**** BEGIN LOGGING AT Fri Jul 16 02:59:56 2010 Jul 16 03:54:54 aditya_1010, if you ask a specific question, you might even get a response. Jul 16 03:56:28 grg: do you use uclibc with mipsel ? Jul 16 03:56:43 khem, i haven't, no Jul 16 03:57:04 khem, could probably justify playing with it at some point though Jul 16 03:58:34 grg: ok Jul 16 03:58:41 right now it dies pretty early Jul 16 03:58:45 mips works ok Jul 16 03:58:50 LE seems to have issues Jul 16 03:58:55 but I will weed it out Jul 16 03:59:37 ok, cool Jul 16 05:52:20 grg saw your lsof patch, good work, but, for my education, can you explain why the file:// has been dropped from +LOCALSRC = "${WORKDIR}/lsof_${PV}/lsof_${PV}_src.tar Jul 16 05:53:06 eFfeM_work, because i just passed LOCALSRC to a tar command directly. And tar doesn't know about file:// Jul 16 05:54:02 ah ok, didn't know LOCALSRC still had the file:// prefix Jul 16 05:55:09 tar gives a weird error too... Jul 16 05:55:11 grg@eye7:~$ tar x --no-same-owner -f file://asdf Jul 16 05:55:11 tar: Cannot connect to file: resolve failed Jul 16 05:59:08 grg this means tar has the notion of things like http://, guess one can do tar xvf http://xx.tar Jul 16 05:59:50 i noticed that somewhere in the past tar xvf xx.tar.gz will also work even if the z flag is not there, this used to be different in the old days Jul 16 06:00:44 yeah, i only noticed that this year... not sure how long gnu tar has supported that. Works for .tar.bz2 and .tar.Z too. Jul 16 06:03:13 the colon notation appears to be hostname:file format, a la scp. I'll never know the real deal though as gnu people write shitty incomplete man pages, because they still think info is a good idea Jul 16 06:06:39 tar xf is what you need Jul 16 06:09:58 i like to see that it is actually doing something so I typically add a v Jul 16 06:10:22 especially with c, so I can see the right files are added Jul 16 06:10:33 eFfeM_work: sorry for loss in WC final :) Jul 16 06:11:00 Jul 16 06:11:02 I reckon you are in netherland Jul 16 06:11:07 I don't care too much about soccer Jul 16 06:11:12 yes I am in NL Jul 16 06:20:27 eFfeM_work: I did not approve the glibc recipe removal patch because I think someone who wants to use latest OE but still wants to use older tools should not be barred Jul 16 06:21:04 khem, no problem Jul 16 06:21:28 the one I was referring to in my email reply from 30 mins or so ago was the glibc 2.5 cleanup that I've sent before Jul 16 06:21:48 eFfeM_work: although I agree that we should remove old recipes Jul 16 06:22:03 it should be agreed upon as OE's future direction Jul 16 06:22:17 thats ok Jul 16 06:22:19 an acked Jul 16 06:22:54 * khem has had enough of python today Jul 16 06:23:16 wrt the removal: the versions themselves are not pinned and esp the gcc 2.3.3 and 2.3.5 which have a mainline version and two cvs variants seem tobe a little bit overkill Jul 16 06:23:33 also taking into account they are versions from 2004 Jul 16 06:23:37 and 2005 Jul 16 06:24:04 I agree on these accounts but there has to be consensus Jul 16 06:24:48 may be we should only support handful of fully functional combinations at a given point of time Jul 16 06:24:50 hard to get consensus on that, some people do not seem to understand the powers of a revision control system fully Jul 16 06:25:28 there is some TSC ruling on it Jul 16 06:25:36 eFfeM_work: well revision control is not going to help think of if one wants gcc 3.4.x + glibc 2.4 Jul 16 06:25:36 let me check Jul 16 06:25:41 true Jul 16 06:25:46 and we removed it Jul 16 06:25:53 he has to roll back Jul 16 06:26:00 and also lose the other OE changes Jul 16 06:26:07 which may not be what he wants Jul 16 06:26:54 true (but to some extend overlays can help there) Jul 16 06:27:01 We just have to convince that there is no enough use cases Jul 16 06:27:30 i think we have too much stuff around that has been created ages ago and not been cared after Jul 16 06:28:00 btw the tsc discussion is minuted at http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-May/020047.html Jul 16 06:28:24 yes Jul 16 06:28:27 it would make sense to remove all the older versioned recipes (e.g 2.3.x) except the most recent minor versions Jul 16 06:28:42 thats the reason for things like sane-toolchain.inc Jul 16 06:29:04 so, given that there is a 2.3.6, all the other 2.3.x should be removable Jul 16 06:29:18 (i feel similarly about gcc recipes) Jul 16 06:29:19 grg: for gcc in future we will only track major releases Jul 16 06:29:28 like 4.5 and 4.6 etc Jul 16 06:29:29 grg, I sumbitted a patch for glibc to remove all versions that are not pinned Jul 16 06:29:41 nomore 4.5.x Jul 16 06:29:44 eFfeM_work, i saw that. I like. Jul 16 06:30:06 grg, feel free to express that in the thread and/or ack Jul 16 06:30:08 khem, so you're just going to track the gcc-4.x branches directly? Jul 16 06:30:14 yes Jul 16 06:30:28 makes sense Jul 16 06:30:35 khem, I'm wondering if it is useful to keep rarely used stuff like gcc 4.1.0 around Jul 16 06:30:57 eFfeM_work: as I said if there are no use cases then may be not Jul 16 06:31:09 its hard to tell Jul 16 06:31:16 things one doesnt use seems useless Jul 16 06:31:24 but that may not be case Jul 16 06:31:35 (and actually apparently we skipped 4.4.3) Jul 16 06:31:37 we have to RFC is properly Jul 16 06:31:41 you could intentionally commit a breakage to those old recipes and see how long it takes for anyone to notice Jul 16 06:32:14 grg: heh well many people dont sync daily Jul 16 06:32:28 khem, or even yearly Jul 16 06:32:34 and some have branches syncing every now and then Jul 16 06:32:38 khem, I agree Jul 16 06:33:07 (on rfc and also on syncing) Jul 16 06:33:21 as I said its good think to remove them but it has to be widely advertised beforehand Jul 16 06:33:42 btw it would be fine with me to keep those old recipes around for a while in the stable branch Jul 16 06:34:34 but dev head is bleeding edge and should aim to move fwd; if you do want e.g. gcc 3.3.3 then you should probably not be using dev but stable Jul 16 06:34:38 OE is like a shipyard you pick and chose and make your ship Jul 16 06:34:40 :) Jul 16 06:34:58 so more kind of spare parts much bigger the shop Jul 16 06:35:06 khem, nice analogy Jul 16 06:35:30 I believe thats the strength of OE Jul 16 06:35:33 in one sense Jul 16 06:35:37 but unfortunately lots of parts do not go together well, so it is harder to construct a good ship Jul 16 06:36:08 I have worked at a job where we tried to shovel same set of versions to every customer and it did not work Jul 16 06:36:14 yes. And the more choice between broken parts there is, the harder it becomes to make your ship Jul 16 06:36:50 you try to pick the ones that are broken in the least obscure ways Jul 16 06:37:37 grg: embedded world is very fragmented Jul 16 06:37:48 I still get request to fix 2.95 gcc Jul 16 06:37:54 so I'd rather have one casco (is that a good english word?) and allow people to pick their options Jul 16 06:38:03 khem 2.95, ouch Jul 16 06:38:05 not in OE ofcourse Jul 16 06:38:12 hope you are paid well for it :-) Jul 16 06:38:14 2.95 was a nice release Jul 16 06:38:32 nice and fast compared with 3.x Jul 16 06:38:34 eFfeM_work: I fix atleast 2-3 bugs a month Jul 16 06:38:38 in 2.95 Jul 16 06:38:53 well not so much lately Jul 16 06:39:12 i think i did a project with it very long ago, I'm in the embedded sw business for most of my career and I know how it works Jul 16 06:39:21 at times it ranges from 2.95 to 4.5 Jul 16 06:39:30 make product, derive variant, derive variant Jul 16 06:39:49 right and upgrades are hard sell Jul 16 06:40:09 yes Jul 16 06:40:37 then there are folks who care about GPLv3 Jul 16 06:40:56 and then people come up and say I have this 2.4.something kernel, but we need to have feature x which is only available in 2.6.33+ Jul 16 06:40:56 and want to stay away at any cost Jul 16 06:41:40 I think sometimes its a choice between backport or forward port Jul 16 06:41:41 khem, my previous employer was in that group, they had quite a lot of basic patents that they did not want to make publically available Jul 16 06:41:44 one has to balance Jul 16 06:42:19 it costs both ways just see which one is cheaper and faster to market Jul 16 06:42:25 GPLv3 was an absolute no over there, they even had a guy checking all kinds of licenses Jul 16 06:42:34 heh Jul 16 06:43:02 actually there is something called GPLv3 contamination Jul 16 06:43:19 which means even if you looked at the code your brain cells are tainted Jul 16 06:43:42 haha Jul 16 06:43:46 that's gold Jul 16 06:43:48 :-) Jul 16 06:44:17 gplv3 needs a "wetware" exemption :) Jul 16 06:49:18 well I have gone thru all this crap Jul 16 06:49:27 :) Jul 16 07:08:56 morning Jul 16 07:09:09 anyway to conclude the glibc discussion, I don't mind doing some housekeeping and cleaning up some recipes, but I am not really interested in wasting time cleaning up recipes that are apparently not used Jul 16 07:09:13 hi hrw, gm Jul 16 07:12:30 eFfeM_work, yep. I totally understand. Jul 16 07:13:04 and that's why large scale recipe clean ups don't happen when they should - no one else wants to clean up the old recipes that aren't used either Jul 16 07:13:40 because it means *testing* them too. Jul 16 07:18:03 eFfeM_work: fetch mail Jul 16 07:19:10 