**** BEGIN LOGGING AT Wed Jul 13 02:59:57 2011 Jul 13 04:29:32 htns: it seems it did not include linux/input.h Jul 13 06:03:50 Does any one know how to change the machine ID in the kernel ? Jul 13 08:34:37 Morning everyone Jul 13 08:35:06 Anyone willing to share wpa_supplicant config snippet for ad-hoc with WEP/WPA? Jul 13 08:36:36 uhm adhoc Jul 13 08:38:34 woglinde: yeah, I am still playing with that Jul 13 08:38:53 Googled examples, but no WEP or WPA works with wext/nl80211 Jul 13 08:39:15 wpa_supplicant: PSK (ASCII passphrase) - hexdump_ascii(len=8): [REMOVED] Jul 13 08:41:21 "No keys have been configured - skip key clearing" Jul 13 08:50:12 davidlt still the sdio driver? Jul 13 08:50:18 yeah Jul 13 08:50:36 hms Jul 13 08:50:44 can you test a usb stick Jul 13 08:50:48 ? Jul 13 08:50:51 I would to that Jul 13 08:50:53 first Jul 13 08:51:01 to get a known working setup Jul 13 08:51:04 I just moved to udev 165, that solved problem with libertas wlan0 preping Jul 13 08:51:33 I am thinking about that for awhile, I do not have WiFi USB dongle Jul 13 08:51:44 uhm the are cheap Jul 13 08:52:08 you dont have a uhm how was that calles, something with shack Jul 13 08:52:13 radioshack Jul 13 08:52:17 or so market Jul 13 08:52:30 Would be nice to get one cheap, with Linux support (AP/ad-hoc) Jul 13 11:17:53 03Yu Ke  07master * rb631e92225 10bitbake.git/lib/bb/fetch2/__init__.py: (log message trimmed) Jul 13 11:17:53 fetcher2: retry mirror if upstream checksum mismatch Jul 13 11:17:53 This patch is for [YOCTO #1085] fix. Jul 13 11:17:53 If the upstream fails a checksum, retry from the MIRROR before giving up. Jul 13 11:17:53 This will add more robust fetching if an upstream serves a bad file or webpage. Jul 13 11:17:54 fetching of distcc prior to the move from samba -> googlecode is a good example Jul 13 11:17:55 of this. Jul 13 11:33:12 03Joshua Lock  07master * r570405f2f5 10bitbake.git/lib/bb/ui/crumbs/tasklistmodel.py: (log message trimmed) Jul 13 11:33:12 ui/crumbs/tasklistmodel: update brought in by column when possible Jul 13 11:33:12 When a package is orphaned we were not correctly updating the brought-in-by Jul 13 11:33:12 column if a later package additon would have brought that package in as a Jul 13 11:33:12 dependency. This patch ensures that orphan packages are correctly re-parented Jul 13 11:33:13 when appropriate. Jul 13 11:33:14 Partially addresses [YOCTO #1218] Jul 13 11:33:22 03Joshua Lock  07master * r4803c6d3d1 10bitbake.git/lib/bb/ui/crumbs/tasklistmodel.py: (log message trimmed) Jul 13 11:33:23 ui/crumbs/tasklistmodel: fix automatic removal of orphaned items Jul 13 11:33:23 The sweep_up() method intends to remove all packages with an empty brought in Jul 13 11:33:23 by column, this patch changes the implementation to be more reliable. Jul 13 11:33:23 Each time a removal is triggered we begin interating the contents model again Jul 13 11:33:23 at the beginning, only once the contents model has been iterated from start Jul 13 11:33:23 to finish without any removals can we be certain that there will be no more Jul 13 12:33:43 I created a simple recipe for a Qt application (ipmon-gui), but building my image fails with Jul 13 12:33:44 | * opkg_install_cmd: Cannot install package ipmon-gui. Jul 13 12:33:44 | * resolve_conffiles: Existing conffile /home/stiller/development/OMAP/OE_TEST/tmp/rootfs/ipmon-image/etc/device_table is different from the conffile in the new package. The new conffile will be placed at /home/stiller/development/OMAP/OE_TEST/tmp/rootfs/ipmon-image/etc/device_table-opkg. Jul 13 12:33:55 My application package contains no "device_table". Jul 13 12:33:56 The tmp/rootfs/ipmon-image/etc/device_table seems to