**** BEGIN LOGGING AT Mon Nov 15 02:59:58 2010 Nov 15 07:07:01 gm Nov 15 07:38:27 do we have a page on how to create a patch? I could not really find it in the wiki Nov 15 07:39:00 yeah Nov 15 07:39:17 http://wiki.openembedded.org/index.php/How_to_submit_a_patch_to_OpenEmbedded Nov 15 07:40:55 mrj10: yeah, ifound that one, but ghis is mostly for the git part, I was looking for a page to send to someone which tells to use diff -urN or quilt and on what dir etc; Nov 15 07:42:25 ah, ok. sorry about that, i'm not aware of a page for that. would be pretty useful though Nov 15 07:42:54 yup Nov 15 08:21:58 morgen Nov 15 08:23:21 hrw: morgen :) Nov 15 08:48:12 good morning Nov 15 08:56:15 morning Nov 15 09:10:26 valhalla: ciao Nov 15 10:18:43 http://tinderbox.openembedded.net/builds/103097/ Nov 15 10:19:00 interesting fail on incremental build of testing-next Nov 15 10:19:11 only tosa was affected Nov 15 10:19:32 ant_work: please look Nov 15 10:21:29 ah.. hehe.. looks like race condition Nov 15 10:22:10 libqtaux2 is in do_fetch state at that moment Nov 15 10:23:06 and oestats log is unuseful at all Nov 15 10:25:18 Jay7: hey Nov 15 10:25:39 I'll post to ML Nov 15 10:25:44 he..I thought I already had seen all race conditions ;) Nov 15 10:25:55 this one seems new to me Nov 15 10:47:04 sh@t.. sent from wrong email Nov 15 10:47:23 gmail web-interface suxx sometimes :( Nov 15 10:53:12 well, sent Nov 15 13:30:21 hi JaMa do you know what to do to get usb0 up at the first boot? Nov 15 13:30:31 I just bitbaken a console-image for the bug device Nov 15 13:30:35 and it's still not working Nov 15 13:30:48 despite of having an up to date openembedded.dev tree Nov 15 13:31:55 GNUtoo|laptop: is g_ether or whatever module loaded during boot and auto in /etc/network/interfaces? Nov 15 13:32:07 I'll look Nov 15 13:32:55 auto usb0 (as I tought) Nov 15 13:32:56 and we had problems with connman shuting all ifaces down by default (resolved by not shiping connman in our images) Nov 15 13:33:23 g_ether 28788 0 Nov 15 13:33:41 ah ok Nov 15 13:33:48 I'll look if there is connman Nov 15 13:33:57 but it's a stock angstrom console-image Nov 15 13:34:02 cat /etc/modutils/g_ether -> g_ether? Nov 15 13:34:30 yes that too is there Nov 15 13:34:57 and connman was not even built Nov 15 13:35:39 [oe] Fwd: [SOLVED]Network init script error?? Nov 15 13:35:41 then I'm out of ideas Nov 15 13:35:49 about that....did the commiter bump PR Nov 15 13:35:52 or something like that Nov 15 13:37:22 what was the commit Nov 15 13:37:26 I don't find it Nov 15 13:37:51 I also see it only in gumstix-user and one fwd in oe Nov 15 13:38:46 86925741fc7a20f1948edd081e1481adfdae69c3 Nov 15 13:39:50 but I don't use ifupdown package Nov 15 13:40:23 ah ok Nov 15 13:40:26 what do we use then Nov 15 13:40:40 ifup from busybox (in shr images) Nov 15 13:40:50 ok Nov 15 13:41:01 GNUtoo|laptop: you can try ifup -a Nov 15 13:41:09 ok Nov 15 13:41:14 maybe just initscript wasn't called during boot.. Nov 15 13:41:27 still doesn't come up Nov 15 13:41:29 here's the issue Nov 15 13:41:47 ifup usb0 => ifup: interface usb0 already configured Nov 15 13:41:48 then I would try to debug ifup -a Nov 15 13:42:49 pb_: ping Nov 15 13:43:08 what could be the cause of "interface usb0 already configured" Nov 15 13:43:24 note that I've serial cable for that hardware Nov 15 13:43:43 can it be a race somewhere? Nov 15 13:43:51 because at second boot it's fine Nov 15 13:43:58 but not a lot of people have the serial cable Nov 15 13:44:11 so....how they are supposed to reboot if GUI doesn't work yet Nov 15 13:44:17 and that they have no usbnet Nov 15 13:44:52 eth0,eth0:avahi,eth1,lo came up Nov 15 13:52:28 JaMa, if I remember well the freerunner suffered from the same problem a very long time ago Nov 15 13:54:14 GNUtoo|laptop: yes but that was from connman Nov 15 13:54:24 ok Nov 15 13:54:28 GNUtoo|laptop: ifup configuring it and then connman just shuting down Nov 15 13:56:14 here's /etc/network/interfaces: http://pastebin.com/95443xem Nov 15 13:56:19 I'll compare with n900 Nov 15 13:58:18 you can try it without setting hwaddress Nov 15 14:01:51 didn't wor Nov 15 14:01:55 *didn't work Nov 15 14:01:59 here's the first boot : Nov 15 14:02:27 http://pastebin.com/h9Hr0mkJ Nov 15 14:02:28 note that: Nov 15 14:03:18 Configuring network interfaces... ifconfig: SIOCGIFFLAGS: No such device Nov 15 14:03:18 ifconfig: SIOCSIFADDR: No such device Nov 15 14:03:32 so maybe it does that: Nov 15 14:03:39 ifup usb0 when usb0 is not there Nov 15 14:03:40 it fails Nov 15 14:03:45 but record the up state Nov 15 14:04:18 maybe the only way is to compile in the knerel image g_ether Nov 15 14:04:22 yes looks like it Nov 15 14:04:23 and not as module Nov 15 14:04:34 or is there any other way to fix that? Nov 15 14:04:41 because it would prevent usb-host I bet Nov 15 14:04:43 well it works ok on freerunner with g_ether as module Nov 15 14:04:53 hmmm Nov 15 14:05:06 maybe reorder initscripts Nov 15 14:05:22 ok Nov 15 14:05:24 I'll look Nov 15 14:06:35 hmmm Nov 15 14:06:39 what is usb-gadget-mode recipe Nov 15 14:06:41 I'll look Nov 15 14:07:17 or you can ping Stefan Schmidt as he is maintaining bug devices IIRC Nov 15 14:07:34 I'm not even sure that he has a bug 1.x Nov 15 14:07:42 I don't have a bug 2.0 yet Nov 15 14:07:57 and want to have at least minimal support for bug 1.x in angstrom/oe Nov 15 14:08:03 but I'll ping him anyway Nov 15 14:08:16 but he works for buglabs, doesn't he? Nov 15 14:10:19 yes he does Nov 15 14:10:34 Stefan has bug 2.0 - atleast prototype Nov 15 14:10:40 yes I know Nov 15 14:10:51 but I wonder if they sent him a bug 1.x too Nov 15 14:11:04 rather not Nov 15 14:11:12 ok Nov 15 14:11:14 bug 1.x is considered out-of-support Nov 15 14:11:21 ok Nov 15 14:11:34 which is why I want it supported by angstrom Nov 15 14:11:58 there are some plan to support bug 1.x but only during buglabs developers's free time Nov 15 14:20:59 moving the networking script works Nov 15 14:54:13 JaMa: hi Nov 15 14:55:17 hey guys Nov 15 14:56:06 hi kergoth Nov 15 15:02:54 pb_: Hi, do you remember why you had to add trailing slash in 952e5e2b7a5c1deefc939594d40b81a71fb16a54 Nov 15 15:03:40 pb_: it's causing some troubles now, see http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-November/026794.html Nov 15 15:05:22 pb_: my guess is because of micro distro collapsing /usr to / Nov 15 15:05:24 JaMa: not offhand, sorry. it was more than two years ago and my memory is not that good. :-} Nov 15 15:05:58 if I had to guess I would say it was probably something to do with either an empty prefix going wrong, or something to do with micro setting prefix=/ not /usr. Nov 15 15:10:16 does anyone know why task-sdk-native isn't actually inheriting task? Nov 15 15:10:26 it breaks adding "dev" to IMAGE_FEATURES, it seems Nov 15 15:10:42 * kergoth_ tests with it doing so Nov 15 15:12:44 GNUtoo|laptop: ^^ then I guess your version with optional_slash should work Nov 15 15:13:18 hmmm Nov 15 15:15:11 GNUtoo|laptop: ah and maybe not if there is some assignement ie version="1" then it would OEPREFIX it too if I see it right Nov 15 15:15:24 hmm, no opinions on the variable typing stuff eh? Nov 15 15:20:20 version with optional_slash and empty prefix; echo version=1 | sed -e 's:=\("\?\)\(/\?