good evening Jul 16 07:19:32 hrw, read it, good idea Jul 16 07:19:51 eFfeM_work: same for other components Jul 16 07:19:56 sure Jul 16 07:20:07 can someone please look at the problem I encounter: http://paste.pocoo.org/show/238207 Jul 16 07:20:33 this is after update and do the fresh compile Jul 16 07:21:25 IgorK, "See `config.log' for more details." Can you post this as well? Jul 16 07:23:24 IgorK: there are two errors in your log Jul 16 07:23:25 IgorK, actually, there's probably no need Jul 16 07:23:31 i see: crti.o: No such file Jul 16 07:23:42 237./238: cannot create exe Jul 16 07:24:21 that is the configure failing Jul 16 07:24:41 and the other one is line 160 where it cannot fidn crti.o Jul 16 07:25:29 the configure failure to find a working compiler is doubtless due to the same missing file Jul 16 07:25:37 grg: do u need an error.log or not? :) Jul 16 07:25:43 IgorK: did you do an rm tmp before this ? Jul 16 07:26:01 yes, freshly new build Jul 16 07:26:11 IgorK, no, it won't help narrow down the problem further Jul 16 07:26:47 i don't know why your toolchain is borked Jul 16 07:26:48 did "git pull && rm -rf tmp && bitbake angstrom-gnome-image" Jul 16 07:27:12 IgorK, can you paste your local.conf? Jul 16 07:28:36 IgorK: but your TMPDIR is called temp, not tmp Jul 16 07:28:57 http://paste.pocoo.org/show/238208 Jul 16 07:29:14 rm -rf /home/igor/OE_HEAD/openembedded/temp && bitbake angstrom-gnome-image Jul 16 07:32:56 well, its 5pm on a friday evening... time for me to scuttle off... Jul 16 07:33:00 bye folks Jul 16 07:33:06 grg, have a nice weekend Jul 16 07:33:45 * eFfeM_work still has 7 hrs to go before weekend starts; the good thing is that I still have weekend when grg starts to work :-) Jul 16 07:35:56 eFfeM_work: ok, will do. Jul 16 07:36:25 Do you know which option I need to reset to not save temp files? Jul 16 07:36:53 IgorK: there is something that tells you to rm things in the work dir Jul 16 07:36:58 don't know details Jul 16 07:40:19 afk Jul 16 07:41:21 http://www.kdedevelopers.org/node/4278 Jul 16 08:07:44 hrw, Re the binutils ?!nios2; couldn't you just have a function in binutils that basically says nios2_compatible=return PV > 2.20 ? Jul 16 08:08:33 thats too but we try to avoid anonymous functions in metadata Jul 16 08:09:24 ah ok, didn't know that Jul 16 08:12:07 i'm surprised that there is no generic set of version comparison helpers (that i can see)? Jul 16 08:15:59 there are functions to compare versions Jul 16 08:16:22 where? Jul 16 08:16:39 i must be blind or something ;) Jul 16 08:19:01 base_version_less_or_equal Jul 16 08:24:36 thanks Jul 16 08:34:56 hi guys. how high is the chance that a gcc 4.4.3 (from Ubuntu 10.4) miscompiles something for amd64 nowadays. I am seeing random segfaults and if I put a fprintf() in the respective code (which looks fine btw) or compile with debug flags the problem does not show up (heisenbug :) ) Jul 16 08:35:31 (just to make it clear it is about the classpath-native recipe in OE) Jul 16 08:36:51 btw: I was countering seemingly strange GCC behavior in classpath before by declaring a variable as volatile. Jul 16 08:37:29 when compiling with -O2 the variable was optimized out, with -O0 the problem did not appear :( Jul 16 09:02:27 03Henning Heinold  07org.openembedded.dev * r0f889c1531 10openembedded.git/recipes/llvm/llvm-common.bb: Jul 16 09:02:27 llvm-common: use do_stage for now, until there is a better solution Jul 16 09:02:27 * remove the extra PR and bump PR Jul 16 09:27:30 thebohemian: I have ubuntu 10.4 on amd64 too; if you have a simple testcase, maybe I can compile it on my box to see if it is consistent Jul 16 09:28:26 I have 10.10-dev on my amd64 - can test too Jul 16 09:30:56 thebohemian: miscompilations certainly aren't unheard-of. could try 4.4.4 or 4.5.0 and see if they behave any better. Jul 16 09:31:18 hm, thanks guys. but I dont know how this could be done. the segfault happens (randomly) in a function that looks completely sane to me: http://pastebin.ca/1901649 Jul 16 09:31:58 btw. is there a way of setting up a program in a way that it automatically starts gdb when the segfault happens? Jul 16 09:34:06 thebohemian: not trivially. you could use catchsegv if you just want a backtrace and crash dump. Jul 16 09:34:28 obviously you can write a SIGSEGV handler to do whatever you like, including launching gdb Jul 16 09:42:50 hm, the problem with launching gdb is that the console is hidden from me (AFAIK) since it runs as a task in bitbake. I was playing around with gdbserver for that job but that required me to start a corresponding gdb for each invocation of the program (of which there are a lot) Jul 16 09:43:40 * thebohemian wonders if screen could be of help here Jul 16 09:43:58 oh, I see. that's a bit tiresome. Jul 16 09:44:07 in that case I would start with catchsegv, which will just go to the logs as usual Jul 16 09:44:18 yes, I will start with that at first Jul 16 09:44:59 for gdb, I guess you could do what the kernel does with "make menuconfig" for example, and run it inside an xterm Jul 16 09:45:21 when I get a backtrace like this: ??:0(main)[0x400656] Jul 16 09:45:21 is there an easy way to find out what line of code in the program it is? Jul 16 09:45:28 addr2line Jul 16 09:46:17 another option with gdb would be to enable core dumps and then debug from the core after the crash rather than interactively at run time Jul 16 09:52:35 03Graham Gower  07org.openembedded.dev * re2cd1d0756 10openembedded.git/recipes/lsof/lsof_4.83.bb: Jul 16 09:52:35 lsof_4.83.bb: Make recipe buildable again. Jul 16 09:52:35 Breakage introduced by 74e32f114272668dfa4c88805741cf55ef965989. Jul 16 09:52:35 Other recipes which call base_do_unpack directly do not share this issue. Jul 16 09:52:35 Signed-off-by: Graham Gower Jul 16 09:53:05 hey thanks a lot phill! will try those approaches to find out what goes wrong in that function Jul 16 11:24:31 hello, is sb there i ve some problem to build an image for an at91 ... qemu failed every time :( Jul 16 11:35:09 i want to build stable/2009 and get this error message NOTE: Tasks Summary: Attempted 314 tasks of which 306 didn't need to be rerun and 1 failed. ERROR: '/home/mobile/stuff/openembedded/recipes/qemu/qemu-native_0.10.3.bb' failed Jul 16 11:43:16 prinzenrolle: it's probably also telling you to look into a log file :) Jul 16 11:43:45 there you can see what happened Jul 16 12:00:38 can you look here http://www.at91.com/forum/viewtopic.php/f,12/t,19390/ Jul 16 12:05:00 install the host qemu Jul 16 12:05:19 and make in ASUMMED_PROVIDED in your local.conf Jul 16 12:06:57 03Robert Schuster  07org.openembedded.dev * ra573f4c1b5 10openembedded.git/recipes/jamvm/ (files/java jamvm-native.inc jamvm-native_1.5.3.bb): jamvm-native: Add automatic restart workaround to native package as well. Jul 16 12:06:58 03Robert Schuster  07org.openembedded.dev * r505ca9b45d 10openembedded.git/recipes/jamvm/ (files/java jamvm-native.inc jamvm-native_1.5.3.bb): jamvm-native: Removed restart workaround, use libffi instead of custom assembly code. Jul 16 12:07:37 woglinde: I assume the random crasher with jamvm is fixed by this (just rebuilding icedtea6-native a 2nd time to be sure) Jul 16 12:07:46 cool Jul 16 12:07:57 thebohemian I am driving crazy with llvm-common Jul 16 12:08:07 woglinde: my assumption is that the custom assembly code that handles JNI calls for amd64 is buggy Jul 16 12:08:20 I have a very suspicious looking coredump Jul 16 12:08:25 uh I saw segfaults on my i686 too Jul 16 12:08:31 but now I saw only hangs Jul 16 12:08:34 but with gdb Jul 16 12:08:47 attaching Jul 16 12:08:57 breaking and continuing Jul 16 12:08:58 where the program flow continued at some random place instead of returning Jul 16 12:09:02 it worked futher Jul 16 12:09:17 woglinde: that could well be another symptom Jul 16 12:09:35 the ZIPEXE override worked Jul 16 12:10:17 neat. one less patch to think about Jul 16 12:12:09 damn Jul 16 12:12:13 I didnt worked Jul 16 12:12:14 *sigh* Jul 16 12:12:31 forgot that I installed zip again Jul 16 12:12:39 on host system Jul 16 12:13:11 not a good day Jul 16 12:13:25 besides wayfinder went opensource Jul 16 12:13:55 :) Jul 16 12:16:27 hm llvm-common still dont works Jul 16 12:17:43 woglinde: oh noes! Jul 16 12:17:55 woglinde: what is the problem? Jul 16 12:18:02 do_stage Jul 16 12:23:07 please help me, why this happens when i fire bitbake nano Jul 16 12:23:28 this gives error; error location http://pastebin.com/vwchXTmm Jul 16 12:23:46 it for all package file Jul 16 12:24:22 akumar you can read and understand english? Jul 16 12:26:44 hm do_stage dont exists for llvm-common Jul 16 12:26:46 hrms Jul 16 12:33:03 woglinde: ?!? Jul 16 12:33:39 akumar: kidding? "Please install following missing utilities: svn" Jul 16 12:34:37 woglinde: what's the issue with llvm-common. all it does is installing a simple shell script into either STAGING_BINDIR (-native) or STAGING_BINDIR_CROSS (non-native). dead simple IMO Jul 16 12:35:35 too bad however that I dont see that this can be done without a do_stage() function to be written. and I somehow have this feeling that STAGING_BINDIR_CROSS had been forgotten when do_stage() was deprecated Jul 16 12:35:35 thebohemian: i have already install svn, but is not resolve the