come from openembedded/files/device_table-minimal.txt and tmp/rootfs/ipmon-image/etc/device_table-opkg from tmp/work/armv7a-angstrom-linux-gnueabi/initscripts-1.0-r128/device_table.txt Jul 13 12:34:36 A lot of people seem to have this problem, but nobody has a solution. Jul 13 12:34:45 Any idea? Jul 13 12:43:57 hii i am using OE and i want to compile all oe packages with some extra options provided to make like CFLAGS="-march=amdfam10" Jul 13 12:44:21 i want this to be done at some higher level than edit all the recipes Jul 13 12:44:43 i find the base.bb class and there i edit oe_runmake Jul 13 12:45:31 make CFLAGS+="-march=amdfam10" ${EXTRA_OEMAKE} Jul 13 12:46:16 but some issues in compiling Jul 13 12:46:46 anyone having idea how to append CFLAGS for all the packages Jul 13 12:48:11 its a bit urgent Jul 13 12:48:22 if anyone can provide some solution Jul 13 12:51:34 Is friendly arm kernel 2.6.36 buggy? Jul 13 13:24:32 hi all Jul 13 13:24:51 i am having trouble with the svn fetcher in bitbake Jul 13 13:25:22 i can manually checkout a package using the svn command in my shell Jul 13 13:26:01 but i get an error when trying to check it out via the svn fetcher Jul 13 13:26:07 can anyone help me Jul 13 13:33:32 can anyone help me with the svn fetcher Jul 13 13:38:53 cboyd: what does your svn url look like? Jul 13 13:39:53 svn://devonit.jira.com/svn/AGENT;module=trunck;proto=https Jul 13 13:40:22 gchiii: svn://devonit.jira.com/svn/AGENT;module=trunck;proto=https Jul 13 13:41:21 trunk with ck? Jul 13 13:41:34 gchiii: i can fetch the directory manually using the svn co command .. but bitbake gets an error Jul 13 13:42:35 Cubi_: typo ... sorry Jul 13 13:43:53 do you need a "rev=" argument? and have you told the computer to "permanantly store" the https key? Jul 13 13:45:01 gchiii: no ... I have not told the computer to "permanantly store" the https key. How would you do that? Jul 13 13:46:34 unfortunately i have no idea, but i remember reading somewhere that in order to get the svn fetcher to handle an https connection I first had to do my checkout manually and when prompted by svn tell it to store the key Jul 13 13:46:54 you can add user=xxx;pswd=yyy Jul 13 13:47:15 sorry, my caffeine intake this morning hasn't yet reached productive levels Jul 13 13:48:38 Cubi_: I already have the user and pswd set, I just omitted that from my description Jul 13 14:34:05 03Marco Cavallini  07master * r37bef33e36 10openembedded.git/conf/distro/kaeilos.conf: kaeilos.conf: changed version identifier Jul 13 14:34:15 03Marco Cavallini  07master * rb59743b913 10openembedded.git/recipes/linux/linux-2.6.28/mh355/defconfig: linux-2.6.28/mh355/defconfig: updated configuration file for mh355 kernel Jul 13 14:41:11 Solution for my question (2 hours ago): My package contained no files (make install didn't work). After fixing this, the problem with etc/device_table disappeared. OE seems to have a problem with empty packets. Jul 13 14:45:10 hm? Jul 13 14:45:27 there are hundreds of empty packages emitted every day. every 'task' package contains no files Jul 13 14:48:02 But tasks don't create a directory in tmp/work/arm* Jul 13 14:49:05 every recipe has a workdir. Jul 13 14:49:18 what exactly are you doing? Jul 13 15:01:47 I just had a reciepe for some software without a working "install" target in the Makefile. So "packages-split", "image" etc. were empty Jul 13 15:04:14 and? Jul 13 15:05:21 and I got this: Jul 13 15:05:22 | * opkg_install_cmd: Cannot install package ipmon-gui. Jul 13 15:05:22 | * resolve_conffiles: Existing conffile /home/stiller/development/OMAP/OE_TEST/tmp/rootfs/ipmon-image/etc/device_table is different from the conffile in the new package. The