**** BEGIN LOGGING AT Mon Aug 22 03:00:00 2016 Aug 22 08:11:12 out of curiosity, how long does it take for layers.openembedded.org to update itself? https://layers.openembedded.org/layerindex/recipe/4179/ has been updated 5 days ago http://git.openembedded.org/meta-openembedded/commit/?id=2ea482d03a1849fbf23ed56262b5b7bb02b0b45e http://git.openembedded.org/meta-openembedded/commit/?id=2cf000c8bd9f714ccacc62e40d91057a30b673e0 but it's still showing the old information (and wireless-regdb can't be found as a recipe) Aug 22 09:11:46 anyone got a wic file to make an sd card with like a 64M fat and 2 8G partitions for a 16 M card Aug 22 09:12:04 (I know the math isn't right, but you get the idea) Aug 22 12:37:41 can I use an image recipe to create an image with wic? Aug 22 14:22:41 i'm just trying to place a binary blob under /usr/lib/node_modules/${PN} - is the following approch correct? http://pastebin.com/42n6nikp Aug 22 15:02:49 (nvm, solved.) Aug 22 15:02:53 (typos) Aug 22 16:17:27 So, a strange, but OE related question. Aug 22 16:17:53 I'm doing some perf analysis of some code, migrating some customer project from a Linaro rootfs to OE Aug 22 16:18:55 And I see that taking the program in question from the relatively old (glibc 2.17) to a semi-modern OE is now causing a lot of time to be spent in wchar related libc functions, that are not being directly called by the app in question Aug 22 16:19:27 I'm assuming that there's a knob somewhere that's been set to assume wide char stuff by default, and I'd like to turn that off as I know that's not going to be useful here Aug 22 16:19:39 Or does that sound totally off-base? Aug 22 16:21:04 Like do I want to whack out libc-posix-wchar-io maybe? Aug 22 16:21:27 (... Yes, I may have hit the age old solution of, ask in public which makes you think a tiny bit harder, git grep, blarg) Aug 22 17:05:11 Tartarus: which version of glibc are you on now ? Aug 22 17:05:26 hey khem Aug 22 17:05:35 howdy Aug 22 17:05:39 This is on jethro atm, and I just grabbed meta-musl to do some experiments Aug 22 17:05:53 for musl master is better Aug 22 17:06:12 and krogoth also works to a certain extent Aug 22 17:06:17 Yeah, moving up to master was on my todo list as of last week Aug 22 17:06:28 But this is $customer work :) Aug 22 17:14:40 Tartarus: when you say wchar related functions do you have a list Aug 22 17:16:08 khem: Lemme see if I have the perf dumps around still Aug 22 17:20:47 khem: One of them was __GI_wcrtomb Aug 22 17:21:41 bluelightning: can we nest progress handlers? i.e. have one progress handler as a stage in a multi step progress bar? i.e. for git, we have things to do after the actual clone Aug 22 17:23:02 Tartarus: what locale are you using UTF-8 ? Aug 22 17:23:35 kergoth: individual stages can have progress already - we could extend this to allow nesting Aug 22 17:24:13 kergoth: I know the git fetching is a bit weird at the moment where it appears to reset to 0 Aug 22 17:24:22 khem: That was another thing I meant to go and see about whacking, where do I tweak the default locale for the system? Aug 22 17:24:37 Admittedly I haven't dug into the progress code yet, was just rebasing the shallow bits and thought about the tarring of the repo being part of the progress, and also construction of shallow repo when shallow is enabled, so was curious about the capabilities Aug 22 17:31:56 Tartarus: IMAGE_LINGUAS = "" will disable locales Aug 22 17:32:04 set that in image recipe Aug 22 17:34:38 khem: And with no locales found it will default to 'C' ? Aug 22 17:35:31 khem: Yeah, no good, already did that. Aug 22 17:36:25 Tartarus: yes, Aug 22 17:37:12 Tartarus: you need to have locales for this image ? Aug 22 17:37:17 nope Aug 22 17:37:49 ok so you still see the issue with C locale then I assume Aug 22 17:38:10 khem: iirc yes, since I already have IMAGE_LINGUAS = "" in the image recipe Aug 22 17:53:53 Tartarus: I would recommend to build it with OE master and see if you still see the performance issue Aug 22 17:55:11 Tartarus: so it can be narrowed, ideally a small reproducer case would help Aug 22 17:57:31 khem: Yeah, I know I can't narrow this particular bit down, sorry Aug 22 17:58:09 I'll kick musl a bit harder on jethro and then go see how bad it is to move their stuff up to master instead which I think shouldn't be too hard realy Aug 22 17:58:27 Nearing the point in musl backports already that yeah, no, it's not looking great Aug 22 18:00:01 yeah, musl in jethro could only build minimal images Aug 22 22:14:52 Tartarus: its hard to figure out without more info, since the function you are talking about is very common and I would have expected more forlks to trip over it Aug 22 22:15:06 so if you can say reproduce it with krogoth Aug 22 22:15:32 and atleast we can pin it to a version of glibc ( I am not even sure if its a glibc problem to begin with) Aug 23 00:25:43 khem: Yeah, I moved up to master and that wasn't too hard Aug 23 00:26:00 I'm going to poke musl since I was in the middle of that first, and I'm generating a new glibc image now Aug 23 01:33:59 cool :) you are da man Aug 23 01:36:28 Tartarus: on master, we have world build working with musl Aug 23 01:37:03 of course with some package exceptions since they are using gnu'isms beyond accpetable limit Aug 23 01:37:13 Found a few problems in meta-openembedded Aug 23 01:37:21 nothing I coudln't just hack out for tomorrows experiment Aug 23 01:37:22 ah yeah Aug 23 01:37:36 (libnl and sys/poll.h vs poll.h, rsyslog) Aug 23 01:37:50 I'll try and do some stuff right if this turns out to be what I need to do Aug 23 01:37:56 libnl has a fix on ml Aug 23 01:38:11 sys/poll -> poll.h is simple enough Aug 23 01:38:36 hi khem got LetoThe2nd all sorted. Aug 23 01:38:37 I have asked for a rework on libnl fix but current version is good enough to proceed Aug 23 01:38:52 ka6sox:great thanks for being there Aug 23 01:39:03 kk Aug 23 01:39:05 glad to help..hope thigns are good...LTNS Aug 23 01:39:14 all is good Aug 23 01:41:15 Tartarus: whats the market segment you are looking for musl usecase Aug 23 01:42:05 khem: Hmm Aug 23 01:42:13 Not tiny IoT Aug 23 01:43:06 it doesnt have to Aug 23 01:43:29 I have a XFCE image with musl running in virtualbox Aug 23 01:43:44 with Chromium :) Aug 23 01:43:52 ha Aug 23 01:44:16 and all lmbenchmarks are 40% better on musl Aug 23 01:44:24 I mean not all Aug 23 01:44:35 but 40% of them Aug 23 01:44:44 do better on musl Aug 23 01:48:59 * Tartarus kicks off meta-toolchain + musl, runs away for the night Aug 23 01:52:25 Tartarus: No meta-toolchain, its better to do -cpopulate_sdk **** ENDING LOGGING AT Tue Aug 23 02:59:58 2016