problem. Jul 16 12:36:40 and calling 'svn' on the commandline works as expected? what kind of distro, machine is that? Jul 16 12:37:09 akumar: you should give more information if you need help (local.conf, machine, OS) Jul 16 12:39:58 thebohemian: local.conf at http://pastebin.com/nQDtyU8W, machine: beagleboard, OS: ubuntu Jul 16 12:42:00 akumar: sorry, I'm off for weekend! Jul 16 12:42:15 akumar: but from cursory looking at it, it looks sane Jul 16 12:42:27 fine Jul 16 12:45:12 thebohemian yes should be so simple but not with the new staging system Jul 16 12:46:07 what happens when you type svn help? Jul 16 12:50:06 hi jconnolly|away Jul 16 12:50:14 hi gnutoo Jul 16 12:50:17 hi woglinde Jul 16 12:50:32 any objection about pushing a commit that split license and firmwares Jul 16 12:50:40 and firmwares RDEPENDS on license Jul 16 12:50:56 it's because 2 firmwares install the license in the same place Jul 16 12:51:18 no Jul 16 12:51:29 marvell_gspci and marvell_sdio for 88xxx Jul 16 12:51:33 aka libertas Jul 16 12:51:37 ok thanks Jul 16 12:51:50 it's needed for the bug device to get it's kenrel modules Jul 16 12:51:50 hm yes Jul 16 12:51:52 damn Jul 16 12:52:02 else opkg fails at do_rootfs Jul 16 12:52:11 llvm-common problem Jul 16 12:52:12 and you don't get the kernel-modules Jul 16 12:52:15 ah ok Jul 16 12:52:38 do_stage isnt working properly Jul 16 12:52:52 do_stage is deprecated/removed Jul 16 12:53:09 woglinde: hi! Jul 16 12:53:13 jo zecke Jul 16 12:56:17 what the hell Jul 16 12:56:58 woglinde: yeah? Jul 16 12:57:46 *sigh* we only want a simple package which install a file into staging_hostbla/usr/bin/targetfoo/ Jul 16 12:58:01 but this seems not to work with the BBCLASSEXTENDED Jul 16 12:59:07 do_stage dont exist Jul 16 12:59:51 *sigh* seems I need to split into two .bb again Jul 16 13:01:14 or should I use do_install now? Jul 16 13:01:17 hm Jul 16 13:03:07 hi Jul 16 13:03:12 hi Jul 16 13:03:45 is there a fast way to get all freesmartphone stuff out of my image, I'm targetting an x86 Jul 16 13:04:04 why you build such an image? Jul 16 13:04:09 make your own image Jul 16 13:04:43 that's what I'm onto, I'm using xorg-image as base Jul 16 13:04:54 remove the task Jul 16 13:04:57 and you are done Jul 16 13:05:34 woglinde, do_install I think Jul 16 13:05:48 yeah Jul 16 13:07:39 geo_gcju, hi Jul 16 13:07:45 hi Jul 16 13:07:59 geo_gcju, you've got an x86 phone and what freesmartphone stuff on it? Jul 16 13:08:21 no, I want a distro for a "normal" pc and want fso stuff out of it Jul 16 13:08:26 ah ok Jul 16 13:09:12 so look at the exisgins tasks/images and look how to add freesmartphone stuff to your image Jul 16 13:09:20 I looking at task-base-extended, that seems to be the place these things Jul 16 13:09:27 hmmm no Jul 16 13:09:34 I'll look if you want Jul 16 13:09:42 task-base-extended seem core Jul 16 13:09:47 03Henning Heinold  07org.openembedded.dev * r145cb3f5d6 10openembedded.git/recipes/llvm/llvm-common.bb: Jul 16 13:09:47 llvm-common: use do_install to do the stuff Jul 16 13:09:47 * bump PR Jul 16 13:10:09 add that: Jul 16 13:10:10 task-fso2-compliance Jul 16 13:10:24 to get rid of it? Jul 16 13:10:28 or rather: Jul 16 13:10:34 IMAGE_INSTALL += "task-fso2-compliance" Jul 16 13:11:09 okay, but that's for installing things, not removing or? Jul 16 13:11:22 ok Jul 16 13:11:25 there are 2 approach Jul 16 13:11:34 remove unwanted stuff in an image Jul 16 13:11:40 or add wanted stuff in an image Jul 16 13:11:48 I was explaining the second one Jul 16 13:11:54 which require to choose an image Jul 16 13:12:01 like a gnome or xfce46 image Jul 16 13:12:11 and then adding fso stuff Jul 16 13:12:19 it would be better than the first approach Jul 16 13:12:25 because you'll have a desktop image Jul 16 13:12:33 with a desktop windows manager Jul 16 13:12:35 etc... Jul 16 13:12:43 but do you need phone applications etc...? Jul 16 13:13:05 no, just Xorg and openbox Jul 16 13:13:17 console-image can be good start from then :) Jul 16 13:13:29 so make your own image maybe Jul 16 13:13:37 the manual and existing images can help you Jul 16 13:14:20 * Jay7 should return one of zauruses to usable state.. Jul 16 13:14:28 yes, I have a minimal-base-image, looks good so far Jul 16 13:15:09 thought xorg-image would be just console+xorg, but lots of things behinde the bitbake wall I haven't figured out yet Jul 16 13:27:38 GNUtoo|laptop: now I'll give it a try with: http://pastebin.com/z60kr1db Jul 16 13:28:05 ok Jul 16 13:41:40 thx i installed qemu and test it now Jul 16 13:42:31 i ve again a really stupid question but i dont know it... Jul 16 13:43:45 i ve an hardware from atmel at91SAM9G45-EKES with 256MByte NAND Flash and 256MByte RAM Jul 16 13:44:36 where is the filesystem located..? Jul 16 13:44:43 is in the nand device? Jul 16 13:45:02 free say i ve only 128MB where is the rest ram? Jul 16 13:45:10 root@at91sam9g45ekes:~# df -h Filesystem Size Used Available Use% Mounted on /dev/root 60.0M 31.3M 28.7M 52% / tmpfs 40.0K 0 40.0K 0% /mnt/.psplash none 61.8M 72.0K 61.7M 0% /dev tmpfs 61.8M 232.0K 61.5M 0% /var/volatile tmpfs 61.8M 0 61.8M 0% /dev/shm tmpfs Jul 16 13:45:17 sry for the last post Jul 16 13:46:48 how can i find the info where /media/ram is in which hardware and which hardware / is? Jul 16 13:47:25 NAND or DDR-RAM or? Jul 16 13:47:46 cat /proc/cmdline and look for root= Jul 16 13:48:15 this about rootfs Jul 16 13:48:58 root@at91sam9g45ekes:~# cat /proc/cmdline mem=128M console=ttyS0,115200 mtdparts=atmel_nand:4M(bootstrap/uboot/kernel)ro,60M(rootfs),-(data) root=/dev/mtdblock1 rw rootfstype=jffs2 Jul 16 13:49:23 okay the rootfs is in the NAND and 60Mbyte large right? Jul 16 13:49:29 true Jul 16 13:49:41 thx Jay7 for your great help Jul 16 13:49:49 np :) Jul 16 13:49:53 what about tmpfs on /media/ram type tmpfs (rw,relatime) Jul 16 13:50:11 this is tmpfs (swap-backed ramfs) Jul 16 13:50:24 when you have no swap it eats RAM only afaik Jul 16 13:51:20 may be compressed iirc Jul 16 13:52:28 k -> my swap partition is realized at tmpfs... Jul 16 13:52:40 tmpfs is located in the RAM right? Jul 16 13:52:45 yes Jul 16 13:53:19 i ve 4x micron a 64 MByte Modules.... but free tells me only 126516Bytes Jul 16 13:53:29 tmps is ramfs with limit checking and swapping Jul 16 13:53:40 you can look at top output for memory info too Jul 16 13:54:15 mem=128M Jul 16 13:54:18 is it ok? Jul 16 13:54:39 (from cmdline) Jul 16 13:55:38 tmpfs 61.8M 232.0K 61.5M 0% /var/volatile what is this? Jul 16 13:56:12 why is there 3x times tmpfs? and what is the difference? Jul 16 13:56:16 you have limited your memory size by mem=128M Jul 16 13:56:22 tmpfs 61.8M 0 61.8M 0% /dev/shm Jul 16 13:56:22 hi stefan Jul 16 13:56:30 hi woglinde Jul 16 13:56:31 remove it or increase to your real memory size Jul 16 13:56:40 woglinde: you are so fast saying hello :) Jul 16 13:56:49 Sometimes I think it is a bot :) Jul 16 13:57:05 prinzenrolle: this is 3 tmpfs's for different purposes Jul 16 13:57:13 stefan hm Jul 16 13:57:32 Jay7: ive nothing limted... Jul 16 13:57:39 I'm sure they are using few bytes only Jul 16 13:57:48 128 is the default Jul 16 13:57:53 Jay7: it is so shipped from atmel Jul 16 13:58:05 [17:48] root@at91sam9g45ekes:~# cat /proc/cmdline mem=128M console=ttyS0,115200 mtdparts=atmel_nand:4M(bootstrap/uboot/kernel)ro,60M(rootfs),-(data) root=/dev/mtdblock1 rw rootfstype=jffs2 Jul 16 13:58:19 I've added tag around :) Jul 16 13:58:33 did you have u-boot? Jul 16 13:58:35 but if i increase the ram where is then tmpfs located? Jul 16 13:58:42 yes i ve uboot Jul 16 13:59:12 good morning Jul 16 13:59:30 prinzenrolle: mem is limiting memory that will be seen to overall system Jul 16 13:59:34 tmpfs is included Jul 16 13:59:34 or is the tempfs partion located in the 128MB RAM? and atmel deliver the board with this limitation without any reason Jul 16 13:59:39 is there a setting to remove the source once its compiled? Jul 16 13:59:48 so you have just not using 128Mb of your 256Mb RAM Jul 16 13:59:54 k thx a lot Jay7 Jul 16 14:00:10 IgorK : # Remove build dir after a package was built INHERIT += "rm_work" Jul 16 14:00:13 so check the way to change kernel cmdline in u-boot Jul 16 14:00:37 where is the active kernel located after uncompressing? also in ram? Jul 16 14:01:09 should be there when XIP is disabled Jul 16 14:01:16 ok, thx Jul 16 14:01:39 but XIP can't be enabled with comressed kernel iirc Jul 16 14:02:08 U-Boot 1.3.4 (Mar 9 2010 - 19:36:32) DRAM: 128 MB NAND: 256 MiB Jul 16 14:02:18 you are right Jay7 ;) Jul 16 14:02:30 hehe :) Jul 16 14:02:41 I meant compile the package then remove it, not wait until everything is compiled... Jul 16 14:03:28 btw Jul 16 14:03:41 woglinde: is bitbake from git master usable? Jul 16 14:04:07 latest rev I mean Jul 16 14:04:17 should be Jul 16 14:04:21 if not, open a bug :) Jul 16 14:04:24 * kergoth uses it every day Jul 16 14:04:30 well.. I'll test now then :) Jul 16 14:04:33 :) Jul 16 14:04:50 bitbake -i is known to be broken, though Jul 16 14:04:58 I remember.. Jul 16 14:05:00 it needs to be reimplemented after the client/server split that happened for the UIs Jul 16 14:05:12 just need a new basic console UI with the shell interface Jul 16 14:05:14 basically Jul 16 14:05:25 hm.. that is interesting task.. Jul 16 14:05:27 probably one