\):=\1OEPREFIX\2:' -> version=OEPREFIX1 Nov 15 15:21:21 I guess that setting prefix="/" instead of prefix="" in micro.conf will have also some bad side effect (at least doubled /) Nov 15 15:22:15 doubled / breaks builds on certain os's, i can't remember which though, iirc either osx or cygwin Nov 15 15:22:18 :\ Nov 15 15:22:20 annoying Nov 15 15:23:39 reading Nov 15 15:29:09 what should we do then? Nov 15 15:30:09 03Tom Rini  07org.openembedded.dev * rb95cf40f3e 10openembedded.git/site/powerpc-common: Nov 15 15:30:09 powerpc-common: Fix ac_cv_sizeof_ptrdiff_t size Nov 15 15:30:09 This was set to 0 originally but should be 4. No idea how Nov 15 15:30:09 modphp built with this originally. Nov 15 15:30:09 Signed-off-by: Tom Rini Nov 15 15:31:11 GNUtoo|laptop: well you can add "/usr" case, without using $prefix variable, but that's just ugly Nov 15 15:31:37 pb_: hey, long I don't read of you, hope all's well Nov 15 15:32:13 GNUtoo|laptop: but at least wont break on empty $prefix Nov 15 15:32:27 pb_: could you comment about ML thread titled "lib_package.bbclass: fix packaging of static libs, inherit binconfig"? Nov 15 15:33:52 hmmm Nov 15 15:34:18 or we could check on empty prefix? Nov 15 15:34:34 or use python re? Nov 15 15:35:06 * GNUtoo|laptop should check the micro distro Nov 15 15:35:46 ant_work: I'll take a look Nov 15 15:35:57 ah ok Nov 15 15:36:08 so the prefix is "" and not / Nov 15 15:36:15 in the case of micro Nov 15 15:36:24 we could skip this in case of micro Nov 15 15:36:27 or something like that Nov 15 15:36:29 ? Nov 15 15:36:38 or handle differently Nov 15 15:36:50 ant_work: hm, the whole thing looks pretty bogus to me. what exactly is the problem that these patches are trying to solve? Nov 15 15:37:35 this business of putting .a files into -static is not correct, at least not without additional logic. Nov 15 15:38:44 I'm missing the point too: "Although it packaged ${bindir}/*-config into FILES_${PN}-dev, it did not include the required mangling applied by binconfig.bbclass" Nov 15 15:38:45 I wonder about the case where a recipe doesn't build shared libs at all -- the .a should clearly go into -dev in that case Nov 15 15:39:06 kergoth, may i push these 3? http://busybox.net/~aldot/bitbake/20101115-1628/ Nov 15 15:39:15 JaMa, /usr would break on empty prefix just the same I bet no? Nov 15 15:39:23 it won't be converted Nov 15 15:39:33 kergoth_: right, that's the additional logic I was thinking of. Nov 15 15:39:38 so it wouldn't point to the right dir Nov 15 15:39:40 but at the same time Nov 15 15:39:48 kergoth, nothing fancy, just cosmetics Nov 15 15:39:49 blindvt`: yep, looks good Nov 15 15:39:49 it wouldn't include the -L/usr/lib Nov 15 15:39:56 blindvt`: to master only, of course Nov 15 15:40:08 kergoth_: arguably it does make some sense to relegate the .a files to some -static ghetto if the library is shipping a .so with full functionality (but see glibc for example, which needs .a even for regular linking). Nov 15 15:40:10 wouldn't point to the right dir means staging dir Nov 15 15:40:19 kergoth, done. Thanks! Nov 15 15:40:21 03Bernhard Reutner-Fischer  07master * rde34a403e2 10bitbake.git/lib/bb/cache.py: Nov 15 15:40:21 bb.cache: only log if the respective action was taken Nov 15 15:40:21 This avoids alot of misleading log-messages like "Removing FOO from cache" Nov 15 15:40:21 if FOO was not in the cache and as such is not a removal candidate. Nov 15 15:40:21 Signed-off-by: Bernhard Reutner-Fischer Nov 15 15:40:30 but, clearly, if the .a libs are the only ones you have, a -dev package that doesn't include them is not going to be very fulfilling. Nov 15 15:40:32 03Bernhard Reutner-Fischer  07master * r5b85de2c71 10bitbake.git/lib/bb/parse/parse_py/BBHandler.py: Nov 15 15:40:32 BBHandler: use os.path in inherit() Nov 15 15:40:32 Signed-off-by: Bernhard Reutner-Fischer Nov 15 15:40:33 03Bernhard Reutner-Fischer  07master * r858d704d71 10bitbake.git/lib/bb/parse/parse_py/BBHandler.py: Nov 15 15:40:33 BBHandler: simplify supports() Nov 15 15:40:33 Signed-off-by: Bernhard Reutner-Fischer Nov 15 15:40:37 GNUtoo|laptop: then you can remove trailing slash in sed calls and use something like $prefix_with_at_least_slash instead of $prefix (prepared in that method before calling sed) Nov 15 15:40:38 * kergoth_ nods Nov 15 15:41:02 blindvt`: thanks, appreciate it, there are a lot of things like that we could clean up -- bitbake isn't particularly pythonic today :) Nov 15 15:41:12 GNUtoo|laptop: and with that I meant '/usr', -L/usr/lib should be fine because it's handled without trailing slash Nov 15 15:41:38 GNUtoo|laptop: and also '/usr' wont match because there is '="?/usr' Nov 15 15:41:44 micro.conf has no slash at all: prefix = "" Nov 15 15:42:25 yes and in this case $prefix_with_at_least_slash would be just "/" Nov 15 15:42:44 yes but "" won't be detected Nov 15 15:42:51 and won't be converted Nov 15 15:42:52 the same with exec_prefix Nov 15 15:43:10 but that's right, isn't it? "" shouldn't be converted Nov 15 15:43:40 GNUtoo|laptop: micro now has empty ${prefix} but that's why there is "/" appended in that sed call directly Nov 15 15:43:41 maybe it shouldn't in case of normal distros, but how do you link to something like that: Nov 15 15:43:57 -L{STAGING_LIBDIR}/somedirs/ Nov 15 15:43:59 pb_: he.. I recently committed a fix about unpackaged stuff for libgtkstylus (http://tinyurl.com/3ypj2wr) and was now wondering about alternative ways to do that Nov 15 15:44:02 GNUtoo|laptop: you would move it to $prefix_with_at_least_slash and then you wouldn't need it in sed call Nov 15 15:44:03 using the micro distro Nov 15 15:44:55 where do you get -L{STAGING_LIBDIR}/somedirs/? Nov 15 15:45:02 I'll look Nov 15 15:45:18 anything involving ${STAGING_LIBDIR} ought to "just work" with micro. I don't quite understand what the issue is there. Nov 15 15:45:18 and if anyone is interrested, in http://busybox.net/~aldot/bitbake/20101115-1628/ there are 3 files "commit-*" that are an excerpt of the stuff i have accumulated for oe Nov 15 15:46:08 http://pastebin.com/K4Y4HH9A Nov 15 15:46:35 GNUtoo|laptop: ? Nov 15 15:46:45 JaMa, I'll explain Nov 15 15:46:57 to name them: staging.bbclass: do not fill log with empty .la files Nov 15 15:46:58 native{,sdk}.bbclass: log redundant depends Nov 15 15:47:23 GNUtoo|laptop: if there is -L/usr/lib/something then it has to be changed to -L{STAGING_LIBDIR}/something for sysroot Nov 15 15:47:37 yes Nov 15 15:47:37 and (my favourite that IIRC was sent to the list a couple of months already): base.bbclass: depend unpacking on native unpackers Nov 15 15:47:40 * blindvt` & Nov 15 15:47:42 I'll re-explain Nov 15 15:47:46 blindvt`: should think about using git for that stuff, can easily put bitbake/oe clones on github to share branches around -- thats what I do, can always fork mine Nov 15 15:47:51 let's take micro with prefix="" Nov 15 15:47:55 GNUtoo|laptop: but if there is -L{STAGING_LIBDIR}/something already (probably shouldn't because it won't work on target) then it's fine and you don't need to mangle it Nov 15 15:47:56 prefix="" is not detected Nov 15 15:47:56 blindvt`: ah, yes, the native unpackers one should really go in, will look Nov 15 15:48:31 but if the package in installed in -L{STAGING_LIBDIR}/something Nov 15 15:48:45 how do you know how to link with it without the right flag Nov 15 15:48:46 ? Nov 15 15:49:16 I don't understand you or you don't understand me :) Nov 15 15:49:17 because binconfig will say -l for the libsomehting but not -L because it wouldn't have been mangled Nov 15 15:49:21 indeed Nov 15 15:49:26 kergoth: also, log redundant depends sounds good Nov 15 15:49:27 I'll try to re-explain better Nov 15 15:49:33 give me some time Nov 15 15:49:39 GNUtoo|laptop: there is / for mictor Nov 15 15:49:40 ant_work: yeah, good cleanup thing Nov 15 15:49:41 micro Nov 15 15:49:45 GNUtoo|laptop: maybe you could provide an actual concrete example of what is going wrong. Nov 15 15:49:56 GNUtoo|laptop: but not in prefix variable, but added later in that sed Nov 15 15:50:01 pb_, for micro? I didn't build something yet with it Nov 15 15:50:23 kergoth, i've though of asking for a personal branch (dunno if i ever did officially) but keeping that stuff local is good enough for my testing purposes Nov 15 