", line 167, in main cooker = bb.cooker.BBCooker(configuration) Aug 14 13:31:59 File "/home/moko/bitbake/lib/bb/cooker.py", line 65, in __init_ self.parseConfigurationFile( os.path.join( "conf", "bitbake.conf" ) ) Aug 14 13:32:06 ~pastebin Aug 14 13:32:07 [~pastebin] A "pastebin" is a web-based service where you can paste anything over 3 lines without flooding the channel. Here are links to a few : http://www.pastebin.com , http://pastebin.ca , http://channels.debian.net/paste , http://paste.lisp.org , http://www.rafb.net/paste Aug 14 13:32:26 File "/home/moko/bitbake/lib/bb/cooker.py", line 399, in parseConfigurationFile Aug 14 13:32:26 bb.fetch.fetcher_init(self.configuration.data) Aug 14 13:32:26 File "/home/moko/bitbake/lib/bb/fetch/__init__.py", line 97, in fetcher_init Aug 14 13:32:26 pd.delDomain("BB_URI_HEADREVS") Aug 14 13:32:26 File "/home/moko/bitbake/lib/bb/persist_data.py", line 70, in delDomain Aug 14 13:32:27 self.connection.execute("DROP TABLE IF EXISTS %s;" % domain) Aug 14 13:32:29 sqlite3.OperationalError: attempt to write a readonly database Aug 14 13:33:17 and I dont know what is the problem, this happend whe I try to do "bitbake -c build scr" Aug 14 13:33:32 and also with "bitbake -c clean scr" Aug 14 13:33:47 someone could help me?รง Aug 14 13:34:16 sorry Aug 14 13:38:24 thanks I did "http://pastebin.com/m1e921053" Aug 14 13:43:44 antz: "find /home/moko/ -name *.sqlite3 -exec ls -l {} \;" Aug 14 13:43:59 antz: it looks as if you have some permissions for TMPDIR are wrong. Maybe you used a different user to build earlier? Aug 14 13:55:29 kalev: yes, you can do that in oe if you want. Aug 14 13:55:35 see flow-lossage.bbclass for example Aug 14 13:56:43 Is it just me, or are the oe web sites down today? Aug 14 13:57:08 * jmcmullan just noticed the topic. Aug 14 13:58:00 thanks Aug 14 13:58:05 i will check Aug 14 14:01:00 hail zecke Aug 14 14:09:54 thanks a lot, I try to fix the permissions, but I always compile with the same user I think Aug 14 14:10:04 mickey: oddly, the solution to my re problem seemed to be to wrap the (x|y|z) bit in another level of parenthesis. Aug 14 14:10:09 not quite sure what was going on there Aug 14 14:10:47 oh, wait, I see. I didn't have as many brackets to start with as I thought I did Aug 14 14:10:49 doh Aug 14 14:11:50 heh Aug 14 14:11:57 RE is strange sometimes. Aug 14 14:12:35 yeah Aug 14 14:13:06 my error was that I had "(something)|(something else)|(third thing)(remainder)$" Aug 14 14:13:21 whereas, clearly what I wanted was "((something)|(something else)|(third thing))(remainder)$" Aug 14 14:13:40 * pb___ pokes self in eye with a sharp stick Aug 14 14:14:26 hehe Aug 14 14:14:30 too-many-brackets error in line 1 Aug 14 14:14:31 ;) Aug 14 14:14:47 right :-} Aug 14 14:16:53 pb___: okay, I just looked at your patch Aug 14 14:17:35 pb___: why do we need to write the hook out? Aug 14 14:18:02 zecke: I'm not sure I understand the question Aug 14 14:19:00 if you don't write it out then... well, it won't be there to be executed Aug 14 14:19:34 pb___: I talk about your package.bbclass change and the addition of package_stash_hook. What is the reason to write the hook out? because the hook of one bb file gets executed in another context? Aug 14 14:19:43 right, exactly Aug 14 14:21:15 zecke: did you follow the background of the libgcc issue? Aug 14 14:21:38 pb___: not really, it is a problem that sometimes libgcc does not end up in a rootfs Aug 14 14:22:25 pb___: I think I'm fine with the hook... I would have some style comment and wonder if we could ever change the 'T' var but not set it back properly... but