day's work, just haven't gotten around to it, and apparently neither has RP Jul 16 14:05:37 if you know python, there's a number of existing UIs you can look at :) Jul 16 14:05:43 thx so lot.... Jay7 ... my weekend is rescued Jul 16 14:06:05 prinzenrolle: that's great :) Jul 16 14:06:11 the client/server interface uses xml/rpc in both directions, to send commands to the server, and to receive bitbake events (i.e TaskStarted) Jul 16 14:06:34 there's a helper for it, its basically just a straightforward main loop Jul 16 14:06:37 kergoth: is there any presentation or diagram? Jul 16 14:06:45 and is split happens already? Jul 16 14:06:49 hmm, i don't think RP ever wrote anything up about it Jul 16 14:06:51 yeah it has Jul 16 14:07:02 but by default its not actually used, they're internally the same process today Jul 16 14:07:14 you can enable the xml mode, but its left as is for now Jul 16 14:07:18 so bitbake now start own server then passes task to it? Jul 16 14:07:28 or there is workaround for local mode? Jul 16 14:07:51 it acts like the split is there even though technically they're in the same process Jul 16 14:08:05 from the ui perspective, it doesn't care how it happens to communicate under the hood Jul 16 14:08:22 now, this is where my brain starts to hurt Jul 16 14:08:29 :) Jul 16 14:08:39 from code in lib/bb, you can be called by the ui process, the server process, or any of the worker processes Jul 16 14:08:48 hi JaMa|GoNe Jul 16 14:08:51 and you don't know, in the api functions, which it is :) Jul 16 14:08:54 ouch everybody is gone Jul 16 14:09:07 kergoth: well.. looking good :) Jul 16 14:09:18 master is coming along, there's some nice stuff Jul 16 14:09:23 have you seen RP's bblayers stuff yet? Jul 16 14:09:26 may be I'll try to start python learning with bitbake :) Jul 16 14:09:33 its awesome not needing to use environment variables at all now Jul 16 14:09:43 jay7 hm thats maybee not so good Jul 16 14:09:46 just login to a shell, cd Code/oe; bitbake (i put master in my PATH by default) Jul 16 14:09:57 wow.. that's great about env vars Jul 16 14:10:02 I like it Jul 16 14:10:06 bitbake isnt a good example for python in some parts Jul 16 14:10:07 here, i have an example of it Jul 16 14:10:27 http://github.com/kergoth/OE-BBLayers Jul 16 14:10:39 woglinde: yeah.. but I'll prefer to learn on real examples :) Jul 16 14:10:59 anyway I have enough work for now in kexecboot.. Jul 16 14:11:09 *g* Jul 16 14:11:13 well, its a good example of what not to do! Jul 16 14:11:14 ah.. that's quilt error.. Jul 16 14:11:16 you can learn from that too Jul 16 14:11:18 :) Jul 16 14:11:22 hehe :)) Jul 16 14:11:55 well.. I should go now.. will return later and fix quilt against new patch versions (after 2.6.1). Jul 16 14:13:52 woglinde why do you know that qemu was not installed? i searched for hours why it want compile Jul 16 14:15:35 prinzenrolle aeh???? Jul 16 14:15:43 you can either let oe build qemu Jul 16 14:15:48 or use your host one Jul 16 14:16:44 or you disable NLS Jul 16 14:16:51 and dont need qemu at all Jul 16 14:18:32 woglinde: how do I disable NLS? Jul 16 14:18:38 sorry im a real noob what means NLS? Jul 16 14:19:03 geo_gcju search for USE_NLS Jul 16 14:19:05 and the difference between oe build qemu or the host i didnt understand Jul 16 14:21:18 prinzenrolle: NLS means National Language Support Jul 16 14:22:04 and, as far as I know, the locales get build with qemu, which takes some time Jul 16 14:23:48 where find i those information? its my first time that i build an image ... *sry* Jul 16 14:24:45 incomming... Jul 16 14:24:50 in my case general linux knowledge and observation of build Jul 16 14:24:57 03Denis 'GNUtoo' Carikli  07org.openembedded.dev * rd13ce77283 10openembedded.git/recipes/linux/ (linux-leviathan/defconfig linux-leviathan_git.bb): (log message trimmed) Jul 16 14:24:57 linux-leviathan: switched to a new branch and bump PR Jul 16 14:24:57 This new branch has many improvements: Jul 16 14:24:57 When we started forking the android kernel it was an experimental Jul 16 14:24:57 2.6.32 branch. Jul 16 14:25:07 bitbake -DDD is your friend Jul 16 14:27:50 ahh ive found the faq http://wiki.openembedded.net/index.php/How_to_disable_generation_of_locales Jul 16 14:28:47 i ve nothing set or configured this is the build env. from atmel... Jul 16 14:33:09 xf86-video-vesa, Failed to load module "int10" Jul 16 14:34:56 in rootfs, there is a libint10.la but no libint10.so Jul 16 14:35:40 i i think im missing an important feature Jul 16 14:35:51 oops, laggy session :) Jul 16 14:35:58 ignore me ;) Jul 16 14:37:13 kergoth: I've just added a new feature to bitbake in poky and mailed the lists about it - bbappend files Jul 16 14:37:26 * RP likes this feature a lot Jul 16 14:37:57 saw it, not sure how I feel about it. I like the simplicity of the implementation -- I'm not sure I like reliance upon the basename being unique in the entire bbfiles namespace, as that's not necessarily a requirement Jul 16 14:38:03 * kergoth will give it some thought Jul 16 14:38:56 kergoth: I think recipes will either fall into cases where this makes sense or layers will want to override them Jul 16 14:39:15 kergoth: The case where both happens would get interesting :/ Jul 16 14:40:01 hi, I can't clean and rebuild because I use autorev and fso git seem down....so Jul 16 14:40:05 I also don't know how I feel about using a new extension for the same file format, but I can't think of a better way to indicate it with the filename alone Jul 16 14:40:09 how do I get the id of an image Jul 16 14:40:21 (we already have too many .bb/.inc/etc of the same format :) Jul 16 14:40:28 because I've too many Jul 16 14:40:31 run.do_rootfs.Number Jul 16 14:40:53 kergoth: The file extension is being used to denote the role the file plays, I think that is ok Jul 16 14:41:23 I like amend.inc's greater flexibility, but it can't be sanely integrated into bitbake, since it relies upon filespath, and we don't know that until after finalize, and then we'd need a second finalize after the append... so i see why you went this route Jul 16 14:41:31 maybe i'll try actually using it and see how it feels Jul 16 14:41:31 I've no more my console output from my build Jul 16 14:42:29 kergoth: I'd try it, its solving a real world problem for me quite neatly at least Jul 16 14:42:39 (new machine in a layer Jul 16 14:43:09 it solves the same problem amend.inc does, so i know its a real problem :) its extremely useful being able to add to a recipe from a collection Jul 16 14:43:15 makes collections much, much more useful Jul 16 14:43:34 kergoth: right :) Jul 16 14:43:49 our air conditioner vent has been dripping water all damn night.. called the "emergency" maintenance number.. nothing. Jul 16 14:43:51 * kergoth mutters Jul 16 14:44:59 <_chase_> RP: what list did you send your mail to? I didn't see it on openembedded-devel but I would like to take a look at this. Jul 16 14:45:33 how long does it take to build an image with angstrom and qemu? centrion 1,4Ghz 1,5GByte RAM Jul 16 14:45:35 _chase_: Its gone to bitbake-dev, oe-devel and poky's mailing lists Jul 16 14:45:58 RP: of course, you can add all collections to FILESPATHBASE easily enough, rather than doing it in the .bbappend, if you know the collections share a directory layout. basename of the dirname of ${FILE} + each element in BBPATH Jul 16 14:46:55 * kergoth 'll set up some local collections and play a bit Jul 16 14:46:58 <_chase_> RP: Thanks. I'll try to find it. Jul 16 14:47:36 kergoth: its these assumptions about directory layout that make me nervous Jul 16 14:47:45 kergoth: e.g. poky and OE are different for a start Jul 16 14:47:46 RP: of course, using this means a greater reliance upon a version lockdown, since you're appending to specific versions. if your version shifts, you lose your append Jul 16 14:47:57 well, it depends on the use case Jul 16 14:48:10 you can't safely combine really different collections anyway, due to class changes Jul 16 14:48:18 we need to give more thought to that Jul 16 14:48:32 ideally you could pick and choose metadata from any collection from any source and have it function Jul 16 14:48:54 i've thought about having recipes prefer the classes from their own collections Jul 16 14:48:57 kergoth: this has been in the back of my mine with the layer stuff Jul 16 14:49:00 but that'd make things much more complex Jul 16 14:49:10 class versioning of some sort would help us here a bit Jul 16 14:49:10 kergoth: I didn't want it to break that objective at least Jul 16 14:49:33 kergoth: Not overlapping the class naming between collections would help Jul 16 14:49:47 hmm, thats an idea, let the layer.conf change the FILESPATHBASE Jul 16 14:49:55 kergoth: bbclassappend? Jul 16 14:49:57 * RP hides Jul 16 14:49:59 it has the structure knowledge, and that way its one less thing to set up in .bbappend Jul 16 14:50:01 ugh :) Jul 16 14:50:16 the problem is appending isn't sufficient there Jul 16 14:50:25 a change to an anonymous python function and you're hosed Jul 16 14:50:29 can't override that Jul 16 14:50:41 kergoth: time to name all functions? Jul 16 14:50:50 probably a good idea *anyway* Jul 16 14:50:52 good cleanup task Jul 16 14:50:53 kergoth: and an anonymous flag Jul 16 14:50:56 right Jul 16 14:51:06 