new conffile will be placed at /home/stiller/development/OMAP/OE_TEST/tmp/rootfs/ipmon-image/etc/device_table-opkg. Jul 13 15:06:14 http://bugs.openembedded.net/show_bug.cgi?id=3593 Jul 13 15:07:41 anyone around creating boot media for x86 from OE based builds? wondering what the best solution is - I have been using host based grub but would like to remove that dependency as modern grub is proving a bit flaky for me - would rather use grub built from OE Jul 13 15:08:49 ran across this http://www.cs.nmsu.edu/~pfeiffer/grub.html which I believe is saying to install grub in target filesystem and run that version when installing to removable media via chroot Jul 13 15:09:27 also saw reference to using syslinux but if I understand that correctly it only supports booting media from a FAT partition which I don't want Jul 13 15:40:07 meh Jul 13 15:43:34 hi guys! something went wrong with kernel v2.6.39.3: i get some really crazy output on ttySAC0 after decompression: http://pastebin.com/AuzgNbCG any ideas what might cause it? Jul 13 15:45:42 ug Jul 13 15:46:00 diff'ing the world's config.log from one run to the next shows some slighly worrisome implicit deps Jul 13 15:46:32 * Tartarus makes some quick changes locally to remember to dig at this more for real next Jul 13 15:47:15 diffing the world's config.log? yikes. bet you're glad its only oe-core Jul 13 15:48:16 heh, i remember playing with trying to share configure cache bits between different recipes Jul 13 15:48:18 world on oe-core+meta-oe had some failures, but that was my original plan Jul 13 15:48:32 kergoth, oe-core has a something for that now, kinda Jul 13 15:48:59 On the fly generation of some site info that's re-used later Jul 13 15:55:09 dFence, maybe your serial settings are different between uboot and the linux kenrel Jul 13 15:55:41 GNUtoo|laptop: I didn't change any settings, and with 2.3.36 it works fine Jul 13 15:56:10 dFence, ok Jul 13 15:57:24 what's your bootloader? Jul 13 15:57:27 do you have md? Jul 13 15:57:43 GNUtoo|laptop: superboot Jul 13 15:59:09 does it have an equivalent of md? Jul 13 15:59:17 GNUtoo|laptop: what's md? Jul 13 15:59:27 memory display Jul 13 15:59:37 like devmem2 or peek/poke but in uboot Jul 13 15:59:46 I guess you cannot dump the dmesg Jul 13 15:59:50 like over ssh Jul 13 15:59:51 GNUtoo|laptop: nada, superboot only parses an INI-file and boots whatever I tell it to boot - no output whatsoever Jul 13 15:59:52 right? Jul 13 16:00:28 GNUtoo|laptop: hi Jul 13 16:00:36 hi Jul 13 16:01:35 I am trying to run IBSS/Ad-hoc with libertas on Gumstix with WEP or WPA, but encryption is always dropped by wpa_supplicant. Any thoughts? Jul 13 16:01:48 hmmm Jul 13 16:02:04 Googled several examples, none of them worked. Jul 13 16:02:24 I didn't try IBSS alone yet Jul 13 16:02:29 Running wpa_supplicant 0.7.3 with -dd I get lines with [REMOVE] appended, looks like encryption is removed. Jul 13 16:02:35 I'd like to try a real batmand setup tough Jul 13 16:04:03 Never heard about, but just googled. Is it supported in OE ? Jul 13 16:04:17 it's not interesting for you I guess Jul 13 16:04:54 it's for mesh networking Jul 13 16:04:56 over IBSS Jul 13 16:05:12 and yes it's in oe Jul 13 16:06:18 I just need secure stable WiFi connectivity to the Gumstix, but it's not that easy do accomplish Jul 13 16:07:41 Ad-hoc with libertas, look amazingly stable, but no encryption Jul 13 16:12:33 ok Jul 13 16:26:41 I always get: Jul 13 16:26:43 PSK (ASCII passphrase) - hexdump_ascii(len=8): [REMOVED] Jul 13 16:26:43 PSK (from passphrase) - hexdump(len=32): [REMOVED] Jul 13 16:31:47 and what were you expecting? Jul 13 16:31:58 that's roughly what wpa-supplicant usually outputs Jul 