**** BEGIN LOGGING AT Fri Jan 23 02:59:59 2015 Jan 23 06:32:48 So, yes, pseudo 1.6.3 has a bug that can cause its configure to fail, workaround is to make the directory that will be used with --prefix before configuring it, 1.6.4 patch should be sent to the list shortly. **** BEGIN LOGGING AT Fri Jan 23 08:40:48 2015 Jan 23 09:26:01 morning all Jan 23 09:28:17 (UMT) Jan 23 09:29:42 hi bl Jan 23 09:49:55 good morning Jan 23 09:54:46 I have a situation where one recipe in the shared state sometimes has an empty p Jan 23 09:54:46 opulate-sysroots.tgz. This doesn't happen on first execution of the recipe but s Jan 23 09:54:46 eems to happen when multiple nightly builds use the same shared state. The stran Jan 23 09:54:46 ge thing is that is only happen for one particular recipe. Is there a way of deb Jan 23 09:54:46 ugging this Jan 23 09:55:23 I have a situation where one recipe in the shared state sometimes has an empty populate-sysroots.tgz. This doesn't happen on first execution of the recipe but seems to happen when multiple nightly builds use the same shared state. The strange thing is that is only happen for one particular recipe. Is there a way of debugging this? Jan 23 10:45:54 dguthrie: are you using a common PR server for all builds? Jan 23 10:46:26 We experienced issues when sharing sstate but were not using the same PR server Jan 23 10:50:39 dguthrie: the recipe isn't installing files directly into the sysroot is it? Jan 23 10:53:04 bluelighting: No, it is using ${D} as the destination directory. Jan 23 10:53:38 bachp: What is a PR server? Jan 23 10:55:53 dguthrie: It's a service that takes care of PR numbers. See: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#working-with-a-pr-service Jan 23 10:59:31 I'd be surprised if this related to PR numbering at all Jan 23 11:00:03 dguthrie: would it be possible to pastebin the recipe? Jan 23 11:07:40 bluelightning: http://pastebin.com/UqXBAdYu Jan 23 11:10:14 dguthrie: hmm... unless "make install" for the software being built is doing something odd (since you're doing do_install_append rather than do_install) then I can't see anything that would lead to staging problems Jan 23 11:11:36 dguthrie: I would suggest that "touch *" probably isn't a good idea - with autotools that might lead to the configure process being re-run during do_compile Jan 23 11:11:54 dguthrie: I'm also having a thrift recipe internally. Would it be possible for you to work together on a shared recipe and try to bring that into meta-oe? Jan 23 11:12:19 bluelightning: ok.... I'll keep looking Jan 23 11:13:17 bachp: Ya that would be great. I can shared the native and target recipes that I have with you Jan 23 11:17:03 dguthrie: That would be great Jan 23 11:24:35 bluelightning: I will check what the history of the touch command is and remove it. Thanks Jan 23 11:25:50 dguthrie: also, btw, the standard way to pass those ac_cv_* options is to add them to CACHED_CONFIGUREVARS (or just pass them as arguments in EXTRA_OECONF) Jan 23 11:31:14 bachp: What is the standard way to shared recipes or patches? Jan 23 12:23:36 dguthrie: I think the best way would be a git repository Jan 23 12:24:07 bachp: I will follow up with you on this on Monday Jan 23 12:24:43 Ok Jan 23 12:57:25 could anyone please point why I'm getting this error: package quilt-native-0.51-r1: task do_populate_sysroot: Failed Jan 23 12:58:09 I'm running on ubuntu14.04 and I don't see it on ubuntu12.04 Jan 23 13:04:38 would anyone please look at my query? Jan 23 13:09:12 ramose: i don't think that without additional information (which release of poky? are there logs?) anybody can actually be helpful. Jan 23 15:06:47 otavio, denix: good work on the u-boot-fw-utils work. will there be a new u-boot upgrade to follow? Jan 23 16:35:17 denix: do you mind to rebase your upgrade patch on top of mine? Jan 23 16:54:03 hi folks, how to install npm modules in yocto? Jan 23 16:54:16 at this moment I just copy-paste downloaded modules during install :p Jan 23 16:54:21 yocto isn't a linux distribution Jan 23 16:54:40 I'm wondering if there is something arch independent approach Jan 23 16:54:51 kergoth, I know that Jan 23 16:55:00 the question doesn't make sense Jan 23 16:55:08 I would like to create a recipe that automatically downloads the right modules and installs them in the final image Jan 23 16:55:10 you can't install something