would help bitbake internals and debugging Jul 16 14:51:09 i'd love to see that. Jul 16 14:51:13 more sane stack traces Jul 16 14:51:19 easier tracking for the varref stuff Jul 16 14:51:27 kergoth: shouldn't be too hard to add Jul 16 14:51:32 yeah, agreed Jul 16 14:51:42 backwards compatibility is easy Jul 16 14:51:49 well, wait a second Jul 16 14:51:57 why not do the events flag we've talked about before Jul 16 14:52:05 then all you need is foo[events] += "RecipeParsed" Jul 16 14:52:07 and you have anonymous Jul 16 14:52:14 no need for the "anonymous" concept at all Jul 16 14:52:16 yes, thats even better Jul 16 14:52:22 hi woglinde Jul 16 14:52:24 more explicit Jul 16 14:52:27 less confusing, hopefully Jul 16 14:52:30 kergoth: agreed Jul 16 14:52:50 adding events flag support is trivial, i had it on a branch somewhere once Jul 16 14:53:02 woglinde, mrmoku, found that RDEPENDS_${PN} = "initscripts console-tools" in keymaps_1.0.bb Jul 16 14:53:06 * RP is totally in favour Jul 16 14:53:12 SHR uses initscripts-shr Jul 16 14:53:30 basically it's highly problematic Jul 16 14:53:45 initscript get installed instead of initscripts-shr in shr-image Jul 16 14:53:48 then Jul 16 14:53:58 there is /etc/init.d/devices Jul 16 14:54:14 which is a devfs script according to mickeyl Jul 16 14:54:18 which result in: Jul 16 14:54:22 *no gsm Jul 16 14:54:25 *no sound Jul 16 14:54:30 because of lacking dev nodes Jul 16 14:55:36 RP: okay, the .bbappend thing is growing on me. always a knee jerk reaction to taking away flexibility (relative to amend.inc in this case), but I think we don't really need that flexibility. I'll play with it more today. Jul 16 14:56:09 * kergoth mutters, forgot to make a janitors wiki page to document cleanup tasks, need to do that Jul 16 14:56:36 kergoth: Its always a balancing act between core features in bitbake and the power to do stuff in classes Jul 16 14:56:42 yep Jul 16 14:57:08 it's nice to keep a lot of flexibility, but there are performance implications, among others Jul 16 14:57:09 I love the simplicty of this but its always a question of weighing up the pros and cons Jul 16 14:57:13 * kergoth nods Jul 16 14:57:28 performance hit should be minimal for this Jul 16 14:58:08 yeah, a fair bit less noticeable than amend.inc is, which is a Good Thing Jul 16 15:00:13 trying to figure out how to leverage this variable reference tracking in bitbake internally is breaking my brain. the challenge is the case where you need to re-parse a file in the middle of the stack of datasmart objects Jul 16 15:01:02 * hrw finally bootstrapped Ubuntu cross compiler Jul 16 15:01:09 ~hail OpenEmbedded Jul 16 15:01:10 * ibot bows down to OpenEmbedded and chants, "I'M NOT WORTHY!!" Jul 16 15:01:13 kergoth: we do that? Jul 16 15:01:27 RP: not today, i'm thinking about how we can use this to avoid reparsing the world. Jul 16 15:01:27 kergoth: I thought if something had already been included we skipped it? Jul 16 15:01:41 kergoth: ah, right Jul 16 15:01:51 * RP still has checksuming to play with Jul 16 15:02:02 well, each file is sort of its own entity, each include builds upon the previous data Jul 16 15:02:09 * RP wanted to make task based packaged staging work first though Jul 16 15:02:11 we need variable versioning, like they do in some compilers Jul 16 15:02:36 kergoth: right. We're thought about that before Jul 16 15:02:48 so like, at the point where this include comes in, we know FOO is version 3, and the next guy down the line depends on 5, since this file modifies it twice Jul 16 15:02:49 * kergoth nods Jul 16 15:02:55 I need a whiteboard! Jul 16 15:03:05 * kergoth doesn't have one at home yet :) Jul 16 15:03:42 kergoth: We should convince our employers to lock us in a conference room for a week ;-) Jul 16 15:03:56 * RP dreads to think what we'd do to bitbake :) Jul 16 15:04:09 ooh, that'd be nice. Jul 16 15:04:26 we need to think about splitting up some of our classes, i think Jul 16 15:04:33 we have some truly massive ones Jul 16 15:04:34 kergoth: btw, are you able to make ELC-E in October? Jul 16 15:04:41 with interdependencies back and forth between bb modules in both directions Jul 16 15:04:45 it hurts my brain Jul 16 15:05:01 I might be able to, not sure yet Jul 16 15:07:40 wtf, http://englishrussia.com/index.php/2010/02/17/russian-gadget-hackwrench-religion/ Jul 16 15:09:53 kergoth: http://media.englishrussia.com/gadget_fan_club_from_russia/6.jpg Jul 16 15:09:54 lol Jul 16 15:09:55 ok Jul 16 15:10:06 lol Jul 16 15:10:46 kergoth: I beg thats a joke religion Jul 16 15:10:50 me too Jul 16 15:10:52 like the discordianism Jul 16 15:10:55 yeah Jul 16 15:24:35 anyway Jul 16 15:24:53 GNUtoo|laptop: hope you will find the solution for the usb problem Jul 16 15:25:04 yes later Jul 16 15:25:09 I'm busy with making a release Jul 16 15:25:12 GNUtoo|laptop: ok Jul 16 15:25:15 SHR? Jul 16 15:25:18 waiting endlessly for that is not a good idea Jul 16 15:25:22 kergoth: he has freebsd logo on tshirt Jul 16 15:25:25 after the phone is end of life Jul 16 15:25:33 so basic functionality is working Jul 16 15:25:39 so release Jul 16 15:25:42 k Jul 16 15:25:57 where do you upload the image? Jul 16 15:30:31 leviathan: guess we should build them on the SHR buildhost Jul 16 15:47:22 cya Jul 16 15:48:44 is INHERIT+="rm_work" broken in the dev tree? Jul 16 15:49:15 doesn't look broken here Jul 16 15:49:24 works fine for me Jul 16 15:50:27 hmmm, it was working on 1.8.18 with the stable tree on comp A; isnt working on 1.8.18 and 1.11.0 on the dev tree on comp A and comp B o.0 Jul 16 15:51:19 * JaMa using it with 1.10 and bitbake master Jul 16 15:51:43 i dont think the toolchain is finished building yet, so that may be the reason Jul 16 15:52:14 but if thats the case, that means it takes over 15gb to build the toolchain which shouldnt be the case o.0 Jul 16 15:52:31 bitbake master? Jul 16 15:55:26 i cant find anything about bitbake master o.0 Jul 16 15:56:56 and can i do a bitbake -c rm_work after canceling a build, then resume the build after? Jul 16 15:57:37 warflyr no Jul 16 15:57:49 rm_work removes the source Jul 16 15:57:51 and all Jul 16 16:00:31 hrmm Jul 16 16:00:37 well hopefully 30gb will get me through this o.0 Jul 16 16:00:51 it didnt on the stable build o.0 Jul 16 16:01:40 depends Jul 16 16:03:28 warflyr: you can resume a build fine with or without rm_work, at a recipe level. Jul 16 16:03:40 warflyr: the workdir isn't removed until after the build of a given recipe is mostly complete Jul 16 16:04:12 do_rm_work is executed after do_build Jul 16 16:05:19 INHERIT+="rm_work" in local.conf isnt working for me... is there somewhere else i can also put it? Jul 16 16:05:23 maybe in a recipe? Jul 16 16:05:58 ???? Jul 16 16:06:06 no Jul 16 16:06:11 its working Jul 16 16:06:22 but the tasks may kick in late Jul 16 16:06:32 rgr Jul 16 16:06:46 try something simple first Jul 16 16:06:51 hopefully 30gb will get me by until it does start working Jul 16 16:06:58 the toolchain isnt even built yet Jul 16 16:06:59 like binutils-cross Jul 16 16:07:25 oh its on do_compile bitutils-cross right now Jul 16 16:08:13 do_install... Jul 16 16:09:47 last thing it did was task do_package_stage: Succeeded with no rm_work before Jul 16 16:10:21 last should be do_build Jul 16 16:10:32 when rm_work isnt used Jul 16 16:10:39 hmmm Jul 16 16:11:45 no do_builds in the past 10 packages or so Jul 16 16:12:04 cancel it Jul 16 16:12:12 and do bitbake binutils-cross Jul 16 16:12:17 than you know for sure Jul 16 16:29:43 why do most packages need a native and the cross? Jul 16 16:31:03 most? Jul 16 16:31:18 all? Jul 16 16:31:23 because some packages needs its native package to generate some files Jul 16 16:31:29 ah Jul 16 16:31:59 i figured qemu was doing that on the crossd packages o.0 Jul 16 16:40:22 re Jul 16 16:40:52 wb Jul 16 16:45:31 from where can i have a free private repository on internet? Jul 16 16:45:51 false channel Jul 16 16:46:00 hm but look at github Jul 16 16:47:15 woglinde: thank you Jul 16 16:47:39 woglinde, I've a bug question but you're not in #buglabs Jul 16 16:47:44 you had navit on bug 1.x Jul 16 16:47:47 oh mom Jul 16 16:47:54 I've it too Jul 16 16:47:57 gnutoo I only started it once Jul 16 16:48:02 ohh ok Jul 16 16:48:14 no time for it now Jul 16 16:48:16 I wondered if you succedded calibrating the touchscreen Jul 16 16:48:22 oh Jul 16 16:48:25 hm Jul 16 16:48:30 didn't need it Jul 16 16:48:33 because for using navit the touchscreen must be calibrated well enough Jul 16 16:48:33 worked fine Jul 16 16:48:35 ok Jul 16 16:48:39 hm oh Jul 16 16:48:42 jo effem Jul 16 16:48:47 even the p or l key on the software keyboard Jul 16 16:49:10 gnutoo can you make a picture from the keyboard? Jul 16 16:49:18 yes Jul 16 16:49:21 maybee we can make the font a little bit bigger Jul 16 16:49:37 no Jul 16 16:49:43 it's the touchscreen calibration Jul 16 16:49:47 when I click on p Jul 16 16:49:52 I've I Jul 16 16:49:55 which is more left Jul 16 16:49:57 etc... Jul 16 16:50:19 but that happends only at the middle part of the rightest part of the screen Jul 16 16:56:04 hi woglinde, all Jul 16 16:56:16 only a short time here this evening Jul 16 16:57:16 what are you doing? Jul 16 16:57:27 instead hanging here around Jul 16 16:58:20 03Frans Meulenbroeks  