13 16:33:44 sgw: kicking off a test now Jul 13 16:34:39 Tartarus: k Jul 13 16:36:59 pb_: I would like to have encryption, WEP/WPA Jul 13 16:41:38 well, that text doesn't indicate that you aren't getting it Jul 13 16:41:44 what leads you to believe that you are not? Jul 13 16:42:25 sgw: so far so good, it popped up qemu and I think is doing things Jul 13 16:47:43 hm, emacs and gnome-shell don't seem to play well together at all Jul 13 16:48:03 every time I minimize and unminimize emacs it seems to lose the ability to draw itself properly Jul 13 16:50:16 sgw: You want me to break out a few unrelated fixes in my siteinfo series or just do 'em all at once? Jul 13 16:51:00 sgw: and sec for a url Jul 13 16:51:18 http://git.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/update-siteinfo-preview Jul 13 16:51:25 Tartarus: let me take a look Jul 13 16:52:22 Tartarus: yes make the hicolor and libsamplerate0 separate from the siteinfo. Jul 13 16:58:48 k Jul 13 16:59:14 I'll add 2 more in quick then Jul 13 17:03:01 Is recipe information stored in ipk headers somewhere ? Jul 13 17:03:07 like recipe name Jul 13 17:03:30 yes Jul 13 17:03:33 OE: field seems to have something that resembles Jul 13 17:03:38 is that the one Jul 13 17:04:08 right Jul 13 17:04:12 pb_: Because on my laptop I see it as open and iwconfig doesn't show encryption enabled Jul 13 17:04:13 ok Jul 13 17:04:23 see package_ipk.bbclass Jul 13 17:04:23 fields.append(["OE: %s\n", ['PN']]) Jul 13 17:04:37 pb_: ah Jul 13 17:04:39 right Jul 13 17:04:58 is PN always = recipe name ? Jul 13 17:05:23 pb_: also: No keys have been configured - skip key clearing Jul 13 17:05:26 well, depends on your point of view I guess Jul 13 17:05:33 in some sense, PN is the recipe name by definition Jul 13 17:06:03 if what you actually want is the file name, that's available in PF or some such. I don't think it goes into a header in the output data, though maybe it is included in Source: Jul 13 17:06:04 yes and in some cases we muck with it e.g. BBCLASSEXTEND I guess Jul 13 17:06:43 ah PF cool Jul 13 17:07:13 sorry, no, PF isn't what you want Jul 13 17:07:18 I think ${FILE} is the thing I was thinking of Jul 13 17:07:28 hmm yes Jul 13 17:07:48 pb_: and does license go into the ipk header too Jul 13 17:09:01 hmm I dont think so Jul 13 17:09:08 in oe.dev yes, in oe-core no Jul 13 17:09:37 hmmm I think it will be nices to have it in oe.core as well what say ? Jul 13 17:09:50 yes, I think that would be a good thing Jul 13 17:10:17 afaik, it's only missing from oe-core because nobody has backported the patch yet Jul 13 17:10:32 I don't think there was any deliberate decision to leave it out. Jul 13 17:10:59 ah let me do it Jul 13 17:11:01 I need it Jul 13 17:11:23 it also shows a Build: fields Jul 13 17:11:25 cliff was asking about it on the mailing list too recently Jul 13 17:11:37 yes Jul 13 17:12:01 sgw: http://git.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/misc-preview look good to you? esp the top 2 which are new Jul 13 17:12:09 personally I think Build: is of rather questionable value Jul 13 17:14:07 pb_: and may be Priority: is not needed anymore ? Jul 13 17:14:24 correct, I think it should probably be deleted (or at least made distro-selectable) Jul 13 17:14:31 pb_: since we have decided to drop it from recipes Jul 13 17:14:47 yes, though of course DISTROs are still free to set it if they like. Jul 13 17:15:20 possibly the best answer would be to make the default value in bitbake.conf be the empty string, and have package_ipk only output the header if it's not empty. Jul 13 17:15:34 need to make sure that it doesn't break ipkg-make-index though, it might be checking for that header Jul 