into an umbrella project Jan 23 16:55:34 ah, that clarifies it. i'm not aware of something like that personally Jan 23 16:55:57 sed s/install/do_install\(\)/g Jan 23 16:55:59 :) Jan 23 16:56:06 install in the yocto meaning of course Jan 23 16:56:29 what I did was to run npm on the target machine and create a tarball, put in the recipe and live happy Jan 23 16:56:45 with a different target I need to redo everything, I'm trying to learn if there is a smarter approach Jan 23 16:57:05 thanks, I also think there isn't this approach Jan 23 16:57:13 presumably someone needs to create a class to handle most of the node.js build bits to ease creation of recipes to package its modules Jan 23 16:57:20 * kergoth shrugs Jan 23 16:57:48 yep, maybe :) Jan 23 16:58:37 but who cares? I live happy with a cp in the do_install and two different SRC_URI Jan 23 16:58:39 :) Jan 23 16:58:47 thanks for the great product and the help! Jan 23 17:00:49 whatever works :) np Jan 23 17:11:52 Hi Jan 23 17:12:05 Hmmm Jan 23 18:04:56 kergoth: I'm thinking "crazy" things like dropping update_data() as per http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=f082a2e511c42bd3859ca0bf13868d57c01c5073 (and the preceding patch which tweaks append/prepend handling) Jan 23 18:06:27 kergoth: It appears to 3% faster parsing and throws new warnings. The nature of those warnings needs investigating obviously Jan 23 18:06:48 That does seem like a step in the right direction. I'd rather see the _append/_prepend/etc done at getVar() time as a part of the variable expansion for each variable, make less be done at finalize time in general, though :) Jan 23 18:06:59 * kergoth goes to get caffeine Jan 23 18:07:28 kergoth: the previous patch does the append/prepend move Jan 23 18:07:56 kergoth: this makes finalize happen any time OVERRIDES is changed Jan 23 18:07:59 that seems to do the append/prepend when overrides is altered, not when the var in question is used, though, no? or am i missing something Jan 23 18:08:22 kergoth: the preceding commit: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=5ec2027f1b4b7f5cc59d2cad61efa3e7ea396073 Jan 23 18:08:39 ah! missed that Jan 23 18:08:53 yeah, i'd be on board with this, we'd have to vet the shit out of it and compare previous to new metadata and everything of course Jan 23 18:09:20 kergoth: right, this was just a quick test to see how badly it would explode :) Jan 23 18:09:29 i expect this is one of those things we've both had floating around int eh back of our minds for years, so i'm all for it :) Jan 23 18:09:38 see if it was even feasible Jan 23 18:09:46 * kergoth nods Jan 23 18:10:18 kergoth: initially I was getting DEPENDS=" }SDNEPEDESAB{\$" out of bitbake -e which was kind of amusing Jan 23 18:10:28 the more we can do to reduce the need to know about when specific events happen, the better. most folks don't even know about finalization :) Jan 23 18:10:30 haha Jan 23 18:11:42 kergoth: the fact we have update_data() calls at all kind of bothers me Jan 23 18:12:52 hmm, with the variable dependency tracking, maybe we could eliminate expandKeys(), that'd be another good step toward getting rid of it Jan 23 18:13:03 It was the ordering of d.update_data and d.expandKeys causing "bugs" again the other day which caused me to think about this Jan 23 18:14:37 kergoth: not sure quite how we'd get rid of expandKeys entirely but it means its ordering with update_data() is no longer contentious Jan 23 18:17:48 anyhow, time to get some food. Something to ponder :) Jan 23 19:14:55 hi guys Jan 23 19:15:40 is there any documentation about using rpms to update the host sysroot for cross compiling? Jan 23 19:21:54 ? Jan 23 19:24:17 hi kergoth Jan 23 19:25:37 i'm wondering what the appropriate way is to install libraries to the host sysroot for cross compilation. do i have to rebuild the toolchain again with populate_sdk? Jan 23 19:29:01 okay, the first time you asked taht question you didnt' mention the sdk at all. the question doesnt' make any sense without knowing you're working against a yocto sdk :) Jan 23 19:29:04 makes more sense now Jan 23 19:29:09 sorry :) Jan 23 19:29:33 i suspect you might have to rebuild the sdk to add to it, but i'm honestly not sure, maybe someone else knows Jan 23 19:31:09 I'm giving that a shot - I sent an email to the list a few minutes ago. burned a few hours yesterday! digging