the approach is sane Aug 14 14:26:17 zecke: the underlying issue is that, under some circumstances, calling functions in glibc can cause it to dlopen libgcc (and crash if libgcc isn't installed) Aug 14 14:26:52 so, depending on a somewhat-complicated set of conditions, packages which link against glibc and call certain functions may need a dependency on libgcc. Aug 14 14:27:42 zecke: send me your style comments, I will consider them on their merits :-} Aug 14 14:28:33 pb___: I prefer early exits. instead if os.access(foo, R):\indent indent indent, I tend to do if not os.access(foo, R):\ return :) Aug 14 14:30:03 right, fair enough Aug 14 14:30:10 I guess these are my old C habits. Aug 14 14:31:35 style, anyway, I think it is fine to put that into package.bbclass now... bitbake should gain a function to parse a "fragment" Aug 14 14:31:57 but this can come later... adding it to the huge pile of things that would be nice to have Aug 14 14:31:59 yeah, that would be nice Aug 14 14:34:06 zecke: thanks for the review Aug 14 14:36:25 hrm, I wonder why libcurl puts this "-Wl,-O1" garbage in its .pc file Aug 14 14:36:29 * pb___ stabs it Aug 14 15:00:07 78 Aug 14 15:03:08 er, whoops Aug 14 15:26:54 hi kergoth Aug 14 15:41:50 re Aug 14 15:43:50 * Crofton|work curses this disk of x-rays+mri's that must be viewed on windows Aug 14 15:46:31 Crofton|work: anything good on there ? Aug 14 15:55:43 not sure Aug 14 15:55:48 the arm looks like steak Aug 14 16:12:02 hmm, wonder if anyone's done an m4 macro to detect and use the gold linker if it exists Aug 14 16:14:26 mm, dunno. I would have thought that the linker choice was more of a site/user preference than a package choice, really. Aug 14 16:14:32 people who want gold can export LD=gold or whatever Aug 14 16:18:52 thats true, i was thinking a configure option to switch to it when you're targeting elf/x86, but its easy enough to do outside Aug 14 16:21:25 Are the paths and prefixes in pkgconfig files under my staging dir supposed to refer to the same staging dir as opposed to just plain / ? Aug 14 16:21:35 yes Aug 14 16:21:54 kalev for native yes Aug 14 16:22:00 for target not Aug 14 16:22:10 aeh Aug 14 16:22:19 native points to staging Aug 14 16:22:24 target points to / Aug 14 16:24:01 ok, that's exactly how it is in my build Aug 14 16:24:06 * * OE Bug 1431 has been RESOLVED (INVALID) by Aug 14 16:24:08 * * Build of ixp4xx-kernel includes non-fatal depmod error Aug 14 16:24:10 * * http://bugs.openembedded.net/show_bug.cgi?id=1431 Aug 14 16:24:38 why I am asking is because of some xorg driver packages Aug 14 16:26:00 they get sdkdir=$(pkg-config --variable=sdkdir xorg-server), but the pkg config files they find are in target staging, so they end up trying to get include files from /usr/include/xorg Aug 14 16:26:29 kalev hm Aug 14 16:28:12 * Laibsch "bugs" kergoth about 518 Aug 14 16:28:18 in a very old OE build I had, those pkg config files in my target staging dir referred to staging dir, but I guess it has changed now Aug 14 16:34:52 kalev then you should update Aug 14 16:35:33 yes, I that's what I am doing Aug 14 16:36:12 my problem is that in newest OE I can't use pkg-config to get include dirs any more Aug 14 16:36:20 whereas it used to work in an old version Aug 14 16:37:05 kalev what are you trying to build? Aug 14 16:37:51 we have our own OE based distro Aug 14 16:38:44 trying to build xf86-video-via right now Aug 14 16:41:59 kalev: you might be better off with openchrome anyway Aug 14 16:42:29 that one did build ok, in .dev at least, last time I tried Aug 14 16:42:55 I get a similar error with vesa driver too Aug 