**** BEGIN LOGGING AT Tue Jun 20 03:00:03 2017 Jun 20 03:00:10 yeah i see, and then I can easily do local commits on top and push to origin branch Jun 20 03:02:11 gcc clone —depth 1 took 9.22s user 1.74s system Jun 20 03:02:14 wow Jun 20 03:03:09 $ du -sh . (06-19 20:02) Jun 20 03:03:09 780M . Jun 20 03:03:09 i wish we could use shallow clones rather than just using shallow tarballs, but at least it saves time for future builds from the rmirror, just not for the person having to populate the mirror Jun 20 03:03:48 the shallow tarball won't be used if you have a local clone, it prefers the real thing to a shallow version, so keep that in mind if you're looking to test it Jun 20 03:04:00 so if I dont maitain a mirror with shallow tarballs Jun 20 03:04:23 then it will do normal clone and create local tarballs which will be shallow right ? Jun 20 03:04:44 yep Jun 20 03:04:46 but if I create a local mirror then it will get the shallow tarballs Jun 20 03:05:01 provided I use the above vars Jun 20 03:05:20 yeah Jun 20 03:05:36 ok Jun 20 03:05:56 the problem with a shallow clone is it alwyas goes by depth from HEAD, and we ahve no idea how manyc ommits we need from branch HEAD to get to our SRCREV Jun 20 03:05:59 even if there is no mirrors local rebuilds will be quicker I suppose Jun 20 03:06:02 we could support it for AUTOREV recipes, though Jun 20 03:11:10 local rebuilds is debatable, since local full clones are pretty fast too due to hard linking Jun 20 03:11:16 depends on what you have Jun 20 08:36:08 Hi, I'm trying to build my own dts-file but when I'm trying to compile it via bitbake it says systax error. I couldn't find a good explanation on how to do them and every site on the internet does them slighty different. So maybe just that different enough so that the compiler wouldnt recognize it. So does anyone maybe have an example for a dts for one pin only? Jun 20 09:34:47 hello Jun 20 09:34:56 does yocto support macosx? Jun 20 09:42:08 uaelinux: no Jun 20 09:43:03 any plans for support? or porting to it? Jun 20 09:45:59 uaelinux: not really. There have been a few discussions on mailing list but no commitment. There were some non-trivial details IIRC Jun 20 09:46:24 ty jku Jun 20 09:50:39 what about Haiku OS, ReactOS and Plan9? :-) Jun 20 09:50:42 uaelinux: there is CROPS, a neat way to run yocto inside a linux docker on mac Jun 20 09:51:19 uaelinux: AGL provide a docker container iirc Jun 20 09:51:44 yes CTpollard Jun 20 09:51:48 let me then try docker Jun 20 09:54:24 uaelinux: https://events.linuxfoundation.org/sites/events/files/slides/2017%20ELC%20-%20Yocto%20Project%20Containers.pdf Jun 20 09:54:49 uaelinux: https://youtu.be/JXHLAWveh7Y Jun 20 12:43:45 Facing this error | *** Libhogweed (nettle's companion library) was not found. Note that you must compile nettle with gmp support. while bibake gnutls_3.5.3.bb Jun 20 12:44:54 Ramose: are you using -b ? Jun 20 12:45:20 sorry with bitbake ? Jun 20 12:50:19 bluelightning: did n't get what you mean by -b ? Jun 20 12:50:43 Ramose: yes, I mean are you using bitbake's -b option Jun 20 12:53:38 bluelightning: No , I am just simply doing bitbake -c configure gnutls Jun 20 13:06:13 Ramose: which branch / version are you building with? Jun 20 13:08:29 bluelightning: From here I am pulling nettle recipe, https://github.com/openembedded/openembedded-core Jun 20 13:08:59 Ramose: ok so which branch are you on from there? Jun 20 13:11:46 bluelightning: don't know how to find it out ? can you tell me please Jun 20 13:23:03 rburton: nothing has happened with recipe assignments, so I just resent an update version of my patch Jun 20 13:23:27 good reminding i'll do that now :) Jun 20 13:24:37 phase 1 sent Jun 20 13:28:34 rburton: looks good, except why drop core-image and package groups recipes? Jun 20 13:28:51 I would prefer that everything listed here http://recipes.yoctoproject.org/rrs/recipes/2.4/All/ has a maintainer Jun 20 13:33:07 i actually wanted a way of saying "not relevant" Jun 20 13:36:55 What is the link between the Meta