15:50:42 blindvt`: the problem with personal branches on git.oe is the limitations -- it doesn't let you remove your own branches, or push non-fast-forwards Nov 15 15:50:46 GNUtoo|laptop: if you move / from sed to prefix_with_at_least_slash only in case prefix is empty, then you'll fix !micro distros and won't break micro Nov 15 15:50:47 which is why i use github for it now Nov 15 15:51:31 GNUtoo|laptop: mmt I'll show diff Nov 15 15:51:35 ok Nov 15 15:51:42 kergoth_, i also like that "log redundant deps" for classextended recipes. Just a thought Nov 15 15:51:46 * blindvt` nods kergoth_ Nov 15 15:52:00 blindvt`: check out https://github.com/kergoth/openembedded/compare/master...typing as an example of the nice interface for showing people your changes Nov 15 15:52:07 can even add notes to specific lines in the commits, as review Nov 15 15:52:13 :) Nov 15 15:52:20 * kergoth_ nods at log redundant deps Nov 15 15:52:37 about this... Nov 15 15:53:04 kergoth_: somehow removing an unused dep (zlib) solved a race issue with lzma-native :( Nov 15 15:53:18 I still don't see how... Nov 15 15:53:43 blindvt`: my one concern with the log redundant patch is the duplication of code -- course, there's already a bunch of duplicated code between those classes, but still Nov 15 15:53:47 I think I know where is the misunderstanding Nov 15 15:53:50 it's from me Nov 15 15:53:53 # Collapse /usr into / Nov 15 15:54:02 prefix = "" Nov 15 15:54:07 that's from micro.conf Nov 15 15:54:13 blindvt`: the other concern i had at first glance was the virtclass "magic" in the native unpackers commit, haven't had time to look closer to see what the logic was Nov 15 15:54:23 maybe others have Nov 15 15:54:31 JaMa, you're thinking that the prefix is prefix="/" Nov 15 15:54:39 and I'm thinking that the prefix is prefix="" Nov 15 15:54:41 right? Nov 15 15:55:04 I wonder how it goes to "/" tough Nov 15 15:55:42 GNUtoo|laptop: http://paste.pocoo.org/show/291633/ Nov 15 15:56:51 ok Nov 15 15:57:55 pb_, kergoth_: what do you think (about idea I haven't even tried it yet), is it too ugly? http://paste.pocoo.org/show/291633/ Nov 15 15:58:42 JaMa: it does seem fairly ugly. how about allowing the regex to accept either / or end-of-line at that point? Nov 15 15:59:00 also, ?= is a bitbake-ism, not a python thing, afaik Nov 15 15:59:12 and the bitbake idiom is that the var is *unset*, not *empty* Nov 15 15:59:17 unless i'm missing something Nov 15 15:59:44 actually, I guess that wouldn't work very well since you would end up with any empty var being set to the sysroot, not necessarily what you want. Nov 15 15:59:45 heh Nov 15 16:00:33 i think matching / or eol would be ideal, indeed Nov 15 16:00:47 I suspect this is one of these problems that isn't really amenable to a completely general-purpose solution. the var assignments in the generated binconfig files just don't contain enough semantic information to be able to tell unambiguously whether they are meant to be pathnames or not. Nov 15 16:01:07 and what if it points to something else like =/srvdir/blah? Nov 15 16:01:08 it was my first thought when i saw the email -- but i'm not sure if you can use $ in a | block in sed Nov 15 16:01:12 heh Nov 15 16:01:28 JaMa: just an absolute pathname? then it would be left alone, right? Nov 15 16:02:03 yeah, the binconfig sedding is a hack, but one that catches a good chuck of them -- maybe the answer is these particular cases should append extra sed's in the recipe Nov 15 16:02:03 pb_: but I guess that /srvdir is installed in sysroot too Nov 15 16:03:11 03Koen Kooi  07org.openembedded.dev * r2dfc5b7c9e 10openembedded.git/recipes/inkscape/inkscape_0.48.0.bb: inkscape: add 0.48 Nov 15 16:03:38 pb_: empty var wouldn't point to sysroot (if I fix the syntax to use / everytime it's empty == micro) Nov 15 16:04:04 pb_: because it would look like ="?/ at least Nov 15 16:04:37 should really audit the binconfig usage and just kill them all, and patch everything that deps on them to use pkgconfig, or something -- maybe that would be a good task for the janitors page Nov 15 16:04:43 heh Nov 15 16:04:59 pb_: and if I match for / or eol, then with empty prefix I'll get exactly that, empty var pointing to sysroot Nov 15 16:05:06 ERROR: 3:SyntaxError:' SLASH_PREFIX ?= "/"' Nov 15 16:05:12 with micro Nov 15 16:05:16 GNUtoo|laptop: like i said, that's not valid python :) Nov 15 16:05:20 ?= is bitbake Nov 15 16:05:22 ah ok Nov 15 16:05:25 indeed Nov 15 16:05:27 sorry Nov 15 16:05:29 :) Nov 15 16:05:47 that was just idea explanation (I didn't expect someone to try it now) :) Nov 15 16:06:34 * kergoth_ finally decided to upgrade his workstation that's mostly from 2006, time to get a quick personal development machine rather than doing nearly all of his oe work on his work laptop Nov 15 16:06:49 JaMa: well, yeah, I don't think there is much you can do about random absolute paths in general. no good way to tell whether they are meant to be prefixed with sysroot or not. Nov 15 16:07:27 or, more accurately, no good way to tell whether a given var is actually a filename (which needs prefixing) or just a text string (which doesn't) Nov 15 16:07:30 JaMa, ok converted to python and trying Nov 15 16:07:42 I think kergoth's plan to just have a big bonfire of all the binconfigs is probably the best one. Nov 15 16:08:02 that's a lovely image. i'll bring the marshmellows Nov 15 16:08:26 failing that, I think you just need to find the minimal hack that works for the case at hand and not worry too much about what other undiscovered horrors might or might not be in other packages. Nov 15 16:08:35 pb_: yes I can find some example where it was prefixed where shouldn't (and I had to strip prefix in recipe to fix it) Nov 15 16:09:08 if it weren't for the fact that binconfig is legacy cruft, it would be worth spending some time to fix it properly. but it is, and hence it isn't. Nov 15 16:09:37 * JaMa consider SLASH_PREFIX as minimal hack (being as much backward compatible as possible) Nov 15 16:10:07 03Alex Ferguson  07org.openembedded.dev * r624ddf456a 10openembedded.git/recipes/gmu/gmu_0.7.1.bb: (log message trimmed) Nov 15 16:10:08 gmu-0.7.1: Move RDEPENDS to DEPENDS. Nov 15 16:10:08 Merged RDEPENDS into DEPENDS, because there is no practical difference Nov 15 16:10:08 (I believed there was) and bitbake fails to build gmu under certain Nov 15 16:10:08 conditions (such as building with BB_NUMBER_THREADS set) due to Nov 15 16:10:08 half-built dependencies. Nov 15 16:10:09 Signed-off-by: Alex Ferguson Nov 15 16:10:18 03Alex Ferguson  07org.openembedded.dev * r14db95b844 10openembedded.git/recipes/greq/ (3 files in 2 dirs): (log message trimmed) Nov 15 16:10:18 greq-0.9.4: Add greq, a program for building gtk+-1.2 dialogs from scripts. Nov 15 16:10:18 Added tab-delimiter.patch, which changes the delimiter between multiple Nov 15 16:10:18 return arguments from \v to \t because busybox ash doesn't support arrays Nov 15 16:10:19 and accessing individual arguments with cut is easier when they're tab Nov 15 16:10:19 delimited. Nov 15 16:10:20 Added wmclass-option.patch which adds an option for setting the window's Nov 15 16:10:20 03Alex Ferguson  07org.openembedded.dev * rffdc040b8a 10openembedded.git/recipes/gtk+/gtk+-1.2-1.2.10/small-filesel.patch: (log message trimmed) Nov 15 16:10:21 gtk+-1.2: Modify small-filesel.patch for better usability. Nov 15 16:10:21 Modified small-filesel.patch, which now omits the "Rename", Nov 15 16:10:22 "Delete" and "Create Directory" buttons and adds the file Nov 15 16:10:22 selection text area to the bottom. The previous arrangement Nov 15 16:10:23 didn't work so well for saving files, just opening them. Nov 15 16:10:23 Signed-off-by: Alex Ferguson Nov 15 16:10:24 03Alex Ferguson  07org.openembedded.dev * r8ff50704be 10openembedded.git/recipes/dzen2/ (dzen2_svn.bb files/Makefile files/dzen-extras.tar.gz): (log message trimmed) Nov 15 16:10:25 dzen2: Modularize dzen-launcher, update recipe. Nov 15 16:10:36 added greq which is somewhat like xdialog, and which is needed for Nov 15 16:10:36 some dialogs. Nov 15 16:10:37 Signed-off-by: Alex Ferguson Nov 15 16:10:37 Signed-off-by: Kristoffer Ericson Nov 15 16:10:38 03Alex Ferguson  07org.openembedded.dev * r57a6f591f8 10openembedded.git/recipes/jlime/ (3 files): Nov 15 16:10:40 heh "Merged RDEPENDS into DEPENDS, because there is no practical difference" Nov 15 16:10:46 jlime-extras: Add stub jlime-extras-machine recipes. Nov 15 16:10:46 Added stub jlime-extras packages for jornada6xx, jornada7xx and Nov 15 16:10:46 mp900c, which are needed by the jlime-extras recipe when building Nov 15 16:10:46 for the respective MACHINEs. Nov 15 16:10:46 Signed-off-by: Alex Ferguson Nov 15 16:10:46 Signed-off-by: Kristoffer Ericson Nov 15 16:10:54 eh, there is a practical difference, they likely just misunderstood what the difference is Nov 15 16:11:06 heh Nov 15 16:11:22 to be fair, i don't think the bitbake manual covers it well, not sure about the oe one Nov 15 16:11:34 Alex Ferguson ping Nov 15 16:12:20 kergoth: heh Nov 15 16:12:44 but looking on that commit, the message is wrong, but the result is better than it was Nov 15 16:12:46 even people who've used oe a while seem to occasionally misunderstand it Nov 15 16:34:40 03Chris Larson  07master * rb35109935b 10openembedded.git/recipes/tasks/ (task-sdk-native.bb task-sdk-native.inc): Nov 15 16:34:40 task-sdk-native: inherit task, drop DEPENDS Nov 15 16:34:40 Signed-off-by: Chris Larson Nov 15 17:03:41 re Nov 15 17:03:46 funkt wieder Nov 15 17:04:09 sorry wrong screen Nov 15 17:11:39 kergoth_, 'the virtclass "magic" in the native unpackers commit' was (IIRC) that otherwise we'd end up in a dep-loop if you choose xz_git Nov 15 17:18:33 blindvt: i'd say we should make the logic smarter then Nov 15 17:18:49 instead of just going is the string 'xz' in SRC_URI, maybe we should split and check for extension specifically Nov 15 17:18:55 perhaps Nov 15 17:19:00 I submitted a patch http://patchwork.openembedded.org/patch/3538/ Nov 15 17:19:20 are there plans to integrate the patch to add ext4 IMAGE_ROOTFS support into master? http://patchwork.openembedded.org/patch/2592/ Nov 15 17:19:28 then splitted it into two patches, as Michael Smith recommended Nov 15 17:20:14 * kergoth_ ponders Nov 15 17:20:31 but he said, I need "please use INC_PR. e.g. set INC_PR = "r0" in shorewall.inc and set PR = "${INC_PR}.0" in the recipes which include it. " for shorewall patch... I don't undestand where to write PR = "${INC_PR}.0" Nov 15 17:21:23 after "require shorewall.inc" ? Nov 15 17:22:37 yes Nov 15 17:22:39 mrj10, ugh. I promised to add the required fsck to make it a real ext4 image. it's lying around somewhere in my tree, not sure if i sent the updated patch to the list yet.. sorry Nov 15 17:23:29 not a problem, was just curious. i may splat the patch in locally for the time being (and add the fsck suggested) Nov 15 17:23:32 mrj10, but (apart from some noise that prevents it to apply cleanly nowadays) it's straight forward. Feel free to redo it and submit it instead of that 2592 Nov 15 17:24:03 ok, i will do so if i get around to testing an image on my board Nov 15 17:24:44 kergoth_, well.. http://paste.debian.net/99894/ Nov 15 17:24:50 sorry for stupid question :) I did it Nov 15 17:25:49 kergoth_, i admit that the INHIBIT_AUTOTOOLS_DEPS may have been a quick-hack but iirc i tried to solve it in a different (also clean, IMHO) way and didn't quite see how Nov 15 17:27:13 kergoth_, but it would be interesting to see what somebody that knows oe (as opposed to /me ;) would do to avoid such a construct :) Nov 15 17:28:45 kergoth_, perhaps you can imagine an elegant way ? Nov 15 17:35:11 kergoth_, please note that "xz-0.8.15" does _not_ match as extension (we're looking at the uri which is .bz2 for that xz release tarball). IIRC the loop was this: virtual/coreutils-native (coreutils|busybox; both as .bz2) The former uses autotools, to unpack autotools we'd need an unpacker and that unpacker itself needs autotools -> kaboom() (or something along those lines) Nov 15 17:36:24 ah, yes, avoiding autotools deps for -native unpackers is reasonable, and should be something that can be done fine in the recipes -- its the class changes that struck me as odd Nov 15 17:40:33 kergoth_, i don't understand; which class changes do you mean? Nov 15 17:42:56 nevermind, if you're confused, i'm probably just remembering wrong, don't have the patch in front of me. i'll apply it to a tree and poke around later Nov 15 17:50:37 qwt and/or qt does not appear to build with gcc 4.5 Nov 15 17:54:49 khem, ping? Nov 15 17:56:14 blindvt: i was remembering right -- i was wondering if we can manage it without the changes to autotools.bbclass and base.bbclass. it doesn't really belong there, unless there's really no alternative. i'll play with it and see if i can come up with something Nov 15 17:56:20 khem, did you see that odd strong alias breakage w/ gcc trunk and did you, by chance, track it down already or should i have a look? Nov 15 17:56:37 Crofton: qt-x11-free-4.7 builds here fine with gcc Nov 15 17:58:02 https://gist.github.com/700677 - sheesh, getting quite the collection of un-pushed topic branches Nov 15 17:58:12 khem, two of the symptoms are that fcntl and atoi are non-existing; seen on micro-base-image for qemux86 with current uClibc master Nov 15 17:58:38 dv: is FILES_${PN} += "/usr/share/shorewall/*" really needed? the same in shorewall-shell Nov 15 17:59:16 JaMa, what version gcc Nov 15 17:59:29 there is ${datadir}/${PN} in FILES_${PN} in bitbake.conf Nov 15 17:59:31 what about qwt? Nov 15 17:59:32 Crofton: 4.5 Nov 15 17:59:35 k Nov 15 17:59:55 I can try qwt.. Nov 15 18:00:50 kergoth_, cool TIA! Nov 15 18:06:04 JaMa, needed. look, there are two variants of shorewall. one is -shell, second is just shorewall. both of them need /usr/share/shorewall/*, but first one needs /usr/share/shorewall-shell/* also Nov 15 18:06:56 JaMa, I don't know why, but they was not included in my image without such a FILES_${PN}... Nov 15 18:08:33 * kergoth_ kicks off a bitbake xz Nov 15 18:09:42 dv: ah ok, I didn't notice that shorewall-shell is using same .inc, but still strange shorewall-shell needs /usr/share/shorewall-shell/* Nov 15 18:09:59 it shouldn't, yeah. also should be using ${datadir} Nov 15 18:10:01 hrm Nov 15 18:10:06 dv: and better to use {datadir} then /usr/share Nov 15 18:10:12 :) Nov 15 18:11:07 JaMa, yes. It uses different parser, components of which is installed into different location together with common compopnentss in default localtion... about {datadir}... yes, you're totally right :) Nov 15 18:12:17 JaMa, do I have to change my patch ? Nov 15 18:17:25 dv: I mean keep {datadir}/shorewall/ in .inc file, but don't add {datadir}/shorewall-shell in shorewall-shell*bb as it should be included automatically Nov 15 18:17:32 dv: and please update that patch Nov 15 18:18:31 Crofton: qwt fails here too /OE/tmpdir-shr/sysroots/armv4t-oe-linux-gnueabi/usr/include/qt4/QtCore/qstring.h:187:17: note: the mangling of 'va_list' has changed in GCC 4.4 Nov 15 18:20:26 qwt_abstract_scale_draw.cpp:406:1: fatal error: error writing to -: Broken pipe Nov 15 18:27:07 i think amazon has spoiled me. now i get annoyed when some online store doesn't ship the same day Nov 15 18:27:34 or newegg and their $3 extortion? Nov 15 18:27:48 heh, indeed Nov 15 18:28:07 * kergoth_ is talking about newegg actually, no way he's paying that, bastards Nov 15 18:28:13 * kergoth_ shakes fist Nov 15 18:29:59 I updated my receipes :) sorry for