07org.openembedded.dev * rdd7a49ac11 10openembedded.git/recipes/glibc/ (55 files in 6 dirs): (log message trimmed) Jul 16 16:58:21 glibc: remove old/obsolete versions Jul 16 16:58:21 as mentioned before on the OE ML. Jul 16 16:58:21 removing some old versions which are not pinned by any distro Jul 16 16:58:21 and are fairly old: Jul 16 16:58:27 03Frans Meulenbroeks  07org.openembedded.dev * r5fa7c51fc7 10openembedded.git/recipes/glibc/ (22 files in 3 dirs): Jul 16 16:58:27 glibc 2.10.1: cleaned up; patches only in local dir Jul 16 16:58:27 moved all patches to glibc-2.10.1 Jul 16 16:58:27 removed references to other dirs. Jul 16 16:58:27 Signed-off-by: Frans Meulenbroeks Jul 16 16:58:45 woglinde: friday evening is for sports, (indoor soccer, and afterwards some beers with the guys) Jul 16 16:59:19 ah cool Jul 16 16:59:22 indoor? Jul 16 16:59:32 with this nice weather Jul 16 17:00:00 its so hot out in CO, US today o.0 Jul 16 17:00:18 here too in germany Jul 16 17:00:22 but in evening its nice Jul 16 17:00:34 woglinde btw on binutils, once gcc-cross finished building, its starting to rm_work; thanks ;) Jul 16 17:00:45 ;) Jul 16 17:00:47 as I said Jul 16 17:00:54 rm_work will kick in later Jul 16 17:01:08 kergoth ahs some plan to make the tasks better Jul 16 17:01:13 woglinde: actuallywe play throughout the year with only a 6 wk summer break Jul 16 17:01:23 even let them depend on your disk io Jul 16 17:01:29 disk-io is now the bottleneck Jul 16 17:02:05 yeah i noticed that... hm we need a guard for the java stuff Jul 16 17:02:16 which bails when one class has size 0 Jul 16 17:02:30 but my linux servers dont have enough HD space to get through it o.0 Jul 16 17:03:07 *g* delete something Jul 16 17:03:29 how would i use amend.inc to add some peers to the ppp recipe? Jul 16 17:04:08 03Tom Rini  07org.openembedded.dev * r29e63e74ab 10openembedded.git/recipes/lmbench/ (lmbench_2.0.4.bb lmbench_2.5.bb): Jul 16 17:04:08 lmbench: Fix builds by pulling in siteinfo stuff Jul 16 17:04:08 At some point an automatic import of autotools.bbclass was removed. Jul 16 17:04:08 Since we only need CONFIG_SITEINFO, inherit siteinfo and set the variable Jul 16 17:04:08 so we're back to where we were. Jul 16 17:04:27 lol i went through tons of stuff and moved it to a NFS mount; i guess i only needed another 2GB to get through the part til it would start rm_work o.0 Jul 16 17:05:10 original plan was to build on the NFS mount, but that didnt work due to glibc needing to chown some things Jul 16 17:05:28 ieehks Jul 16 17:05:32 dont build on nfs Jul 16 17:05:41 yeah that was a complete waste of a day... Jul 16 17:15:27 woglinde, do you have a 1.4.3 bug off hand ? Jul 16 17:15:39 if so could you run cat /etc/pointercal ? Jul 16 17:15:52 on monday Jul 16 17:16:15 ah ok Jul 16 17:16:26 btw buglabs is making a bug 2.0 device Jul 16 17:16:36 maybe it's fast enough to run openoffice.... Jul 16 17:16:53 do you still have the recipe or is it deleted? Jul 16 17:17:11 *g* Jul 16 17:17:15 lol Jul 16 17:17:20 I still have what I begun Jul 16 17:17:25 let me check on monday Jul 16 17:17:27 yes I imagine Jul 16 17:17:32 but it would save me some work Jul 16 17:17:46 I will do that only when bug 2.0 will be available I think Jul 16 17:18:15 so I'm not in a hurry Jul 16 17:18:21 but I'd like to know it still exist Jul 16 17:43:54 is it possible to have one recipe that pulls from two different git repositories? Jul 16 17:47:34 why wouldn't it be? Jul 16 17:47:37 SRC_URI is just a list of urls Jul 16 17:48:36 it won't try to put both in a directory named git? Jul 16 17:48:55 there's a url parameter to control where it goes Jul 16 17:49:51 what is it? i thought there was just tag and protocol. Jul 16 17:51:33 fetching happens in a unique area based on the url Jul 16 17:51:45 * kergoth_ double checks Jul 16 17:52:24 there's a general unpack parameter "subdir" Jul 16 17:52:40 it'll "unpack" the git dir inside of there, relative to workdir Jul 16 17:52:50 subdir=foo -> ${WORKDIR}/foo/git, as far as i can see Jul 16 17:53:01 not ideal, but should get hte job done Jul 16 17:53:16 awesome! thank you. :) Jul 16 17:55:03 kergoth_: is re the best way to search in a string Jul 16 17:56:23 I am using string.find Jul 16 17:56:31 depends on what you're trying to do Jul 16 17:56:49 gchiii, np. and for reference, that was in the base.bbclass code, oe_unpack_file :) Jul 16 17:56:53 I have to get architecture from output of file Jul 16 18:00:33 khem, i'd use re then Jul 16 18:03:48 hmm a is b compares object types ? Jul 16 18:04:05 and a = b compares values Jul 16 18:04:14 they dont seem same to me Jul 16 18:04:49 == compares by value, "is" compares by object *identity*, not type Jul 16 18:04:56 as in, the exact same object in memory Jul 16 18:05:11 if you want to check types, use isinstance(), though most cases you can avoid it Jul 16 18:05:54 identity means same object ? like pointers in C Jul 16 18:06:06 I mean is Jul 16 18:06:46 yeah Jul 16 18:06:49 kergoth: iow whats object identity for a layman Jul 16 18:06:51 basically compare by memory address Jul 16 18:07:06 ok for a C guy thats pointer :) Jul 16 18:07:18 its a common enough notion, identity comparison vs value comparison. many languages deal with it Jul 16 18:07:35 yes Jul 16 18:07:51 its named differently Jul 16 18:07:54 jo khem Jul 16 18:07:59 C++ calls references Jul 16 18:10:12 is there a way to check dependancies on recipes? i copied a console image over and made a few additions to it, but now bitbake is building X software... Jul 16 18:12:02 i added: nmap u-boot-utils netbase modutils-initscripts libssh automake perl gcc g++ python autoconf make screen m4 sqlite sqlite3 curl lighttpd php wget; though i dont believe any of those require X Jul 16 18:12:42 warflyr whats build and ended up in the images is diffrent Jul 16 18:12:53 check bitbake -g foobaar Jul 16 18:13:01 you get 3 .dot files Jul 16 18:13:13 its easy to understand Jul 16 18:14:47 rgr Jul 16 18:32:52 haha.. hackwrench religion Jul 16 18:45:04 kergoth: is there something like default match in dict ? Jul 16 18:45:17 think of something like default: in case statements in C Jul 16 18:57:53 well.. about quilt and newest gnu patch (after 2.6.1) Jul 16 18:58:43 should I provide patch for 0.48 from quilt git or we will just upgrade to git version of that revision Jul 16 19:03:04 khem, either use collections.defaultdict() or the setdefault() method of dict Jul 16 19:05:21 defaultdict is cool Jul 16 19:06:01 kergoth is it easy to teach bitbake to use only 2 unpack or 2 rm_work at the same time? Jul 16 19:07:19 kergoth, did you see my question about bb.msg function's fn arg? Jul 16 19:07:33 blindvt: no idea on that one Jul 16 19:07:39 woglinde, no, you'd have to revamp the scheduler Jul 16 19:07:52 hms okay Jul 16 19:07:54 mickeyl: hi Jul 16 19:08:00 hi leviathan Jul 16 19:08:03 each tasks should have a queue Jul 16 19:08:03 mickeyl: how far is gnutoo with his release? Jul 16 19:08:18 leviathan: no idea, last i heard he had some build problems Jul 16 19:08:28 mickeyl: hmm, okey Jul 16 19:08:37 gnutoo is here Jul 16 19:08:39 or was Jul 16 19:08:40 kergoth, it sounds like it ment to hold the caller functionname, but it's pretty dangerous/unpleasant to just ignore that arg (as can be seen in filter_environment()), isn't it Jul 16 19:08:47 woglinde: I know Jul 16 19:08:48 it seems pretty stupid, anyway Jul 16 19:08:54 woglinde: he has connection problems Jul 16 19:08:55 if you *really* need the caller, you can inspect the stack Jul 16 19:09:02 indeed Jul 16 19:09:18 but i doubt it really needs it anyway, and very few spots in the code actually usei t Jul 16 19:09:49 woglinde: +1 to your request about queuing. There will be 3 queues - downloading, checksumming/unpacking/rm_work, building Jul 16 19:09:49 kergoth, it provokes errors, serves no purpose i can see and should thus be removed altogether, shouldn't it Jul 16 19:10:07 each queue can have own max number of threads Jul 16 19:10:47 each queue is creating own load (internet, disk io, cpu) Jul 16 19:14:21 jay7 seems that will your first task Jul 16 19:14:26 when working on bitbake Jul 16 19:14:27 hehe Jul 16 19:14:32 * Jay7 hides Jul 16 19:15:12 I can still see you Jul 16 19:15:37 * Jay7 hides better Jul 16 19:43:42 I wonder how can you hide *better* Jul 16 19:45:38 re khem Jul 16 19:45:58 hey woglinde Jul 16 19:46:25 uclibc on mipsel doesnt boot :( Jul 16 19:46:37 but mips does Jul 16 19:46:46 so something in mipsel Jul 16 19:48:09 hm uh Jul 16 19:48:11 or qemu Jul 16 19:50:56 check this diff please: http://pastebin.ca/1901970 Jul 16 19:51:19 I've applied changes only for quilt-native_0.48.bb Jul 16 19:54:39 Jay7: you need to patch configure.ac Jul 16 19:54:44 not configure Jul 16 19:54:50 khem: it does not working Jul 16 19:54:58 quilt have bundled configure Jul 16 19:55:11 yes then regenerate configure Jul 16 19:55:26 and add both to patch Jul 16 19:55:38 hm.. Jul 16 19:55:52 just patching configure is always errorprone Jul 16 19:56:06 I understand that Jul 16 19:57:01 so should I patch configure.ac then regenerate configure and place into tree diff between original files and current? Jul 16 19:57:52 quilt have diff already: http://git.savannah.gnu.org/cgit/quilt.git/commit/?id=2f0eb85bf8c0e1dbc4b73b294ed1eb478f8b7004 