Vendor and the all project based on yocto project ? Like let say i would like a new version of my kernel. But the BSP Vendor layer is based into another kernel linux. So the work is to overlay the project to put my own layer and upgrade it? Jun 20 13:54:29 kanavin: for the dnf recipes, assuming the SHAs correspond to release tags, can you just rename the recipe to have the PV embedded in the filename? Jun 20 13:54:36 i prefer that for clarity Jun 20 13:56:31 rburton: sure, just a moment Jun 20 13:56:46 I prefer that too, but you know, inertia :) Jun 20 14:31:49 Hello! I am in the process of buying my ticket to the ELCE. Anyone know is there is any plan for a Yocto Developer Day Europe? (Sorry for the cross-post, from #oe i just realised this is the group where I should post this ) Jun 20 14:50:01 ribalda I would expect there will be Jun 20 14:50:43 I've not looked at the dates of ELCE yet.. but it's usually just before or just after.. since ELC is 3 days.. YP day is immediate day before or after (during week, not weekend) and the OE meetings (if there is oen) is usually the day after (or before) Jun 20 14:51:04 i.e. if ELC-E is Wed/Thr/Fri.. YP is usually Tuesday and OE usually Monday.. Jun 20 14:51:20 if it's Tue/Wed/Thr -- then one side is usually YP and the other OE Jun 20 14:51:34 Mon/Tue/Wed - Thr is usually YP and Fri OE Jun 20 14:52:02 hopefully that helps.. (You can look for Jefro on IRC he may have more info if anything has been scheduled) Jun 20 15:08:37 What is the yocto component responsible of choosing the kernel version ? Is it the poky we install if we base with poky? Could it be the BSP Vendor also in his layer? Jun 20 15:08:53 * yocto component => i mean meta layer sorry. Jun 20 15:09:00 fray OE meeting is Sunday at ELCE, YP dev day is Thursday Jun 20 15:09:48 ChrysD: it could be specified anywhere really, though typically it is the BSP Jun 20 15:09:48 Ahh so you made it earlier.. :P Jun 20 15:09:58 yeah Jun 20 15:10:04 thus my 'usually' comment.. ;) Jun 20 15:10:07 figured waitin guntil Friday would hurt attencance Jun 20 15:10:12 bluelightning : Ok thanks. Jun 20 15:10:13 no really good choices Jun 20 15:12:02 bluelightning : so basically what is the impact of yocto in software versionning ? Like for exemple if my all project is in morty version. So basically and logically i should take meta-qt5 if I want qt5. But the version of qt5 is maybe not the one I want. So the yocto branch name is a kind of picture of the state of the version you can use? Jun 20 15:12:28 bluelightning : i should take the morty branch of meta-qt5* Jun 20 15:12:30 the branch name is the corresponding YP branch Jun 20 15:12:39 so morty of meta-qt5 belongs with morty of the YP.. Jun 20 15:12:46 yeah i know Jun 20 15:12:48 any other 'mix' of branch names, you are on your own as far as compatibility goes.. Jun 20 15:13:04 it might work, it might not.. if it doesn't.. there likely isn't anyone who may be willing to help Jun 20 15:13:20 ChrysD: you will need to take the branch corresponding to the poky /. OE-Core branch that you are using. If you need a different version you will probably need to backport it Jun 20 15:13:20 So somebody that supply morty branch of his own meta it means that should be compatible to the morty of the YP branch Jun 20 15:13:28 yes Jun 20 15:13:33 But when we speak about YP branch, we speak about poky right? Jun 20 15:13:57 Poky is an instantiation of OpenEmbedded + bitbake + meta-yocto* layers Jun 20 15:14:21 the branch namings on all of the parts (except bitbake) should generally use the same format.. Jun 20 15:14:37 bitbake uses a numeric system, which can sometimes make it a bit tricky to know what matches Jun 20 15:14:45 bluelightning : So if i want let say qt 5.9 but the meta-qt5 provides me 5.8, i'm still able to do it, but i have to do it in my own Jun 20 15:15:21 yes.. and in the case of 5.8 -> 5.9.. usually the upgrade isn't all that bad.. Jun 20 15:15:34 but i have to be aware of all the changes i need into it Jun 20 15:16:10 Because for