unconvinience - I'm a beginner Nov 15 18:31:31 no worries, thanks for contributing, we always (or *should* always) appreciate newcomers Nov 15 18:32:49 03Dvorkin Dmitry  07master * rb79c5f4e3f 10openembedded.git/recipes/pstree/pstree_2.33.bb: Nov 15 18:32:49 pstree: added recipe Nov 15 18:32:49 Signed-off-by: Dvorkin Dmitry Nov 15 18:32:49 Acked-by: Michael Smith Nov 15 18:33:00 03Dvorkin Dmitry  07master * r3e3253ae59 10openembedded.git/recipes/shorewall/ (4 files): Nov 15 18:33:00 shorewall: updated to 4.2-shell and 4.4-perl versions Nov 15 18:33:00 Signed-off-by: Dvorkin Dmitry Nov 15 18:33:00 Acked-by: Michael Smith Nov 15 18:34:41 03Michael Smith  07master * r50d3e0bc5d 10openembedded.git/recipes/ (pstree/pstree_2.33.bb shorewall/shorewall.inc): Nov 15 18:34:41 pstree/shorewall: fix AUTHOR Nov 15 18:34:41 I pushed the wrong version of the patch from patchwork. Nov 15 18:34:41 Signed-off-by: Michael Smith Nov 15 18:36:26 ups Michael pushed older version with /usr/share/shorewall-shell/ :/ Nov 15 18:36:58 heh, no big deal to push the fix in a subsequent commit Nov 15 18:37:17 I'm working for 14 hours and tired too :) Nov 15 18:37:26 * kergoth_ should spend more quality time on the ${B} vs ${S} branch Nov 15 18:37:27 is the current details on LICENSE documented on wiki yet? Nov 15 18:38:07 JaMa, same here, thansk Nov 15 18:38:14 Crofton: yw Nov 15 18:42:16 khem, I didn't quite understand your description of LINCENSE for the recipes I sent to maillist: here is iftop licnese http://pastebin.com/8S0YCMzp - are you saying that should be GPLv2 or GPLv2+ ? Nov 15 19:26:16 03Alex Ferguson  07org.openembedded.dev * rb9ea13f7bd 10openembedded.git/recipes/jlime/fileselector_1.0.bb: (log message trimmed) Nov 15 19:26:16 fileselector: Bumped PV from 1.0 to 1.1. Nov 15 19:26:16 Updated upstream source file with some bug-fixes along with a small Nov 15 19:26:16 program called 'loading' which simply exists to display a black screen, Nov 15 19:26:16 providing a stable background in scripts which use fileselector. Crappy, Nov 15 19:26:17 I know, but it only uses xlib and is lightweight so why not. Nov 15 19:26:17 Signed-off-by: Alex Ferguson Nov 15 19:26:26 03Alex Ferguson  07org.openembedded.dev * r930c7c8a53 10openembedded.git/recipes/jlime/jlime-extras_1.0.1.bb: Nov 15 19:26:26 jlime-extras: Modify checksums for upstream source file change. Nov 15 19:26:26 This one is a quick modification of the checksums to mirror a Nov 15 19:26:26 small change in the upstream source file. Nov 15 19:26:26 Signed-off-by: Alex Ferguson Nov 15 19:26:26 Signed-off-by: Kristoffer Ericson Nov 15 19:26:28 03Alex Ferguson  07org.openembedded.dev * r95a71d22c8 10openembedded.git/recipes/dzen2/ (dzen2_svn.bb files/Makefile files/dzen-extras.tar.gz): (log message trimmed) Nov 15 19:26:28 dzen2: Modify dzen-extras, update Makefile and recipe. Nov 15 19:26:28 Modularized the battery applet by moving the charge calculation Nov 15 19:26:28 in an external script and adding per-platform functions. Made the Nov 15 19:26:29 process a bit more efficient by adding a small program written in Nov 15 19:27:13 C, dzen-battery, which calculates a battery percentage given two Nov 15 19:27:13 standard files in /sys/class/power_supply and by moving processes Nov 15 19:34:54 03Michael Smith  07master * re384e345cd 10openembedded.git/recipes/python/python-pyopengl-accelerate_3.0.1.bb: Nov 15 19:34:54 python-pyopengl-accelerate: add Nov 15 19:34:54 Signed-off-by: Bram Cymet Nov 15 19:34:54 Signed-off-by: Michael Smith Nov 15 19:35:04 03Michael Smith  07master * r8ded640ef1 10openembedded.git/recipes/python/python-pyopengl_3.0.1.bb: Nov 15 19:35:04 python-pyopengl: add Nov 15 19:35:04 Signed-off-by: Bram Cymet Nov 15 19:35:04 Signed-off-by: Michael Smith Nov 15 19:42:15 03Michael Smith  07master * rd9997db446 10openembedded.git/recipes/xorg-driver/ (2 files in 2 dirs): Nov 15 19:42:15 xf86-input-elographics: allow limiting the touchable area of the display Nov 15 19:42:15 Add ScreenWidth and ScreenHeight configuration options to override Nov 15 19:42:15 the screen size. Nov 15 19:42:15 see also: https://bugs.freedesktop.org/show_bug.cgi?id=31397 Nov 15 19:42:16 Signed-off-by: Michael Smith Nov 15 19:48:35 tharvey: the statement here Nov 15 19:48:39 tharvey: the statement here "# Nov 15 19:48:39 either version 2 of the License, or Nov 15 19:48:41 # (at your option) any later version." Nov 15 19:48:54 means that its GPLv2+ Nov 15 19:48:58 good morning all Nov 15 19:49:34 blindvt`: no I did not investigate yet the breakage due to aliases Nov 15 19:49:46 blindvt`: was occupied heavily last week Nov 15 19:52:57 khem, ok I'll resubmit as 'GPLv2' Nov 15 19:55:29 hi khem Nov 15 19:57:13 blindvt`: xz-native and xz both built fine with your patch minus the class changes, here anyway Nov 15 20:06:04 morning again Nov 15 20:06:18 kergoth_: have you seen my race report in ML? Nov 15 20:06:55 GPLv2+ Nov 15 20:10:27 silly question (i asked on #angstrom, but that channel appears to be mostly dead): anyone know what's happening when you boot an angstrom image (i assume other distros do this as well but haven't checked) for the first time and it says "Configuring x.. Configuring y.. Configuring z.." for a good while? Nov 15 20:10:41 init scripts Nov 15 20:10:48 its running the postinst scripts for the packages that require they be run on the target Nov 15 20:11:00 what kergoth said Nov 15 20:11:08 the scripts are in the opk Nov 15 20:12:42 ok cool, thanks guys Nov 15 20:20:28 hi Nov 15 20:20:37 I just git pulled Nov 15 20:21:16 building angstrom minimal-image I face to a scary error "git was compiled without libcurl support" Nov 15 20:21:59 I always considered angstrom the distro taiolred on OE (or vice versa) Nov 15 20:22:21 what's happening to OE? Nov 15 20:23:41 03Khem Raj  07master * r7b4734daf2 10openembedded.git/recipes/binutils/ (15 files in 2 dirs): Nov 15 20:23:41 binutils: Use git instead of cvs for SRC_URI for upstream version of binutils Nov 15 20:23:41 * Use the upcoming 2.21 branch as SRC_URI so that we can test it Nov 15 20:23:41 and provide feedback to upstream based on our testing Nov 15 20:23:41 use PREFERRED_VERSION_binutils-cross = "git" to select this Nov 15 20:23:41 and test it in builds and reports issues. Nov 15 20:23:42 Signed-off-by: Khem Raj Nov 15 20:24:33 mckoan: was it talking about the git on your host machine, or the git it's building for the target as part of the image? (i don't know offhand if angstrom-minimal-image includes git) Nov 15 20:25:37 mrj10: it must be git on host machine, but is a Debian Lenny, so I wonder what could be wron there Nov 15 20:25:51 that seems like a legitimate problem Nov 15 20:25:53 mckoan: well, I think you are using git-native Nov 15 20:26:17 the bandaid solution is to make sure you dont try to fetch any git repositories using the http protocol, use the git protocol instead Nov 15 20:26:32 but it seems like the easier way is to get a version of git that does have libcurl support Nov 15 20:26:45 khem: what do you mean with "using git-native" ? Nov 15 20:27:34 mckoan: the problem is happening on target or on the build machine ? Nov 15 20:27:53 I changed nothing, I simply pulled, and it worked until 2nd november Nov 15 20:28:57 I am assuming its the git-native thats being used. In this case you should look into git-native build logs specifically into log.do_configure and see why it disables libcurl support because as far as I see there is no specific construct to disable it in the recipe if it finds it then it should have enabled it Nov 15 20:29:16 hmmm Nov 15 20:29:40 could be that libcurl native has issues Nov 15 20:29:58 cant say much without looking into the config.log for git-native Nov 15 20:30:31 * mckoan is enabling tinderbox Nov 15 20:31:14 mckoan: I dont think tinderbox has logs for the successful builds Nov 15 20:32:12 also tinderbox is not working Nov 15 20:32:17 angstrom-2008.1 console-image is built fine for my targets Nov 15 20:32:30 with exception of that race Nov 15 20:32:38 Jay7: are you using debian? Nov 15 20:32:49 mckoan: debian testing Nov 15 20:32:50 DISTRO = "angstrom-2008.1" Nov 15 20:32:57 x86_64 Nov 15 20:33:00 ah.. Nov 15 20:33:05 but this is testing-next branch Nov 15 20:33:35 was forked 2010-11-12 iirc Nov 15 20:33:52 s/forked/merged/ Nov 15 20:34:09 or even updated from master Nov 15 20:34:55 tinderbox is half-life.. Nov 15 20:38:00 maybe I missed the new goals of OE, I have to reconsider it in a new perspective Nov 15 20:38:07 mine is reliability Nov 15 20:39:13 mckoan: here are my builds: http://tinderbox.openembedded.net/builders/Jay7-tb/ Nov 15 20:39:44 Jay7: thank you Nov 15 20:39:45 but almost all are on testing-next this time Nov 15 20:40:04 khem: thx Nov 15 20:40:12 I'm testing angstrom-2008.1 for lot of machines with console-image, x11-image and opie-image Nov 15 20:40:44 Jay7: but looks like the real problem is Debian Lenny x86_64 Nov 15 20:41:06 mckoan: I've updated from lenny about a week Nov 15 20:41:39 that was because I was needed bitbake master to test too Nov 15 20:41:49 but lenny have only python 2.5 Nov 15 20:44:19 anyway, you can't say 'reliability' about OE master :) Nov 15 20:45:38 Nov 15 20:46:00 Jay7: You're right ;-) Nov 15 20:46:13 but we are on the way to release Nov 15 20:53:49 shouldnt MACHINE_RDPENDS = "wpa-supplicant" include that package in all images built? Or am I missing something? Nov 15 20:54:01 MACHINE_EXTRA_RDPENDS Nov 15 21:08:31 03Martin Jansa  07master * ra02d3d0eef 10openembedded.git/recipes/xorg-driver/ (2 files in 2 dirs): Nov 15 21:08:31 xf86-video-glamo: fix stack corruption from overflowing cmdq Nov 15 21:08:31 Signed-off-by: Martin Jansa Nov 15 21:12:15 btw Nov 15 21:12:33 khem: did you still care about efikamx? Nov 15 21:12:48 my builds failed on kernel unpacking Nov 15 21:13:15 khem: http://tinderbox.openembedded.net/packages/977058/ Nov 15 21:14:47 btw, oestats doesn't show this as a failed build.. Nov 15 21:23:29 03Sebastian Krzyszkowiak  07master * r08b1bf363f 10openembedded.git/recipes/freesmartphone/cornucopia.inc: Nov 15 21:23:29 cornucopia: bump SRCREV Nov 15 21:23:29 Signed-off-by: Martin Jansa Nov 15 21:54:36 hi all Nov 15 21:54:50 i'm trying to get a recipe going for the newest version of valgrind, which supports armv7+ Nov 15 21:55:11 there's a COMPATIBLE_HOST line in the existing valgrind recipes that masks out all ARM hosts Nov 15 21:55:27 i modified it in my recipe to allow arm angstrom hosts (will make more robust later) Nov 15 21:55:35 i think the COMPATIBLE_HOST line should be satisfied now Nov 15 21:55:53 how can i get bitbake to ignore the fact that its cache says to skip that recipe? Nov 15 21:56:04 i tried 'touch'ing local.conf and rebuilding, but that didnt seem to do it Nov 15 22:05:29 03Martin Jansa  07master * r664ad859bc 10openembedded.git/recipes/orrery/ (files/orrery.png orrery/use.GdkPixbuf.patch orrery_2.7.bb): Nov 15 22:05:29 orrery: use GdkPixbuf for more effective drawing Nov 15 22:05:29 * also s/files/orrery/ for faster lookup Nov 15 22:05:29 Signed-off-by: Martin Jansa Nov 15 22:10:05 anyone know why eglibc-dbg has INSANE_SKIP on it? Nov 15 22:10:07 Jay7: yes I do Nov 15 22:10:22 i'd like to be able to build a debug version of either glibc or eglibc for angstrom, to use with valgrind Nov 15 22:10:53 mrj10: how is INSANE_SKIP barring u from building debug version Nov 15 22:11:18 well, i will plead ignorance here. the symptom of my problem is: Nov 15 22:11:25 bitbake eglibc-dbg Nov 15 22:11:26 ... Nov 15 22:11:30 no Nov 15 22:11:42 ERROR: Nothing PROVIDES 'eglibc-dbg' Nov 15 22:11:45 its already emitted while bitbaking eglibc Nov 15 22:11:54 there is no separate target called eglibc-dbg Nov 15 22:12:06 the error says it doesnt it Nov 15 22:12:51 not that i can tell http://pastebin.com/yKPvcuYS Nov 15 22:13:10 maybe my rootfs has it already and i just need to figure out how to make valgrind use those versions of eglibc libraries Nov 15 22:13:42 what does it expenct libraries with profiling info ? Nov 15 22:14:05 i tried using it and got: http://pastebin.com/7KGnFWfU Nov 15 22:14:25 it wants symbols exported so it can replace malloc, memcpy, etc. with its own versions Nov 15 22:16:41 so i guess it has to do with ld-linux.so, i'm not sure how that library gets built yet Nov 15 22:19:09 mrj10: ok did you install the -dbg package into the RFS Nov 15 22:19:17 mrj10: there are separate ipks Nov 15 22:19:27 for debug info of a package Nov 15 22:19:35 they may not be installed by default Nov 15 22:19:39 you have to install it Nov 15 22:20:00 with opkg on device or include it offline during build Nov 15 22:20:09 i'm downloading an ipk now from angstrom-distribution.org, might look around for a built ipk in my deploy dir Nov 15 22:20:41 should ld-linux.so be in the eglibc ipk's Nov 15 22:20:42 ? Nov 15 22:24:38 i tried scp'ing over the ipk and installing it on the target Nov 15 22:24:44 and i get: Nov 15 22:24:47 Collected errors: Nov 15 22:25:11 * pkg_init_from_file: Malformed package file eglibc-dbg_2.11-r15.8+svnr10690.9_armv7a.ipk Nov 15 22:33:58 mrj10, can you extract the control file from this .ipk and put it in a pastebin? Nov 15 22:34:42 e.g. ar x eglibc-dbg_2.11-r15.8+svnr10690.9_armv7a.ipk; tar xf control.tar.gz Nov 15 22:36:38 grg: sure, i'll try that. maybe one more thing first. i scp'ed the file again. i didnt check the md5sums before, but they match now Nov 15 22:36:54 and i tried opkg installing again, and it gave me an md5sum mismatch with whatever opkg has in its cache Nov 15 22:37:03 where's that cache stored. i'll try to override that manually and install it Nov 15 22:37:10 mrj10, hmm ok, that's probably more of a problem Nov 15 22:37:27 /var/lib/opkg maybe... i forget Nov 15 22:37:38 yep, thanks Nov 15 22:40:11 03Philip Balister  07org.openembedded.dev * rd8d33031dc 10openembedded.git/recipes/tasks/task-sdk-gnuradio-native.bb: task-sdk-gnuradio-native.bb : Bump PR so builds pick up last change. Nov 15 22:43:27 vmware are using ipkg... wtf? Nov 15 22:43:43 ? Nov 15 22:44:19 and it looks like they have a patch fragment that is not in opkg Nov 15 22:44:40 (but they missed a bazillion other bugs) Nov 15 22:54:08 Hi all, quick question: I have in my Makefile.am files paths to directories such as: "AM_CXXFLAGS = -Wall -I/usr/include" and "LDADD = /usr/lib/libname.la". I think I need to include a symbol from my recipe so that I can point these to the appropriate fs staging dir "${STAGING_DIR_TARGET}/usr/lib" for example. Not sure how to do this though... Nov 15 22:55:38 ouch Nov 15 22:56:14 -I/usr/include should not be needed, it shows be default for sane cases Nov 15 22:56:28 unless standard includes are disabled Nov 15 22:56:42 not sure about the LDADD Nov 15 22:56:45 sure, but also have subdirs of include that I need to point at Nov 15 22:56:48 maybe use a relative path Nov 15 22:56:57 relative paths ... Nov 15 22:57:03 is what I would try Nov 15 22:58:36 just -lname, there's no need to point at the .la directly, ever, as libtool will handle that for you Nov 15 22:59:19 well, relative to what? If I start with -I./ I'll be referring to my Work dir, no? Nov 15 22:59:56 what exactly are you trying to do? Nov 15 23:00:07 you never have to point into staging yourself, oe passes the necessary args to the compiler to do that for you Nov 15 23:00:16 just drop the -I, use -l for the lib, and you'll be fine Nov 15 23:00:47 trying to include headers from /usr/include/subdirofinclude Nov 15 23:01:33 normally, one would use a binconfig script or pkg-config to get those sorts of