Jul 16 19:58:03 but there is more changes Jul 16 20:00:07 btw, can I force configure regeneration from recipe? Jul 16 20:00:24 ? Jul 16 20:00:36 regenerate configure from configure.ac Jul 16 20:00:36 autotools.bbclass is doing it Jul 16 20:01:04 * woglinde sometimes wonders Jul 16 20:01:14 you arent a noob jay7 Jul 16 20:01:20 I wonders too.. Jul 16 20:01:52 I've done first patch against configure.ac but that was changed nothing in configure Jul 16 20:02:15 quilt have both configure and configure.ac Jul 16 20:02:30 patching configure is a bad idea Jul 16 20:02:31 will bitbake regenerate configure from configure.ac in this case? Jul 16 20:02:40 Makefile.in too Jul 16 20:02:49 inherit autotools Jul 16 20:02:53 then seems I've missed something.. Jul 16 20:02:57 will recheck Jul 16 20:02:57 examine do_configure look Jul 16 20:03:09 bitbake isn't what regenerates configure, autotools.bbclass is, as woglinde says Jul 16 20:03:57 ups Jul 16 20:04:01 s/look/log/ Jul 16 20:06:42 freetype-2.4 anyone? Jul 16 20:18:06 hey someone in Spain wants to buy that octopus :) Jul 16 20:18:26 some wants him to be honorary guest in sea food festival Jul 16 20:19:11 Already a celebrity in Spain after the semifinal prediction - Prime Minister Jose Luis Rodriguez Zapatero offered to send a security team to protect the 2 1/2-year-old floppy octopus from the hungry Germans Jul 16 20:42:54 where is ${DEBIAN_MIRROR} defined at? (i saw it earlier today looking through files but now i cant find it...) Jul 16 20:43:15 khem yes yes Jul 16 20:43:31 but this species lives only 3 years max Jul 16 20:46:18 woglinde: next year it will be dead Jul 16 20:49:06 yes Jul 16 20:49:49 heh someone is offerring 30K euro Jul 16 20:50:06 and german authorities are sayinh np Jul 16 20:50:08 no Jul 16 20:51:30 jupp thats us Jul 16 20:54:49 sell the idiot and take 30k Jul 16 20:55:04 anyway he did not help germany in any way :) with his predictions Jul 16 20:55:08 traitor Jul 16 20:55:18 hm for bets Jul 16 20:55:28 he said all winners Jul 16 20:55:52 * khem thinks this indian tea is super Jul 16 21:04:21 well Jul 16 21:04:39 configure is not regenerated by bitbake -c configure quilt-native Jul 16 21:05:02 autotools is inherited Jul 16 21:05:46 hm ah right Jul 16 21:05:58 because we dont have autotools as this stage Jul 16 21:06:08 quilt-native is one of the firstpackages build Jul 16 21:06:09 hm hmh Jul 16 21:08:28 I see no good way then.. Jul 16 21:08:39 only have to patch both configure and configure.ac Jul 16 21:08:50 or call host's autoconf Jul 16 21:09:34 yeah sorry Jul 16 21:09:40 didnt thought about it Jul 16 21:09:50 so patching configure in this case is okay Jul 16 21:09:57 np, but now we should decide how to do further Jul 16 21:10:21 khem: any opinions? Jul 16 21:10:38 ant__: hello Jul 16 21:10:40 Jay7: patch both configure and configure.ac Jul 16 21:10:47 hi Jul 16 21:10:58 khem: ok Jul 16 21:11:39 jo re ant Jul 16 21:13:48 pb_, ping Jul 16 21:14:24 re florian Jul 16 21:15:54 re Jul 16 21:16:43 pb_, using PKGV instead of munging PR doesn't seem to work for me.. i say PKGV = "4.6+svnr${SRCPV}" but this has no (visible?) effect since i end up with gcc-cross-svn-r0 no matter what. Ideas? Jul 16 21:17:30 you know PKGV only affects the binary packages, not the rev you'll see in the build messages, right? ;) Jul 16 21:18:31 * blindvt shakes his head Jul 16 21:18:40 wow, i'm an idiot Jul 16 21:18:43 hehe Jul 16 21:18:48 "PKG" :) Jul 16 21:19:10 kergoth: is there some run down on srctree stuff Jul 16 21:19:37 i don't think so. i can write something up if you like. if it wasn't for variants (native, cross, etc) the class would just set S = ${FILE_DIRNAME} and call it done Jul 16 21:19:54 there's a danger of task interleaving within a shared ${S} if you have native/cross of the recipe too Jul 16 21:19:55 that wud be cool Jul 16 21:20:07 so it kind of munges the intermediate build tasks together into a single operation Jul 16 21:20:14 its a hack, and probably needs some TLC Jul 16 21:20:21 khem, may i send you bumps for gmp/mpfr/mpc for you to test? Jul 16 21:20:32 blindvt: sure Jul 16 21:20:35 of course, if you use separate bin vs obj dir, or don't have native/target, its enough to just set S = ${FILE_DIRNAME} and leave SRC_URI empty Jul 16 21:20:40 s/or don't/and don't/ Jul 16 21:20:47 er, no, i was right the first time Jul 16 21:21:01 * khem will try to get gcc-runtime work for poky into OE over weekend Jul 16 21:21:26 khem hm what will it usefull for? Jul 16 21:21:27 planning on starting some OE blog posts, i could use the writing practice Jul 16 21:21:28 heh Jul 16 21:21:39 kergoth, and how would i trick oe into spitting out gcc-cross-svn-r4711 for rev 4711 (with PV kept at e.g. "svn")? Jul 16 21:21:56 not sure what you mean by that Jul 16 21:22:00 woglinde: it will separate out gcc target libraries from gcc-cross Jul 16 21:22:02 and gcc Jul 16 21:22:02 what do you mean by "spitting out"? Jul 16 21:22:38 woglinde: so you gcc-cross recipes wont generate any packages thats target specific Jul 16 21:22:42 kergoth, using that as pr, i think. Jul 16 21:22:53 khem eh? Jul 16 21:23:08 well, you'd have to mangle PR for that, but I don't really see why you'd need to :) Jul 16 21:23:28 kergoth, nm, i'll ask pb on the list. He was asking me not to munge pr Jul 16 21:24:03 I guess I could see wanting it there if you wanted to ensure that changing the SRCREV resulted in rebuilding it Jul 16 21:25:23 woglinde: after patching dietlibc it builds. kexec-tools-dietlibc fails do_configure, though... (compiler cannot crate executables) Jul 16 21:25:25 new version of quilt changes: http://pastebin.ca/1902045 Jul 16 21:25:25 that would be the plan, yes (it makes sense to rebuild if SRCREV was changed, but maybe that's just me ;) Jul 16 21:25:36 woglinde, khem: have look please Jul 16 21:26:19 * kergoth nods Jul 16 21:26:47 i think it makes a certain amount of sense to have the rev in the PR, but the auto incrementing crap for autorev definitely belongs in PKGR, i'd think, heh Jul 16 21:26:55 woglinde: some magic in site? Jul 16 21:26:57 * kergoth shrugs, pb_ knows this stuff better than he does Jul 16 21:27:54 Jay7: looks ok also bump PR and mention about distro which shows this issue up Jul 16 21:27:59 in commit log Jul 16 21:28:17 khem: I see no need to bump PR Jul 16 21:28:31 why Jul 16 21:28:34 people who can build quilt have no such error Jul 16 21:28:48 but you might introduce one for them Jul 16 21:28:55 people who have newer GNU patch - can't build quilt :) Jul 16 21:29:02 so if you bump PR they will rebuild it and report back sooner than later Jul 16 21:29:08 ah, this point Jul 16 21:29:11 well, ok Jul 16 21:29:28 I dont see this problem on f12 Jul 16 21:29:40 it uses newest GNU patch Jul 16 21:29:51 may be they patced it to behave old way ? Jul 16 21:29:59 khem: ALT Linux have GNU patch that is build from git Jul 16 21:30:09 I'm using testing branch of ALT Jul 16 21:30:27 oh boy living on doubly dangerous ends Jul 16 21:30:31 ant hm do have the logs somewhere? Jul 16 21:30:38 building OE unstable on an unstable distro Jul 16 21:30:46 I can show you GNU patch revision that changes this Jul 16 21:31:09 well you know what you are upto I always use ubuntu dev version Jul 16 21:31:26 khem: hehe.. so I'm always have some new errors ;) Jul 16 21:31:27 Jay7: ok I believe you Jul 16 21:31:48 hmm, maybe i should do a blog post on a few of the lesser known bbclasses, maybe the ones i did and pushed at MV Jul 16 21:32:04 kergoth: yes that wud be nice Jul 16 21:32:12 we all will link that in our blogs Jul 16 21:32:43 classes/ is a mess right now Jul 16 21:32:50 so I'll bump PR to r3, note distro in commit message and commit then Jul 16 21:32:50 its unclear which are optional via INHERIT Jul 16 21:32:53 which are 'inherit'able Jul 16 21:32:53 kergoth: yes I feel so Jul 16 21:32:56 etc Jul 16 21:33:03 kergoth: where is you siteinfo work Jul 16 21:33:24 i think thats on github, i still need to check sanity, since i saw that one build failure from it Jul 16 21:33:35 need to just bitbake -e and compare the CONFIG_SITE i think Jul 16 21:33:47 hrm Jul 16 21:33:51 kergoth: may be write this on wiki? Jul 16 21:33:53 maybe its not there.. damnit Jul 16 21:33:55 i have too many trees Jul 16 21:34:57 okay, *now* its on github :P Jul 16 21:35:14 http://github.com/kergoth/OpenEmbedded/commit/a8c7a468260822fead10c242ce17e1a1a70b8bf5 Jul 16 21:37:33 I'll go sleep Jul 16 21:37:39 gn8 Jul 16 21:37:41 woglinde, logs coming..rebuilding gcc-cross Jul 16 21:38:13 Jay7, probably a good idea, start a page at least documenting the purpose and mode of operation of each class.. Jul 16 21:38:15 * kergoth adds to todo Jul 16 21:38:51 dokuwiki is easy to setup and require no SQL storage :) Jul 16 21:38:54 JFYI :) Jul 16 21:42:09 * khem debugs portmap to have some fun Jul 16 21:42:15 03Yuri Bushmelev  07org.openembedded.dev * re05f5e4e8f 10openembedded.git/recipes/quilt/ (2 files in 2 dirs): Jul 16 21:42:15 quilt-native_0.48: Patch to properly detect newer versions of GNU patch is added. Jul 16 21:42:15 New versions of GNU patch present themselves as "GNU patch" instead Jul 16 21:42:15 of just "patch". Jul 16 21:42:15 NOTE1: both configure.ac and configure are patched because we have Jul 16 21:42:52 I've