14 16:43:36 I am not sure how it is supposed to work Aug 14 16:43:57 should pkg-config find the right includedir in my native staging dir? Aug 14 16:44:03 or target staging? Aug 14 16:44:44 if it is the first option, then I think I just lack some xorg native packages Aug 14 16:44:50 Laibsch: thanks, i needed that Aug 14 16:45:14 anyone played with ditz? Aug 14 16:45:23 kalev: if you're trying to build for the target, you want the target pkgconfig files Aug 14 16:45:50 ... but the target pkgconfig files refer to /usr/include Aug 14 16:46:01 that sounds bogus. I don't think they should be doing that. Aug 14 16:46:07 * * OE Bug 4474 has been marked as DUPLICATE of bug 3241 by Aug 14 16:46:09 * * meta-toolchain fails do_populate_sdk when GLIBC_GENERATE_LOCALES is set Aug 14 16:46:11 * * http://bugs.openembedded.net/show_bug.cgi?id=4474 Aug 14 16:46:44 ok, that's what I thought too, but woglinde earlier said that they should refer to / :) Aug 14 16:49:24 oh, anyone play with http://code.google.com/p/quagmire/ ? Aug 14 16:49:29 seems cute Aug 14 16:49:47 emulates autofoo style behavior with pure make Aug 14 16:49:51 kergoth`work: heh, you're a veritable fountain of new packages today Aug 14 16:50:03 i expect it must use a lot of crazy bleeding edge gnu make features Aug 14 16:50:18 heh, i find myself suddenly in the mood to catch up on such things Aug 14 16:50:21 yah, I guess so Aug 14 16:50:33 i used to know the best projects to use for a given purpose, at least in the embedded field.. but now i'm behind :P Aug 14 16:52:30 been debating trying to get out of software development and more into an architect role, but i've got a lot of work ahead of me to get there, knowledge wise Aug 14 16:52:31 heh Aug 14 16:58:25 pb___: any idea what could make my pkg-config files point to wrong directories? Aug 14 17:01:30 kalev: not offhand, sorry. I don't have an up to date build tree to look at here. Aug 14 17:01:41 ok, thanks Aug 14 17:01:41 all the trees I have on hand are using .stable, and I guess you are on .dev Aug 14 17:01:52 yes Aug 14 17:02:16 I do have a .dev build tree on my laptop but I'm not near it right now Aug 14 17:02:30 if you haven't figured it out in an hour or two, ask me again and I'll look at it then Aug 14 17:02:43 kergoth`work: heh, right Aug 14 17:09:06 * * OE Bug 3716 has been RESOLVED (FIXED) by Aug 14 17:09:09 * * git-native 1.5.3-r1 doesnt build Aug 14 17:09:11 * * http://bugs.openembedded.net/show_bug.cgi?id=3716 Aug 14 17:16:21 http://www.balister.org/~balister/shoulder-xray-dye/ Aug 14 17:16:26 for your entertainment Aug 14 17:17:07 * * OE Bug 3214 has been RESOLVED (FIXED) by Aug 14 17:17:09 * * Unable to build klibc Aug 14 17:17:11 * * http://bugs.openembedded.net/show_bug.cgi?id=3214 Aug 14 17:25:00 RP: Can we please see the stuff from bug 4437 in linux-rp as well? Aug 14 17:26:45 pb___: in the old OE version pkgconfig.bbclass used to mangle the dirs inside the files, but it doesn't do it any more. I wonder why. Aug 14 18:01:04 anyone here used makeqpf? Aug 14 18:06:55 kalev: oh, right. I guess it relies on PKG_CONFIG_SYSROOT_DIR now. Aug 14 18:08:17 obviously not what he wanted to hear Aug 14 18:08:19 oh well, heh Aug 14 18:23:56 anyone hear of a lock (watchdog timeout) in the rt_sigreturn path from an app getting signaled? Aug 14 18:26:10 never heard of that, why do you ask? Aug 14 18:34:02 customer bug Aug 14 18:34:10 * kergoth`work brushes up on kernel signal handling Aug 14 18:43:46 ah, pesky customers Aug 14 18:43:58 indeed Aug 14 18:44:04 indeed Aug 14 18:46:19 Hey guys! I want to have a package that only gets included in an image