arguments automatically Nov 15 23:03:54 Thx, I'll look into it (as is probably very obvious) i'm quite the n00b when it comes to autotools and linux toolchains in general... Nov 15 23:04:05 khem; 35881fcc6feb9670947056ff868e03b0a0900a51 busted sh4 ;9 Nov 15 23:04:09 broken shift Nov 15 23:04:17 9in my kbd atm0 Nov 15 23:05:04 sounds like my problem lies in non-standard practices in the existing build system, and I should correct that rather than apply hacks Nov 15 23:05:26 :) Nov 15 23:05:34 I think there are ways to make it work Nov 15 23:06:34 but I never figured them out, just split things in to seperate autotools packages so I could do the library builds after they were installed Nov 15 23:06:42 mekoniak: you can always patch out the -I@includedir@/subdir and CFLAGS += "-I${STAGING_INCDIR}/subdir" in the recipe Nov 15 23:06:52 that way upstream still works, and so does OE Nov 15 23:09:40 Crofton: FYI, I'm building each module with its own recipe now, as you suggested. Have system ossie working, working on recipes for components now... Nov 15 23:09:52 ah Nov 15 23:09:55 Tartarus: yes it was busted before Nov 15 23:09:59 Tartarus: for gcc 4.5 Nov 15 23:10:08 that should cut the need for the funky flags Nov 15 23:10:13 Tartarus: I realised that I fixed one problem but created another one Nov 15 23:10:37 * Crofton thinks Tartarus has been drinking Nov 15 23:10:45 Tartarus: one fine day I need to take few quite hours and hack gcc to fix it for good rather than beating the bush Nov 15 23:11:26 03Chris Larson  07master * rfd2d95283d 10openembedded.git/classes/checkbashisms.bbclass: Nov 15 23:11:26 checkbashisms: add initial class to run against ${S} Nov 15 23:11:26 Currently requires that you already have checkbashisms available somewhere in Nov 15 23:11:26 your PATH. Runs against all '#!/bin/sh' scripts in ${S}, to attempt to find Nov 15 23:11:26 and fix issues with /bin/sh not being bash. Nov 15 23:11:26 Signed-off-by: Chris Larson Nov 15 23:12:19 kergoth_: re. bashism did you add a friendly message if someon inherits it and dont have installed checkbashisms Nov 15 23:12:37 I really did not look into your patch so find it out for myself Nov 15 23:12:40 inheriting won't do anything, its its own task Nov 15 23:12:52 but no, not a friendly one, yet Nov 15 23:13:02 -c checkbashisms / -c checkbashisms_all Nov 15 23:13:07 kergoth_: ok may be a sanity message Nov 15 23:13:12 * kergoth_ nods Nov 15 23:13:14 Crofton: I guess the reason these are hard coded in the existing Makefile.am is that they were using relative paths to point at the compiled-but-not-installed libraries and headers in the source dirs. Since they're now installed to the target FS, I should be able to use a more standard convention Nov 15 23:13:29 mekoniak, exactly Nov 15 23:13:57 thanks Crofton and kergoth. Backed me out of my corner :-) Nov 15 23:14:04 khem: So... I'm a little puzzled :) Nov 15 23:14:05 mekoniak: now we do a proper install into staging so yes its better Nov 15 23:14:09 khem: sh4 is broken atm Nov 15 23:14:32 Tartarus: I think I should revert that fix Nov 15 23:14:43 khem: And lock down to 4.4.x if needed, yeah Nov 15 23:14:44 so atleast 4.4 will work Nov 15 23:14:49 or go read gnuradio autofoo and learn to build against libraries before they are installed Nov 15 23:14:51 righto Nov 15 23:14:57 although that is some tough reading :) Nov 15 23:15:04 Tartarus: 4.5 has a problem in multilibs Nov 15 23:15:12 and its a hairy problem in gcc Nov 15 23:15:34 03Chris Larson  07master * r29655d79ed 10openembedded.git/classes/checkbashisms.bbclass: Nov 15 23:15:35 checkbashisms: add nicer error when checkbashisms isn't available Nov 15 23:15:35 Signed-off-by: Chris Larson Nov 15 23:15:35 I have a bug open for it but no one looked into it in gcc community which means I should do it myself Nov 15 23:16:07 Tartarus: for the release sane toolchains I plan to lock 4.4 for sh4 Nov 15 23:16:53 indeed. GNUradio is next on my list... Is it using UHD drivers now? I think that's required for USRP-e, right? Nov 15 23:17:03 want me to push that now khem? Nov 15 23:17:06 right Nov 15 23:17:06 got it queued now Nov 15 23:17:27 oe builds uhd and gnuradio that uses uhd Nov 15 23:17:36 and installs usrp firmware Nov 15 23:18:37 good to know Nov 15 23:18:52 thanks again. Nov 15 23:18:59 I haven't tested it on anything puublically available :) Nov 15 23:19:51 doesn't UHD also work with USRP1? Nov 15 23:19:57 Tartarus: I can reverse it moment Nov 15 23:20:08 it should Nov 15 23:20:18 that is the untested piece Nov 15 23:21:02 Crofton: I was testing GNU Radio from angstrom-distro repo on BB with USRP1 about 2 weeks ago. I assume those changes are newer? Nov 15 23:21:47 not sure Nov 15 23:21:58 pretty close though Nov 15 23:22:08 git log gnuradio_git.bb :) Nov 15 23:23:03 03Khem Raj  07master * rd827b36a57 10openembedded.git/recipes/gcc/gcc-cross4.inc: (log message trimmed) Nov 15 23:23:03 Revert "gcc-cross4.inc: We dont support multilib so empty out the with-libdir-list for SH4" Nov 15 23:23:03 This fixed building the toolchain though but broke Nov 15 23:23:03 kernel as it expect gcc to be able to compile no-fpu Nov 15 23:23:03 code. Which means it will break all gcc 4.x Nov 15 23:23:04 So we revert it and keep the status quo of 4.5 being broken Nov 15 23:23:05 but older 4.x gcc can still build for SH based machines. Nov 15 23:23:06 either way, I'll be doing some more BB/USRP/GNUradio testing later in the week. I'll let you know what I get. Nov 15 23:23:17 dinner time now, though :-) Nov 15 23:23:27 I'm in here and #gnuradio Nov 15 23:23:33 heh Nov 15 23:23:39 I am in CA Nov 15 23:23:42 so mid afternoon Nov 15 23:24:20 Tartarus: and this one http://pastebin.com/e0PkebTb Nov 15 23:24:34 should pick 4.4 for sh3/sh4 Nov 15 23:24:37 we should be set Nov 15 23:25:12 Crofton: did you get over the last issue about that gnome thingy Nov 15 23:25:19 well Nov 15 23:25:24 I commented it out :) Nov 15 23:25:26 for now Nov 15 23:25:40 Crofton: ok. I could not reproduce it Nov 15 23:25:47 otherwise I could have fixed it Nov 15 23:25:53 thanks Nov 15 23:26:00 I need to check with koen Nov 15 23:26:05 whatever works for me is not a bug :) Nov 15 23:26:11 I am wondering if it it F13 specific Nov 15 23:26:14 :) Nov 15 23:26:44 could be. I dont have a F13 box here to try Nov 15 23:26:47 my ability to partitiion sd cards has gone i the toilet Nov 15 23:26:52 what to test tonight? :) Nov 15 23:27:02 Jay7: sh4 Nov 15 23:27:18 Tartarus: do u think I should commit the sane-toolchain change too ? Nov 15 23:27:21 hm.. which distro/machine/image? Nov 15 23:27:40 Jay7: qemush4/minimal/x11-image Nov 15 23:27:52 or angstrom or whatever I really dont care Nov 15 23:27:52 khem: oe master? Nov 15 23:27:55 yes Nov 15 23:28:09 Tartarus: btw. did ppc boot on qemu ? Nov 15 23:29:07 Cisco's reports were not that encouraging hmmm Nov 15 23:32:37 khem: build started, check results on tinderbox under 'Jay7' Nov 15 23:32:47 khem: no Nov 15 23:33:04 Tartarus: hmmm ok Nov 15 23:33:28 Tartarus: it dies in kernel right ? Nov 15 23:33:31 khem: and yeah, please update sane-toolchain Nov 15 23:33:38 khem: yeah qemuppc panics on mounting root Nov 15 23:33:41 Tartarus: ok Nov 15 23:33:46 Tartarus: ah I see Nov 15 23:33:55 Tartarus: then please choose -O2 Nov 15 23:34:06 as global optimization instead of Os Nov 15 23:34:09 and retry Nov 15 23:34:24 I thought it was something in kernel that broke Nov 15 23:34:33 ok Nov 15 23:34:48 Tartarus: there is a lingering issue with ppc and Os where it emits stuff into libgcc Nov 15 23:34:52 that it needs Nov 15 23:35:00 and that part is messed up Nov 15 23:35:06 with O2 its not needed Nov 15 23:35:10 and it should work Nov 15 23:36:16 03Khem Raj  07master * r448d705bee 10openembedded.git/conf/distro/include/sane-toolchain.inc: Nov 15 23:36:16 sane-toolchain.inc: Choose GCC 4.4.4 for SH bases machines Nov 15 23:36:17 Signed-off-by: Khem Raj Nov 15 23:36:22 Jay7: you have to update to latest and then restart the build plz Nov 15 23:36:35 Jay7: and yes use distro minimal or micro Nov 15 23:36:49 or any other which uses sane-toolchain Nov 15 23:37:02 khem: ok Nov 15 23:37:18 I'll build qemush4/minimal/x11-image Nov 15 23:37:27 woohoo, valgrind 3.6.0 is working on my beagleboard Nov 15 23:37:29 with changes above ^^ Nov 15 23:37:57 khem and grg: thanks for the help! Nov 15 23:38:30 * Jay7 is thinking about upgrading wife's workstation to phenom x4 or x6 too :) Nov 15 23:38:38 + distcc/icecc Nov 15 23:38:43 ended up scp'ing over the ipk file from my build machine, changing the md5sum opkg had from when i "opkg update"d, and installed it Nov 15 23:50:55 mrj10: What beagle board do you have? Nov 15 23:51:11 xM rev A2 Nov 15 23:51:28 mrj10: Hmm, nice. Nov 15 23:52:14 khem: build was restarted Nov 15 23:52:17 yeah, somebody did a google summer of code project to improve FFTW on Cortex-A8s like the beagleboard by adding neon intrinsics in places and putting in hooks to use ffmpeg's hand-optimized fft's as primitives Nov 15 23:52:44 after some work i got it all to build, but when i enabled the ffmpeg ffts, i got some double-free errors, and didn't have a good way to debug them Nov 15 23:52:52 thus, valgrind Nov 15 23:54:28 mrj10: Does it run remotely? Nov 15 23:54:41 no it runs on the target Nov 15 23:54:53 idk if valgrind can run remotely or not, never tried Nov 15 23:55:17 mrj10: OK, thanks. Nov 15 23:55:58 mrj10, you might ask in #beagle also Nov 15 23:58:03 Crofton: ask what? Nov 15 23:58:15 Crofton: oh, about the double frees? Nov 15 23:59:57 yeah Nov 16 00:00:53 yeah, i get the sense nobody has actually tried the guy's code. it doesn't build without some handholding, some of the features he implemented are turned off by default in his tree, etc. i commented on his final blog post with some suggestions and questions but no response yet (been a few weeks at this point) Nov 16 00:01:38 hmm Nov 16 00:01:41 try emailing him also Nov 16 00:02:03 SoC projects tend to result in starting points Nov 16 00:02:46 the big thing we know is adding the code to the existing fftw simd structure is not so good Nov 16 00:03:13 first boot on complex images takes forever Nov 16 00:07:36 Crofton: I'll try emailing, see if i'm missing something. Ditto on first boots, its a couple hours for me on this slow Class 2 microSD card Nov 16 00:08:02 I am working on a card with gnome Nov 16 00:08:18 fftw's codebase is like a fine watch, it's crazy to see how it all fits together but very hard to load it all into your working set and make meaningful changes Nov 16 00:08:25 yeah, gnome takes forever Nov 16 00:08:26 before we dupe it, I'll let firstboot run :) Nov 16 00:08:40 lack of public scm is annoying Nov 16 00:08:42 i just built and booted beagleboard-demo-image this afternoon, i feel your pain Nov 16 00:09:22 annoying, but typical for software produced by academics, they likely don't have time to contribute to it anyway so a fork with their blessing is probably the way forward if people are interested Nov 16 00:10:02 what are you going to do with fftw? Nov 16 00:11:08 it's a case study for a research project i'm doing. i want to suss out if there are meaningful ways in which one can compile (or optimize at runtime) for total energy consumption or instantaneous power consumption rather than performance Nov 16 00:11:58 the case study is to look at a bunch of fftw plans and see if there are plans which may be slower than the one FFTW currently settles on, but which provide a net energy win by making up for their performance with disproportionately lower power Nov 16 00:12:36 as an example, you could imagine that using NEON would result in a performance win, but marshalling data to and from a separate unit may be energy-inefficient Nov 16 00:12:50 cool Nov 16 00:12:54 :) Nov 16 00:13:42 my null hypothesis is that compiling for performance == compiling for low energy, since any power savings in the pipeline will be dwarfed by additional power you spend elsewhere in the system and you just want to get your work done as quickly as possible Nov 16 00:14:47 hopefully i'll be able to find some counterexamples. i think one area that is ripe for such counterexamples is trading off recomputing something vs. storing it and fetching it from memory. if you have to go all the way out to DRAM for something, there's a chance you're better off recomputing it instead in terms of energy Nov 16 00:15:43 the 2 big prerequisites to running the study are geting an autotuner like FFTW or SPIRAL running on the board, and sufficiently instrumenting the power rails and performance counters to be able to characterize the energy of each plan Nov 16 00:15:48 i'm getting there, slowly but surely :P Nov 16 00:16:19 let us know if you publish anything Nov 16 00:16:47 i will, i only wouldn't have otherwise because i didnt know if folks would be interested, this community and #beagle have been really helpful in getting this all rolling along Nov 16 00:18:21 Crofton: have you ever talked to jonpry in #beagle? Nov 16 00:18:30 not much Nov 16 00:18:56 ah. j/w if you knew about his project building a tablet w/ an OMAP3 and an FPGA on it for hifi audio stuff Nov 16 00:21:26 I know about ut Nov 16 00:21:54 that's all :P Nov 16 00:22:06 valgrind appears not to handle all the neon instructions Nov 16 00:22:22 going to look up what instruction 0xEE09BF5C corresponds to now Nov 16 00:24:09 * Crofton already has an omap3 attached to an fpga Nov 16 00:24:35 what interface are you using? Nov 16 00:25:02 * mrj10 knows he is derailing the channel, and will shush if requested Nov 16 00:34:46 I have two recipes with different names, that create the same packages Nov 16 00:35:06 Is this a virtul provider case? Nov 16 00:49:31 Crofton: that depends on what you're wanting to do. the only reason to use a PROVIDES is if you have two things that provide the same functionality, from the perspective of another recipe that needs to depend on it Nov 16 00:49:51 for runtime, if say, a task depends on one of those binary packages, bitbake will use your preferred providers to pick between the two recipes that emit that package Nov 16 00:50:22 I have a srctree recipe Nov 16 00:50:43 and I want packages from that to over write the "primary" recipe Nov 16 00:50:48 and vice versa Nov 16 00:55:43 well, like i said, for runtime dependencies just set the preferred provider to your srctree recipe. if something DEPENDS on it and needs its files in the sysroot, thats another matter, and you'd have to use a PROVIDES Nov 16 00:57:59 * kergoth ponders Nov 16 01:03:44 Does meta-toolchain-qte support canadian cross? (can I build i586 with x64? How?) Thanks. Nov 16 01:06:20 No Nov 16 01:06:49 The canadian-sdk target is for a general build on linux for windows toolchain and it's not quite right atm, sigh, need to dust off those patches Nov 16 01:08:42 OK. I like setting up build machines, anyway. If I was smart enough, I'd try to help... thanks for the insight. Nov 16 01:10:29 what's the easiest way to disassemble a single instruction? Nov 16 01:11:39 nevermind: i just objdump'ed the whole library Nov 16 01:11:42 khem: build fails Nov 16 01:11:51 turns out valgrind doesn't support mrc/mcr, not NEON Nov 16 01:12:03 * kergoth thinks about working on bitbake master w/ python 3 issues Nov 16 01:12:05 NOTE: package ltrace-0.5.3-r2: task do_compile: Failed Nov 16 01:12:07 I use MCR to peek at the cycle counter Nov 16 01:13:45 khem: http://tinderbox.openembedded.net/packages/980550/ Nov 16 01:15:49 * Jay7 -> sleep **** ENDING LOGGING AT Tue Nov 16 02:59:58 2010