run into a strange problem, bitbake is failing to build eggdbus claiming it has a mis-match on md5 and sha256 yet when I download the tar balls directly and run md5sums and sha256 the values match what the recipe is looking for. What should I check next? Jul 16 21:43:06 checksums.ini Jul 16 21:43:14 checksums.ini is out Jul 16 21:43:23 only stable is using it Jul 16 21:43:25 ah Jul 16 21:44:15 ill probably be getting a fail then on my custom packages soon if a checksum is required in the recipe o.0 Jul 16 21:44:29 Andy_ODT: if you download it dirrectly are the balls the same as what you have in downloads dir? Jul 16 21:45:06 I'm not certain where the downloads are at that the script gets... my source dir? Jul 16 21:46:03 no the version I downloaded has different sum than the one the script seems to have downloaded Jul 16 21:46:28 i think the stable repository seems more stable o.0 Jul 16 21:46:59 Andy_ODT add the new checksums in the recipe and off oyu go Jul 16 21:47:24 woglinde: alright, thx Jul 16 21:47:28 sometimes upstream changes without Jul 16 21:47:31 noticing Jul 16 21:47:34 or they got hacked Jul 16 21:47:39 cannt decide it Jul 16 21:50:11 Andy_ODT: if I understand you right.. just downloads balls + ball.md5 from your download dir to force redownload Jul 16 21:50:46 maybe it was only partially downloaded and updating checksums will just allow broken download to unpack.. Jul 16 21:51:07 jama maybee Jul 16 21:51:46 JaMa: well I thought that I must have just had a hickup in my network, but I cleaned the recipe... re ran it and got the exact same download from the eggdbus repo Jul 16 21:52:25 * Andy_ODT shrugs Jul 16 21:52:29 cleaned recipe? it doesn't remove source from downloads Jul 16 21:52:55 ah I thought it would, I'll give that a go then Jul 16 21:53:10 it would be marvelous if we could wipe all versions of m4 but the latest (and 1.4.8 if that's really relevant for om-2008, how relevant that one itself may be by now) Jul 16 21:53:40 somone knows why we use libffi_2.0+gcc4.3.2.bb for native libffi? Jul 16 21:54:07 blindvt: nobody touched om-2008 for logn.. I guess that m4 version would be one of easier issues with om-2008 now :) Jul 16 21:54:38 JaMa, there's om-2009 so i wonder why that 2008 is still in there in 2010 ? Jul 16 21:58:52 the same apply to om-2009.. Jul 16 22:01:20 woglinde, checking for arm-angstrom-linux-gnueabi-gcc... diet arm-angstrom-linux-gnueabi-g Jul 16 22:01:22 cc -march=armv5te -mtune=xscale -mthumb-interwork -mno-thumb Jul 16 22:01:24 checking whether the C compiler works... no Jul 16 22:01:28 http://pastebin.ca/1902059 Jul 16 22:01:42 config.log ^^^ Jul 16 22:03:06 CC='diet arm-angstrom-linux-gnueabi-gcc -march=armv5te -mtune=xscale -mthumb-interwork -mno-thumb' Jul 16 22:03:33 armv5te-angstrom-linux-gnueabi/kexec-tools-dietlibc-2.0.1-r0/kexec-tools-2.0.1/configure: line 2543: 16024 Segmentation fault $CC --version 1>&5 Jul 16 22:04:46 ant config.log? Jul 16 22:04:49 ant__, something's a bit broken, from the looks Jul 16 22:05:10 blindvt it was broken anyway but worked somehow Jul 16 22:05:49 hehe.. I have error with libbackend.a while compiling gcc-cross-initial-4.3.3 Jul 16 22:06:05 * blindvt goes to play in his own corner Jul 16 22:06:34 khem, i've only slightly tested these 3 bumps, so please be careful Jul 16 22:06:59 but you are anyway Jul 16 22:09:01 where did i leave? right, i'm miscompiling something in libcpp, fun Jul 16 22:09:03 * blindvt & Jul 16 22:10:01 woglinde, works if I run configure by hand... Jul 16 22:10:07 blindtv: ^^ Jul 16 22:10:20 checking for gcc... gcc Jul 16 22:10:20 checking whether the C compiler works... yes Jul 16 22:10:20 checking for C compiler default output file name... a.out Jul 16 22:13:32 ant eh? Jul 16 22:14:02 yes, going in workdir and running configure Jul 16 22:14:12 I have no erors Jul 16 22:14:22 sure you are using host gcc Jul 16 22:14:26 checking for gcc... gcc Jul 16 22:14:31 omg... Jul 16 22:14:33 non diet Jul 16 22:14:36 sure :/ Jul 16 22:14:44 what did the config.log says? Jul 16 22:14:50 now you overwritten it Jul 16 22:15:00 +have Jul 16 22:15:01 see pastebin Jul 16 22:15:20 I am blind= Jul 16 22:15:21 ? Jul 16 22:15:30 which pastebin Jul 16 22:19:21 ant hm my patch is okay btw. Jul 16 22:19:24 after looking at it Jul 16 22:19:38 DIETHOME=$(PWD) bin-$(MYARCH)/diet $(CROSS)$(CC) $(CFLAGS) -o $@ contrib/dnsd.c Jul 16 22:19:47 DIETHOME=$(PWD) sets DIETHOME Jul 16 22:20:12 but maybee my ccache patch isnt the best Jul 16 22:23:21 oh, and if somebody has the nerves, here's a bison bump that fails to autoreconf for me due to some m4_include(some_variable_that_points_to_a_file_that_in_fact_does_exist) error that m4 wouldn't be able to *open* the *variable* : busybox.net/~aldot/oe/recipes_bison.tar Jul 16 22:23:49 ae? Jul 16 22:24:58 m4_define(foo,/tmp/bar.txt) m4_include(foo) complains that the file `foo´ does not exist. You'll see what i mean Jul 16 22:25:47 m4_define is evil anyway Jul 16 22:25:48 *g* Jul 16 22:27:39 woglinde: ok, I see Jul 16 22:27:54 thenm we are back to Jul 16 22:27:55 | make: *** [bin-i386/elftrunc] Segmentation fault Jul 16 22:28:05 | make: *** [bin-i386/dnsd] Segmentation fault Jul 16 22:28:22 compiling dietlibc today Jul 16 22:28:24 as I said Jul 16 22:28:35 try to compile with ccache Jul 16 22:28:38 if it works Jul 16 22:28:49 you will know my ccache patch is borken Jul 16 22:28:53 hm..I don't have ccache Jul 16 22:29:01 install it Jul 16 22:29:51 can try, but surely wasn't installed when long ago I built dietlibc. Jul 16 23:09:21 ah finally sorted out why this thumb overrides did not work Jul 16 23:09:39 hmmm super now those mysterious segfaults are gone too Jul 16 23:09:42 sweet Jul 16 23:09:57 folks expect a radical RFC :) Jul 16 23:10:28 in any case I am making something to work well which did not thus far so its a better world :) Jul 16 23:23:42 khem: go ahead Jul 16 23:25:25 ant__: I am doing a full build from scratch atm Jul 16 23:25:36 once thats successful I will send Jul 16 23:26:05 khem: great Jul 16 23:27:09 some portions of gcc runtime were not being compiled with interworking Jul 16 23:27:16 and thumb was failing miserably Jul 16 23:27:23 in some apps Jul 16 23:27:27 system would boot Jul 16 23:27:36 but some applications will have random failures Jul 16 23:27:48 that's the cause for the arm-mode override Jul 16 23:27:56 ? Jul 16 23:28:07 dont know Jul 16 23:28:21 we can test some recipe forcing arm and not thumb Jul 16 23:28:23 but we have thumb option and if we do then we should make it work well Jul 16 23:28:43 this was meant as temporary workaround :/ Jul 16 23:29:09 unfortunately only a few have the skills to debug such things (I think at uou and pb) Jul 16 23:29:13 with my RFC DISTOs will get the power to decide to use thumb if thumb is supported on selcted machine or not Jul 16 23:29:15 *you Jul 16 23:29:19 which is right way imp Jul 16 23:29:23 imo Jul 16 23:29:42 right Jul 16 23:29:46 a machine which supports both ARM and THUMB modes doesnt care what you use Jul 16 23:29:56 its only distro who should care Jul 16 23:30:14 btw remember to push arm4 to EABI in minimal Jul 16 23:30:17 iow thumb is a machine feature that can be used/not used by distro Jul 16 23:30:30 ant__: yes I have that patch lined up too Jul 16 23:30:40 will go in along with this RFC Jul 16 23:30:42 great, thx Jul 16 23:31:00 ELC-E 2010 : your proposal has been accepted Jul 16 23:31:02 !!! :) Jul 16 23:31:06 atm there is much buzz around thumb2 Jul 16 23:31:21 yes thats where I am headed Jul 16 23:31:32 once I have the thumb1 settled Jul 16 23:31:41 and thats the motivation too Jul 16 23:31:56 because thumb2 will use same machine properties Jul 16 23:31:58 I see Jul 16 23:32:13 we need to get the design right in place Jul 16 23:32:30 then we'll need to patch the failing recipes Jul 16 23:32:34 usually is asm :/ Jul 16 23:32:47 yes thats right Jul 16 23:32:59 Jay7: congrats Jul 16 23:33:07 khem: thanks! Jul 16 23:33:08 are u talking in ELC Jul 16 23:33:26 yes, I will present kexecboot there Jul 16 23:33:34 oh ok Jul 16 23:33:47 kexecboot is like a bootloader if i am not wrong Jul 16 23:33:54 using kernel as bootloader Jul 16 23:34:38 true Jul 16 23:34:53 nice does it work on qemu Jul 16 23:35:06 Jay7: time to watch tons of english telenovelas :p Jul 16 23:35:08 qemu-arm can't do kexec afaik Jul 16 23:35:35 ant__: russinglish will be ok :) Jul 16 23:35:43 ant__: yeah.. I've found tutor already ;) Jul 16 23:36:00 as long as you leave the vodka out of it Jul 16 23:36:32 well.. I hope CELF will sponsor this travel too.. Jul 16 23:36:44 Jay7: even better Jul 16 23:36:52 but I doubt Jul 16 23:37:29 we will check soon :) Jul 16 23:50:55 man, i dunno wtf i was smoking when i did patch.bbclass Jul 16 23:51:01 but that has got to be fixed up Jul 16 23:51:34 * Jay7 -> sleep Jul 17 01:50:01 good evening, anybody around? Jul 17 01:50:16 I got a problem with the recent git... Jul 17 01:51:39 "Unpacking recipes/glibc/glibc-2.9/etc/ld.so.conf to temp/work/armv4t-angstrom-linux-gnueabi/glibc-initial-2.9-r36.3/ Jul 17 01:52:54 cp: cannot stat `/home/igor/OE_HEAD/openembedded/recipes/glibc/glibc-2.9/etc/ld.so.conf': No such file or directory" Jul 17 01:53:12 anybody know how to fix this? **** ENDING LOGGING AT Sat Jul 17 02:59:56 2010