if a particular MACHINE is set, and otherwise just gets skipped. How can I do that? Set COMPATIBLE_MACHINE? PACKAGE_ARCH? Something else? Aug 14 18:47:34 DEPENDS_machine += "blah" Aug 14 18:47:36 my first guess... not that ive done it... would be Aug 14 18:47:38 RDEPENDS_machine += "blah" Aug 14 18:47:43 yep Aug 14 18:48:32 XorA|gone: Thanks, I'll try Aug 14 18:49:29 re Aug 14 18:49:36 XorA|gone: And it won't complain about a missing dependency if MACHINE is not bla? Aug 14 18:49:49 abraxa_: no Aug 14 18:50:10 it might need to be RDEPENDS_append_machine = "blah" Aug 14 18:50:16 try both see what works out Aug 14 18:50:23 XorA|gone: ok Aug 14 18:51:07 XorA|gone: Just to make sure - blah is the machine I want to verify against and both lines go into the bb that is supposed to become machine-specific? Aug 14 18:51:42 abraxa_: no, lines go into image.bb file Aug 14 18:51:49 and only one line RDEPEND Aug 14 18:52:06 XorA|gone: Ahhh... okay, thanks Aug 14 18:52:15 or something like that anyway Aug 14 18:52:24 * XorA|gone is a little tired Aug 14 18:53:12 re kalev Aug 14 18:54:18 XorA|gone: I appreciate the help regardless :) - and is there a quick way to describe the difference between COMPATIBLE_MACHINE and PACKAGE_ARCH? Aug 14 18:55:27 hi woglinde Aug 14 19:12:23 pb_: ok, so the problem is now this: some xorg drivers use "pkg-config --variable=sdkdir xorg-server", but PKG_CONFIG_SYSROOT_DIR only affects --libs and --cflags Aug 14 19:13:05 I guess I'll just have to patch those makefiles to do the right think and submit those patches to bugzilla Aug 14 19:44:17 wuff Aug 14 19:44:28 ant_: helo :) Aug 14 19:44:34 kalev: right, yes Aug 14 19:44:40 ant_: how it going? :) Aug 14 19:49:58 hey Aug 14 19:50:24 I'm looking for a cmdliner parser (initramfs-klibc-dash=no regexp) Aug 14 19:50:39 ant hehe Aug 14 19:50:41 aks thesing Aug 14 19:50:49 I'm trying with smthg like this one: http://bugs.archlinux.org/task/10381?getfile=2128 Aug 14 19:51:14 woglinde: yes..I don't like to reinvent a tokenizer... Aug 14 19:51:52 but thesing gently asked me to seek for a solution (mtdparts not passed to kexec kernel, well whole cmdline is not passed...) Aug 14 19:52:01 ah okay Aug 14 19:52:02 sorry Aug 14 19:52:07 I thought Aug 14 19:52:12 no sorry, help :-) Aug 14 19:52:22 I hope I can do it from init Aug 14 19:52:26 thesing had now a solution Aug 14 19:52:45 woglinde: http://git.linuxtogo.org/?p=thesing/kexecboot.git;a=tree Aug 14 19:52:53 lokke at the two .c files Aug 14 19:53:32 devicescan.c expects /boot/cmdline or defaults to null Aug 14 19:53:52 } else Aug 14 19:53:52 178 bl->list[count]->cmdline = NULL; Aug 14 19:59:24 what's kexecboot? Aug 14 20:00:14 is the new '2nd stage bootloader' kexecing different kernels/rotfs from a framebuffer menu Aug 14 20:00:29 ah Aug 14 20:00:56 finally old Zaurus can boot from sd Aug 14 20:02:25 hmm i 've got a do_compil error as i attempt to build fso Aug 14 20:02:36 openmoko-terminal2_svn.bb hang the build process Aug 14 20:02:48 http://rafb.net/p/oJIdoH53.html Aug 14 20:03:29 rather odd Aug 14 20:05:08 ant_: script from url looks like not dash (or what shell is in kexecboot?) Aug 14 20:06:12 I found dash in the sources Aug 14 20:07:08 ant_: did you want split mtdparts from cmdline and append to kexec'ed? Aug 14 20:07:17 .