13 17:15:59 * pb_ go home now Jul 13 17:18:56 pb_: yeah, I will defer this to be a separate patch Jul 13 18:04:47 anyone opposed to adding the standard /usr aclocal search paths to the aclocal search paths for native recipes autotools execution? without that, its not possible to safely add a -native to ASSUME_PROVIDED if it provides a macro something else needs Jul 13 18:07:12 that reminds me, we can use a dirlist in the default search location for the macros, to avoid having to pass all the cmdline args Jul 13 19:43:28 does anyone understand the layer stuff in oe-core? Jul 13 19:44:38 cbrake: "layer stuff" is a little generic :) Jul 13 19:45:21 RP__: :-) well, I'm pondering whether to set up a new project using git submodules (like I usually do), or if the layer tools are a better option Jul 13 19:45:37 * cbrake takes another look at them ... Jul 13 19:46:26 cbrake: ah, right. I might not be much help there as I've not managed to migrate my workflow over to them yet myself :/ Jul 13 19:46:37 i can see why yocto/poky needed a different solution, but I don't see why other people hate submodules so much, i'm pretty happy with them for my personal setups, anyway Jul 13 19:46:45 no reason other than the downtime of the transition Jul 13 19:47:27 kergoth__: yeah, same here, I tend to keep finding myself using submodules Jul 13 19:47:48 my problem with submodules is simply if I'm trying to track a given "branch" (as opposed to a commit) submodules is difficult to keep in sync Jul 13 19:47:54 i know a lot of people dislike that they don't track a branch, but i like it, personally. its easy to update them if you need to, but at least you have a known sane point to go back to Jul 13 19:48:05 git subdo co master; git subdo pull -- done Jul 13 19:48:07 * kergoth__ shrugs Jul 13 19:48:11 Now if I want to track a given commit in my project.. I have used submodules successfully Jul 13 19:49:17 the problem with submodules is when someone updates a module, they need to also update the reference to the submodule.. and it's something -very- easy to forget Jul 13 19:49:41 some projects want/need that behavior though.. so submodules as-is are useful.. just not as useful as say "track master" Jul 13 19:50:12 fray: yeah, there is that problem, but typically in a production build system I need that anyway. Would be similiar to updating a SRCREV in a recipe, etc Jul 13 19:50:59 fray: though we have also tracked the HEAD in production build systems as well. Well, it would be nice if git submodules would give us that option. Jul 13 19:51:05 of course.. but where I've tried to use it in the past is manageing a hierarchy of layers.. I wanted to track the development of the layer -- not lock down to a given version of the layer.. Jul 13 19:51:30 what ended up happing is every time someone changed a submodule layer.. they forgot to update the item pointing to it.. and bugs that were fixed weren't showing up as fixed.. etc.. Jul 13 19:51:36 for me, I'd say the main repo *shoudln't* be updated for eveyr module change. I want to make sure the change actually worked, *then* update the references, the refs act as a last-known-sane state Jul 13 19:51:41 (or even the case where the maintainer of the layer doesn't have rights to my repo) Jul 13 19:51:59 but i guess it depends on the use case Jul 13 19:52:06 kergoth__ and thats exactly opposite of what I wanted.. submodules work and do what they're supposed to.. but my use case is simply different Jul 13 19:52:33 that is why I don't like submodules for projects where I want to segregate the work.. but they have a common development group.. it adds unnecessary overhead Jul 13 19:53:23 it sounds like what you want is