through populate_sdk_rpm task to see if i find anything useful :) Jan 23 19:31:25 cheers Jan 23 20:13:06 PREFERRED_PROVIDER_bluez4 = "bluez5" Jan 23 20:13:10 This just seems wrong. Jan 23 20:13:38 there's a pending patch series to improve the bluez switching stuff. don't think its been merged yet Jan 23 20:14:01 kergoth: thanks! you've restored my faith in humanity. :-) Jan 23 20:18:44 gabrbedd: pabigot/rfc-v2/bluez5 in poky-contrib. arranges to use DISTRO_FEATURES for selection of 4 vs 5 Jan 23 20:18:49 not sure what its status is, however Jan 23 21:02:14 Is anyone else seeing qt4-embedded build failures with the latest poky? (https://gist.github.com/af0bcdaedf456c733575) Jan 23 21:23:47 kergoth: thanks again! Jan 23 22:05:47 kergoth: I had a look at the issue with the patch, it doesn't interact well with PROVIDES since the append persists after finalise and then is applied after native.bbclass does it BBCLASSEXTEND magic :/ Jan 23 22:06:40 so holding application of append/prepend until variable access does have issues :( Jan 23 22:07:47 Hmm, interesting Jan 23 22:25:34 kergoth: it gets worse, there is a second issue, we rely on the fact that some event handlers run before update_data is called and they change OVERRIDES. With my code you can't undo class-target for example Jan 23 22:26:16 I guess we can check OVERRIDES at variable access time Jan 23 22:26:35 hmm, add an event when the internal finalize occurs, and add handlers of that event, perhaps? Jan 23 22:33:38 Crofton: hey philipster Jan 23 22:33:50 mranostay, How is it hanging Jan 23 22:34:13 * Crofton is in an airport bar charging all his devices and conducting optical experiments Jan 23 22:36:49 Is there a way to get bitbake to skip the populate_sysroot task? I'm looking for ways to share the compilation burden between my osx box and a linux VM. I'm going to be using OSX to do most compiliation and sharing sstate with a linux VM for the parts that require linux to build... populating sysroot in the VM is completely surperflous... Jan 23 22:52:07 Crofton: charging them with beer? Jan 23 22:52:42 charginf me with beer Jan 23 22:52:51 stuffing electrons into my LiPo Jan 23 22:55:44 kergoth: it gets worse. We change the values in OVERRIDES indirectly through other variables. We need to trigger internal_finalise when any dependent variable changes :/ Jan 23 22:57:08 heh, time for the datastore to get its own isolated event handling mechanism? d.fire(), blah Jan 23 22:57:18 sleepy Jan 23 22:57:55 kergoth: I think there may be a way to resolve all this but its getting a bit messy :/ Jan 23 22:58:29 might be time to take a step back and make sure this is the right approach rather than continuing to run down the rabbit hole :) Jan 23 22:58:37 not that i have any better ideas Jan 23 23:05:41 kergoth: I have something that is so close but not enough :) Jan 23 23:05:53 kergoth: time to sleep on it I think Jan 23 23:06:02 probably a good idea Jan 23 23:06:24 kergoth: I think it can be made to work, if we defer all override expansion until variable access time Jan 23 23:09:14 Hi, I'm trying to patch compat-wireless reciepe, the problem is that the backports src_uri has a destsuffix so when I add my patches in the src_uri they get in the root of the workdir, and the patches fails since the need to be in the same place as the git is Jan 23 23:09:34 How can I fix this? should I copy them manually in do_configure? Jan 23 23:15:09 you can change the strip level of the patches (-p arg to patch) as well as the dir the patch is applied in (patchdir url parameter, if it's absolute it's considered relative to ${S}) Jan 23 23:52:49 We had a really strange problem once involving the use of OVERRIDES_append to try to append something to OVERRIDES. Jan 23 23:52:51 The net result was that it wouldn't take effect for the primary multilib, but if you were doing a multilib build it would take effect for secondary multilibs. Jan 23 23:53:05 I think. Debugging it was really hard. Jan 23 23:53:53 Hmm. Jan 23 23:53:55 Wouldn't that be super-expensive? Or would it happen once per variable, I suppose? Jan 23 23:56:45 Copy them or regenerate the patches with different paths, I think. Jan 24 00:44:00 autobuilder downtime starting now. Jan 24 02:52:10 https://gist.github.com/kergoth/7d2e9f6c7d0fd759d7f2#file-check-python-deps-txt **** ENDING LOGGING AT Sat Jan 24 02:59:59 2015