-)) Aug 14 20:07:35 or here is more? :) Aug 14 20:07:46 or write it to /boot/kernel-cmdline Aug 14 20:07:57 before launching kexecboot Aug 14 20:08:01 ah Aug 14 20:08:35 but I'm now wondering how to proceed: parse only for root= and mtdparts= ? Aug 14 20:09:01 imho, yes Aug 14 20:09:10 this two is only needed Aug 14 20:09:28 rootfstype? Aug 14 20:09:32 hm.. Aug 14 20:09:44 wait, it is deducted Aug 14 20:09:55 probing for boot devices Aug 14 20:11:21 pb_: I'm playing with ubi/ubifs. saw you were querying Aug 14 20:11:45 for me is enough just some editor to set up what to boot and how to boot from kexecboot :) Aug 14 20:12:21 dfoley: ah, very good. I'd be interested to know how you get on. Aug 14 20:12:25 but automation is needed.. Aug 14 20:12:35 for others :) Aug 14 20:13:47 but what happens when we just pass cmdline as is? Aug 14 20:14:13 well...it reads mtdparts from sharpsl.c Aug 14 20:14:36 ( and /home could overwrite /root) Aug 14 20:14:42 no, when we pass u-boots cmdline Aug 14 20:14:53 pb_: mounting and copying files seems Ok. haven't really extreme tested though, but seems to work Ok Aug 14 20:15:02 read it from u-boot env and store to /boot/kernel-cmdline Aug 14 20:15:09 without parsing Aug 14 20:15:21 dfoley: how does the performance compare with jffs2? Aug 14 20:15:26 yes, thesing said it would be complicated (u-boot-utils) Aug 14 20:15:42 but the string there is clean Aug 14 20:15:52 zero parsing = no errors Aug 14 20:16:05 hm.. Aug 14 20:16:28 and if you install debian you don't need mtdparts hardcoded Aug 14 20:16:33 in the kernel Aug 14 20:16:43 same for pdaxrom Aug 14 20:16:49 or android Aug 14 20:17:21 is same /boot/cmdline-kernel is used with any kexec'ed kernels? Aug 14 20:18:24 well, look at devicescan.c Aug 14 20:18:35 pb_: haven't really got that far, but when I load my files the mount is instant where jffs2 is many seconds, but I really need to do more testing Aug 14 20:19:23 damn apr Aug 14 20:19:27 stuff Aug 14 20:19:34 dfoley: ok, great. sounds encouraging. Aug 14 20:20:12 if you do any measurements of total boot time (say, power on to /sbin/init starting, or some such) then I'd be interested to see the results Aug 14 20:22:05 pb_: for sure Aug 14 20:23:03 ant_: well, here can be one cmdline-file on every bootable partitions.. Aug 14 20:23:25 not ideal... Aug 14 20:23:30 now it's clean Aug 14 20:24:07 it's in scan_devices() Aug 14 20:24:27 btw, zImage is hardcoded here :( Aug 14 20:24:42 ah, kexec.. Aug 14 20:24:49 well, zImage / zmage.bin is a mess... Aug 14 20:24:59 zImage.bin of course Aug 14 20:25:30 I've thinking about uImage but remember that here is kexec Aug 14 20:27:53 Jay7: we could set the mtdparts in /boot/kernel-cmdline at compile time using the same CUSTOM_ROOTFS_SIZE like u-boot, but then you'll need patched (foreign) kernels for that exact mtdparts Aug 14 20:28:48 mm... ok Aug 14 20:28:52 otherways soon or later one distro will wipe your partitions... Aug 14 20:29:24 we needed mtdparts only for kernel that booting from NAND Aug 14 20:29:29 is it right? Aug 14 20:29:41 only if you mount mtd's Aug 14 20:30:05 when I boot from SD what chances that I want mount mtd?.. Aug 14 20:30:18 only for /home.. Aug 14 20:30:19 you share /home Aug 14 20:30:22 :-) Aug 14 20:30:24 grr.. :) Aug 14 20:30:43 then we need append mtdparts for every cmdline.. Aug 14 20:31:07 but some zaurus kernels can just ignore them Aug 14 20:34:01 simpliest way is forbid using mtd when booted not from NAND :) Aug 14 20:34:44 hmm, http://liw.iki.fi/liw/unperish/README is a neat little app Aug 14 20:35:28 oh, cute Aug 14 20:35:40 I was just fighting with "make dist" a moment ago. Aug 14 20:36:13 reminds me of the old bitbake/oe building out of the package source tree thing, i always wanted that use case to work as well as the rest, but it was always only half there, too different a usage model Aug 14 20:37:05 * * OE Bug 3886 has been marked as DUPLICATE of bug 2891 by Aug 