more like svn externals. the glclient script used by chromium looked interesting for that sort of a case, it's like 'repo', but the metadata is stored in repo, not in an xml file somewhere Jul 13 19:53:42 * kergoth__ hasn't had a chance to poke at it very much yet Jul 13 19:54:16 in our (wind river) version of git, we added something called subgits.. Jul 13 19:55:01 and it does what we want.. basically you add in an external git repository to a list and pointer.. and then simply specify the branch.. on a git pull, and similar it will automatically recurse over that.. and update it.. and keep it on the same branch referenced etc.. Jul 13 19:55:44 http://dev.chromium.org/developers/how-tos/depottools is the chromium one i was talking about. they do things a bit differently since they need to support both svn and git, though Jul 13 19:57:30 one thing that really bugs me about submodules is the disconnect between .gitmodules and .git/config Jul 13 19:57:38 would it really ahve killed them to operate based on .gitmodules directly? Jul 13 20:13:38 kergoth__, yes, some sort of local override of submodule paths is required Jul 13 20:14:20 also think of the case where the repository location of the submodule changed and you want to check out an old version Jul 13 20:15:24 hmm, fair enough Jul 13 20:33:57 i am a bit lost about how mesa-dri now enabled dri modules Jul 13 20:34:17 for me it just enables swrast but I'd like to have more drivers enabled Jul 13 21:36:25 * khem likes submodules Jul 13 21:37:49 hi khem Jul 13 21:47:37 woglinde: hey Jul 13 21:50:43 with oe-core, does BBFILES updated by the layer? Jul 13 21:51:15 I don't see any evidence its being set in my conf directory Jul 13 21:51:54 bit yet it is BBFILES="/scratch/oe/openembedded-core/meta/recipes-*/*/*.bb" when I run bitbake Jul 13 21:54:30 yes, it's normal for each layer's conf/layer.conf to add to BBFILES Jul 13 21:57:38 cbrake: are you using angstrom ? Jul 13 21:58:14 khem: yes Jul 13 21:58:25 khem: I've done a build using the setup scripts Jul 13 21:58:38 khem: but trying to get one set up from scratch that is a little more streamlined Jul 13 21:58:44 khem: just layers I need, etc Jul 13 21:59:31 look at conf/site.conf file Jul 13 21:59:37 under setup scripts Jul 13 22:00:24 cbrake: streamlined Jul 13 22:01:00 cbrake:Do u mean for building ? Jul 13 22:01:17 khem: I only need a few layers for this product, etc Jul 13 22:01:24 oh Jul 13 22:01:45 all that is in layerman Jul 13 22:01:56 khem: what is layerman? Jul 13 22:02:18 khem: can I lock layers down to specific versions for production builds, etc? Jul 13 22:02:25 and conf/bblayers.conf Jul 13 22:02:50 hmm yes you can I suppose Jul 13 22:03:59 see sources/layers.txt Jul 13 22:04:13 cbrake: thats where you can specify which rev you want to lock to Jul 13 22:04:20 or even branch Jul 13 22:04:46 by default we currently track HEAD of master of every layer Jul 13 22:06:12 khem: so do you think the angstrom layer tools are the way to go? Jul 13 22:06:20 khem: vs git submodules, etc Jul 13 22:06:41 cbrake: I prepared slugos master project on makefile/submodules Jul 13 22:07:00 and its good since I have control over the sub modules Jul 13 22:07:12 but with angstrom layerman I have a bit more control Jul 13 22:07:25 the slugos one was based on your templates Jul 13 22:07:55 khem: so if you were setting up a new project, you'd probably go with layerman? Jul 13 22:08:04 cbrake: I am currently happy with angstrom layerman and it can be enhanced easily Jul 13 22:08:13 cbrake: yes Jul 13 22:08:17 khem: ok, you've convinced me :-) Jul 13 22:08:30 khem: I like the ability to track HEAD or version Jul 13 22:08:39 cbrake: yeah me too Jul 13 