14 20:37:08 * * udev-118 initscript is very inefficent Aug 14 20:37:09 * * http://bugs.openembedded.net/show_bug.cgi?id=3886 Aug 14 20:41:05 * * OE Bug 4118 has been marked as DUPLICATE of bug 2891 by Aug 14 20:41:07 * * udev-118 failed to create device nodes on Tosa Aug 14 20:41:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4118 Aug 14 20:58:44 http://madduck.net/blog/2008.01.29:consolidating-packaging-workflows-across-distros/ Aug 14 21:14:08 Laibsch ? Aug 14 21:16:40 bluelightning ? Aug 14 21:16:59 Tartarus: hello Aug 14 21:18:23 bluelightning, for bug 2891, did you guys narrow things down to the same exact image, except for udev versions, for fixing that particular problem? Aug 14 21:18:51 Tartarus: yes Aug 14 21:19:02 OK Aug 14 21:19:30 I wonder if the scripts for doing stuff after it's found the device changed or are missing Aug 14 21:19:45 diffing /etc/udev between the two might help Aug 14 21:20:29 I'm not really in a position to do that atm I'm afraid Aug 14 21:20:33 OK Aug 14 21:20:41 I'll follow up on the bug then, thanks Aug 14 21:20:52 no, thank you :) Aug 14 21:23:13 Jay7: I'm looking at /sources/klibc-1.5/usr/dash/builtins.def.in Aug 14 21:24:22 well there is not so much for string parsing, apart including string.h and go Aug 14 21:25:26 well, strstr is there Aug 14 21:29:14 Crofton, ping Aug 14 21:29:41 pong Aug 14 21:30:17 Crofton|work, finally got back to my issue regarding using COLLECTIONS to provide local overrides for files included by packages such as netbase and base_files Aug 14 21:30:46 Crofton|work, looking at bitbake-1.8.10 I notice that COLLECTIONS is 'only' used for bb files - NOT for local files Aug 14 21:31:04 what are local files? Aug 14 21:31:08 so I'm not sure why it would work to override local files 'without' also providing a local copy of the recipe Aug 14 21:32:13 recall that I'm using the 'overlay method' where I have a local tree and an org.openembedded.dev tree - I desire org.openembedded.dev/packages/netbase for example to use interface file from local/packages/netbase/netbase/interfaces instead of org.openembedded.dev/pacakges/netbase/netbase/interfaces Aug 14 21:33:02 tharvey, I think you need to talk to someone smarter than me :( Aug 14 21:34:24 heh, ok... we were discussing this the other day as this is the same method gumstix-oe uses Aug 14 21:34:35 yeah Aug 14 21:34:43 they have had very good results Aug 14 21:34:52 I am not sure why you would have issues Aug 14 21:35:30 so for example gumstix-oe uses this for base-files, base-passwd, and netbase packages, however for base-files and netbase they 'do' provide their own local bb file with a minor mod to the orig - so I can see how it works for those Aug 14 21:35:36 but I can't see how it would work for base-passwd Aug 14 21:38:13 rehi Aug 14 21:38:48 thesing: hello Aug 14 21:39:02 I'll have to build gumstix-oe and see if base-passwd does what I think it should do - can't see how it would use the local files provided over the org.openembeeded.dev ones Aug 14 21:40:18 that would be good Aug 14 21:40:25 also try getting on here early Aug 14 21:41:03 ya Aug 14 21:43:07 tharvey: I think you're right that COLLECTIONS doesn't do anything special with FILESPATH, but you can bend the latter by hand if you want to. Aug 14 21:44:25 there's nothing magic about that, it's just a regular variable and you can set it in local.conf like anything else. Aug 14 21:49:20 pb_, yes, I was looking at that but its not clear what I would set that to to search in local as well as org.openembedded.dev Aug 14 21:49:44 just put your local paths on the front Aug 14 21:51:13 pb_, I