22:08:49 so I can lock some erratic layers Jul 13 22:08:57 and use head of sane layers Jul 13 22:09:38 khem: I don't like the ~/.oe/.. stuff Jul 13 22:09:57 khem: setup files should be in the project dir it seems as you may have multiple projects Jul 13 22:10:17 cbrake: me too. I have always meant to change it but no time :) Jul 13 22:10:32 if you create a patch send it over and we can work it up Jul 13 22:10:34 khem: ok, perhaps I'll change it then :-) Jul 13 22:10:36 cbrake: thats the whole reason i use ~/.oe, so i can share things across multiple projects :P Jul 13 22:10:46 just dont' use it the way the angstrom folks do :) Jul 13 22:10:55 kergoth__: heh Jul 13 22:11:25 reason was to use same env on NFS mounted home across machines Jul 13 22:13:56 but oe-core is much improvement. my build hardly breaks due to others faults :) Jul 13 22:14:14 iow pulls Jul 13 22:42:15 damn, my motivation is just gone all of a sudden Jul 13 22:51:16 heh Jul 13 22:51:23 I hate it when that happens Jul 13 22:52:26 kergoth__: go for a game of racquetball Jul 13 23:29:01 fray: around ? Jul 13 23:29:05 ya Jul 13 23:29:16 fray: I am buildin core-minimal-image Jul 13 23:29:31 and its not installing opkg bits correctly Jul 13 23:29:45 I've not worked with opkg.. but I can try to help Jul 13 23:29:47 basically the binaries are in rootfs but not var/lib/opkg Jul 13 23:30:10 now var/lib/opkg is an empty directory packaged with opkg itself Jul 13 23:30:43 var/lib/opkg is erased by default in minimal images using the rootfs_ipk format.. there is an option that disables the packaging system (for small builds) and that erases the DB.. you need to disable it Jul 13 23:31:00 but http://pastebin.com/34Ryu2Sk is content of opkg ipk Jul 13 23:31:10 ahhh Jul 13 23:31:24 but it does not remove the opkg binaries Jul 13 23:31:26 thats strange Jul 13 23:31:34 so opkg ends up in image Jul 13 23:31:39 thats because something told the system to include them.. not sure what Jul 13 23:31:43 but when you boot it its dysfunctional Jul 13 23:31:52 opkg i.e. Jul 13 23:32:10 hmm Jul 13 23:32:11 a minimal image with RPM enable, does NOT install rpm, and erases the rpm database.. Jul 13 23:32:18 thats what I would have expected with opkg and deb Jul 13 23:32:23 yes Jul 13 23:32:35 but the installation of the packaging system is based on the dependencies.. Jul 13 23:32:39 but I see opkg being demanded by core-image-minimal Jul 13 23:32:42 in deptree Jul 13 23:33:02 I don't know why that is Jul 13 23:33:32 "task-core-boot" -> "opkg" [style=dashed] Jul 13 23:37:35 hmmm Jul 13 23:38:30 http://pastebin.com/JH0x9KXZ Jul 13 23:38:41 this is snippet from rootfs_ipk.bbclass Jul 13 23:38:46 which punts it out Jul 13 23:39:08 but it does not remove the package manager itself from rootfs Jul 13 23:40:02 it should never have to remove the package maanger from the rootfs.. since it shouldn't have been added Jul 13 23:40:07 figure out why it's added.. Jul 13 23:40:52 ${@base_contains("IMAGE_FEATURES", "package-management", "${ROOTFS_PKGMANAGE}", "${ROOTFS_PKGMANAGE_BOOTSTRAP}",d)} Jul 13 23:40:58 I am sure this is adding it Jul 13 23:41:17 this is POKY_BASE_INSTALL Jul 13 23:41:48 shouldn't be, if package-management is not there.. Jul 13 23:42:04 if package-management is there, then the ROOTFS_PKGMANAGE should be added, otherwise it shouldn't.. Jul 13 23:42:13 opkg, rpm, deb all should operate the same as far as that goes Jul 13 23:44:06 package-management is not in IMAGE_FEATURES Jul 13 23:44:15 so which means it should not have pulled opkg at all Jul 13 23:44:44 correct.. which means something is dragging it in.. Jul 13 23:47:59 * Tartarus wonders if oe.dev's arm mutex thingy is being obeyed, checks