would need to set it to the relative path of the current recipe with a leading 'local' instead of a leading org.openembedded.dev - unless I'm misunderstanding Aug 14 21:51:30 yes, exactly Aug 14 21:51:58 but FILE_DIRNAME (used as the base for all paths in FILESDIR) is full path or relative? Aug 14 21:52:55 full, I think, but it doesn't really matter. you can use ${FILE_DIRNAME}/../../local/packages/${PN} or some such in either case. Aug 14 21:53:26 or just ${LOCAL_PKG_ROOT}/${PN} if you have an existing variable for the root of your local tree Aug 14 21:53:30 ya, but I don't know how many paths to go back... Aug 14 21:53:47 yes, I thin I would have to do it the 2nd way Aug 14 21:54:15 oh, right, you have .bb files at varying depths in the hierarchy? Aug 14 21:54:23 er... actually all packages are in org.openembedded.dev/packages not subdirs of that right? Aug 14 21:54:24 in that case you could write a little python function to figure it out, I guess Aug 14 21:54:58 does the oe tree use varying depths in packages ? Aug 14 21:55:01 yes, in standard .dev I'm pretty sure everything is in org.openembedded.dev/packages/*/*.bb Aug 14 21:55:15 i.e. the depth doesn't vary Aug 14 21:55:55 ahh, this live/dead lock with sigreturn is when there's a segfault inside a signal handler.. wonder if its only when theres a sigsegv in a handler, or if its from any unblocked/masked signal.. some days i hate working with older kernels... Aug 14 21:55:57 * kergoth`work grumbles Aug 14 21:56:35 pb_, ah good... if bb files are only ever 1 dir deep in packages then your first example using ${FILE_DIRNAME}/../../local/packages/${PN} would be the simple solution Aug 14 21:57:02 tharvey: I'm not 100% sure that they are always 1 directory deep, but it would be easy to check using 'find'. Aug 14 21:57:22 pb_, just for my reference, if I 'did' need to write some python code it would be in the syntax of "${@pythoncode}" ? Aug 14 21:57:37 yes, that would be the easiest way Aug 14 21:58:02 annoyingly you can't define functions in .conf files, so the "def foo:" syntax is unavailable to you unless you also write a supporting class file. Aug 14 22:02:48 pb_, its also unclear how I would write any python statement that required multiple lines/indention Aug 14 22:03:55 pb_, to add a class to be used in the ${@...} notation I would need to extend bitbake with a class? Aug 14 22:07:00 tharvey: ${@...} is a single expression. you can't have multiple lines there. Aug 14 22:07:14 not quite sure what you mean by "extend bitbake with a class" Aug 14 22:11:17 pb_, right... and so if I wanted some crazy function to use with ${@...} that wasn't part of an existing class I would have to much with bitbake to add it Aug 14 22:11:33 no, you can just create your own .class file containing that single function Aug 14 22:11:37 you don't need to do anything to bitbake Aug 14 22:11:44 er, .bbclass I mean Aug 14 22:12:36 pb_, regarding the varied depth of packages/ - looks like asside from 'obsolete' and 'nonworking' paths (which I assume have packages that should be avoided?) the only one that stands out is e2fsprogs which has an extra depth for some reason Aug 14 22:13:15 pb_, yes, but if you use your own .bbclass you can use that in a package that 'inherits' it but can't use it in a conf file right? (or can you 'inherit' in a conf?) Aug 14 22:13:37 yes, you can set INHERIT directly in a .conf file Aug 14 22:13:47 see, for example, sanity.conf Aug 14 22:14:17 tharvey: I think the e2fsprogs thing is probably an error Aug 14 22:14:52 pb_, cool - thx for the discussion - I think I'll be able to get what I need done now Aug 14 22:14:59 I see that the sample local.conf suggests BBFILES = "packages/*/*.bb", which would basically enforce a consistent depth for most users Aug 14 22:15:32 and, as you say, obsolete and nonworking are probably not packages you want :-} Aug 14 22:15:41 ok, great. good luck. Aug 14 22:15:43 I think perhaps gumstix-oe doesn't do what the authors thinks it does for a couple of packages b/c of COLLECTIONS being used only for bbfiles and not for files Aug 14 22:16:25 pb_, yes.. I see that about BBFILES - that explains how packages/obsolete and packages/nonworking get avoided Aug 14 22:16:25 yeah, very likely. the semantics of COLLETIONS are probably not all that well documented. Aug 14 22:17:10 the project that I originally wrote the collections support for got canned just after I finished implementing that stuff and I don't think I ever had the motivation to write any docs for it. Aug 14 22:17:10 actually COLLECTIONS is documented pretty well I think - the key is that they call it 'bbfile collections' and not just 'collections' - http://bitbake.berlios.de/manual/ch04s02.html#id870544 Aug 14 22:17:38 ah, very good Aug 14 22:18:16 well... this 'overlay method' (I've seen discussed here but never in any online docs) seems to be a popular choice for those who want to provide local changes to a snapshot of OE - gumstix-oe definately makes use of it and it makes sense there Aug 14 22:18:46 i'm actually kind of surprised that bitbake's method of locating its data hasn't been revamped by now. OEFILES was always intended to just be a temporary hack of a method :P Aug 14 22:20:03 kergoth`work: yah, it's funny. this is one of those things about oe that obviously sucks, but it's hard to think of any alternative that would be better. Aug 14 22:20:12 true Aug 14 22:21:27 * pb_ bedtime now Aug 14 22:21:29 night all Aug 14 22:21:32 night Aug 14 22:21:32 the poky manual has a better description of collections where as the bitbake manual just shows an example - http://www.pokylinux.org/doc/poky-handbook.html#usingpoky-changes-collections Aug 14 22:21:39 night Aug 14 22:22:44 kergoth`work, OEFILES? Aug 14 22:23:00 that's what it was, originally, before things were renamed Aug 14 22:23:18 I really like the concept of COLLECTIONS but I think it should apply towards local files as well as bbfiles/conf/classes Aug 14 22:24:14 i always thought itd be nice for a local overlay to be able to contain deltas against the main repo's .bb files, instead of having to copy them and keep them in sync manually.. never did find a good way to do that sort of thing Aug 14 22:24:45 yes, but deltas (if you mean something like diffs) are difficult to read Aug 14 22:25:16 I just don't want to be in the habit of copying recipes 'just' to allow local file overrides Aug 14 22:27:37 kergoth: to do that, I have a quilt patch tree which I unapplied before mtn update, then reapplied after. Aug 14 22:28:12 yeah, i used to do that too Aug 14 22:39:35 * Jay7 -> sleep() Aug 14 22:58:32 ack! whats the good of defining FILESPATH in bitbake.conf when base.bbclass sets it? Aug 15 00:57:05 * * OE Bug 4498 has been created by  Aug 15 00:57:07 * * webkit_qt_svn fails do_patch Aug 15 00:57:10 * * http://bugs.openembedded.net/show_bug.cgi?id=4498 Aug 15 01:00:05 * * OE Bug 2122 has been RESOLVED (INVALID) by Aug 15 01:00:07 * * libgemwidget build failure Aug 15 01:00:09 * * http://bugs.openembedded.net/show_bug.cgi?id=2122 Aug 15 02:19:54 * rwhitby wishes there was a way to stop other kernels trying to be built when the kernel you want has a fetch failure ... **** ENDING LOGGING AT Fri Aug 15 02:59:56 2008