Jul 13 23:48:22 task-core-boot is dragging it in Jul 13 23:58:37 aha Jul 13 23:58:39 VIRTUAL-RUNTIME_update-alternatives="opkg" Jul 13 23:58:41 there we go Jul 13 23:58:56 and VIRTUAL-RUNTIME_update-alternatives appears in RDEPENDS of core-task-boot Jul 13 23:59:20 so then update-alternatives should likely be packaged outside of opkg so opkg doesn't need to exist Jul 13 23:59:38 and conf/distro/angstrom-2010.x.conf:VIRTUAL-RUNTIME_update-alternatives = "opkg" Jul 14 00:00:46 fray: what do u mean Jul 14 00:01:39 hmm, iirc chkconfig includes a C implementation of update-alternatives, i keep meaning to look at it, and keep forgetting to. one of these days Jul 14 00:02:38 update-rc.d is smartly removed if package-management is not enabled Jul 14 00:02:54 but I wonder if opkg can do the job with out database at all Jul 14 00:03:59 so in package_ipk class when it does Jul 14 00:04:01 opkg-cl ${IPKG_ARGS} --force-depends remove update-rc.d base-passwd Jul 14 00:04:09 probably we should include opkg as well ? Jul 14 00:05:54 if not set default is update-alternatives-cworth Jul 14 00:07:36 I have no idea.. sorry.. I havn't looked at that stuff Jul 14 00:08:52 hrm Jul 14 00:09:06 I am sure in case where its set to update-alternatives-cworth then that goes in into the core-image-minimal Jul 14 00:09:39 * kergoth__ thinks about apis for recipe discovery Jul 14 00:09:41 but if we punt out opkg then distros trying to use opkg to do update-alternatives stuff with opkgs are at loss Jul 14 00:14:00 hmm, maybe i should have some sort of RecipeSource, e.g. DirectorySource("path/to/some/dir"). a given file format takes a collection of sources and uses them to pull in the files of the correct format, and construct nodes Jul 14 00:14:04 * kergoth__ ponders Jul 14 00:15:45 I punted on the parser for now, I'm using json as the recipe format just so I can play around with the api to use, heh Jul 14 00:15:48 hmm Jul 14 00:20:57 I wonder if I should implement layers / prioritization in a general way, or if that should be handled by the given recipe file format Jul 14 00:22:27 * Tartarus pokes db + site stuff Jul 14 00:22:34 * kergoth__ scratches head Jul 14 00:22:34 I think we've got a ton of extra crap Jul 14 00:22:47 khem, is uclibc building in oe-core atm? If so, just a LIBC = "uclibc" needed? Jul 14 00:22:50 heh, i'm not surprised, that's been accruing stuff for many years Jul 14 00:23:25 * Tartarus starts wondering, really about emptying all of the site files and starting from scratch Jul 14 00:26:47 s/LIBC/TCLIBC Jul 14 01:05:55 what's the default rootpw nowadays? Jul 14 01:06:01 in oe-core, sans distro Jul 14 01:06:22 is it not just set to empty? Jul 14 01:07:15 couldn't ssh in Jul 14 01:07:24 could be the nopass patch wasn't applied Jul 14 01:07:29 sgw: ? ;) Jul 14 01:07:50 I had to log in on the console first and set a non-empty password Jul 14 01:08:24 not empty, heh Jul 14 01:08:28 afk a few Jul 14 01:09:21 hrmm, file://allow-nopw.patch;apply=no \ Jul 14 01:10:43 looks like we only enable that patch if IMAGE_FEATURES contains debug-tweaks Jul 14 01:15:09 Tartarus: yes TCLIBC=uclibc should doo it Jul 14 01:18:47 Tartarus: here Jul 14 01:19:04 on a call with china Jul 14 01:24:11 sgw: does the test framework know what the default root pw is? Jul 14 01:25:39 Tartarus: ping jiajun in the #yocto irc, he is in the same meeting as me right now Jul 14 01:25:52 k Jul 14 01:25:54 post dinner Jul 14 01:25:59 sure Jul 14 02:59:23 Hi. I'm trying out console-image, i'm using tip as of 75abeb98e09b045d972b54f831ee6fc30300c257 and on a 2.6.34 board, i'm seeing lots of errs on bootup http://pastebin.com/80dMQPMi **** ENDING LOGGING AT Thu Jul 14 02:59:57 2011