now the BSP vendor supply a krogoth branch for his BSP. And I see that he is using a old kernerl ... linux-imx 4.1 with a old version of gstreamer Jun 20 15:16:19 And I was curious on how i'm dependant of their meta Jun 20 15:16:32 ChrysD: you may be able to use a newer YP branch of meta-qt5 *as a reference for comparison* when you do the upgrade yourself Jun 20 15:16:46 if it's krogoth, presumably that was the last version they've tested... so using anything newer and you are again "on your own".. Jun 20 15:17:06 yeah they use kernel linux-èimx 4.1 and i need kernel 4.9 Jun 20 15:17:10 ChrysD: unfortunately that is an issue for some BSPs, it's really up to the vendor / maintainer how they keep the kernel up-to-date Jun 20 15:17:13 (this is part of the reason as a project we preach work on the latest you can"... Jun 20 15:17:46 as an OSV, we tend to pick the fall releases to sync on.. and then we're forced to upgrade things (like BSPs) that external providers aren't willing to Jun 20 15:18:05 OSV - commercial - Wind River -- not YP.. Jun 20 15:18:08 So what does poky really provide? Jun 20 15:18:32 the base platform.. including the build system, userspace, kernel, and some reference BSPs.. all in a tested/known configuration Jun 20 15:18:48 so if the BSP vendor meta give the kernel Jun 20 15:19:09 so when you start working, you know you have somethign that worked in a specific configuration with a group of people willing to answer questions and accept bugs (no support promise though) Jun 20 15:19:25 Yeah Jun 20 15:19:30 if you want more then that, you really need to go into the commercial realm of the OSV.. or aquire the skills to do it yourself.. Jun 20 15:19:54 Hello! Anyone here working with the meta-raspberry layer by any chance? Jun 20 15:20:20 so basically, if i want to use new kernel version Jun 20 15:20:21 (we find, speaking as a commercial OSV, that often the board vendor layers are generally crap.. they usually are very very specific to a single use-case and often are not generic enough to work for more then a small number of users at during a moment in time..) Jun 20 15:20:23 I do my own meta Jun 20 15:20:27 and i do the recipes for it Jun 20 15:20:37 even if the meta of my bsp vendor is in a lower verison Jun 20 15:20:58 yes.. (or one of the YP/OE advantages for commercial is there are people who you can hire to do it as well.. thus my comment about aquiring the skills) Jun 20 15:21:14 Yeah i'm an intern Jun 20 15:21:20 :) Jun 20 15:21:20 And I would like to have thsoe skills Jun 20 15:21:25 and the main focus of the project i do Jun 20 15:21:39 is to know what yocto is, how to use it Jun 20 15:21:47 And stargting with the boards Jun 20 15:22:00 my recommendation is to attempt to avoid cutting corners.. if you use the vendor kernel tree and upgrade -- then the next (different) vendor you won't be able to re-use anything you do.. Jun 20 15:22:15 But I receive a board choosen from the guy on top of me and make my opinion of how it works and if it doesn't work how to do it Jun 20 15:22:27 if you port the work to the standard YP (linux-yocto) kernel tree.. then on the next project it is usually easy to upgrade/port and otherwise re-use the code for the same or different vendor Jun 20 15:22:44 yeah so if you buy a board to a vendor Jun 20 15:22:51 Crofton|work, the bigger issue is there are 5 other LF events that week Jun 20 15:22:51 you are also very dependant of the software work that they provide Jun 20 15:23:05 yes... Jun 20 15:23:06 yeah Jun 20 15:23:09 ChrysD: on what criteria was the board chosen, just curious? Jun 20 15:23:18 larger companies can afford to standarize (and need to, due to having many more projects) Jun 20 15:23:44 kanavin, what I've see is often the hardware is choosen based on the belief that software is cheap and easy.. :P Jun 20 15:23:52 kanavin : only the hardware Jun 20 15:23:54 then they get to the software side and find the board vendor BSP doesn't match, etc.. Jun 20 15:23:58 fray: yes, that was the subtext of my question :) Jun 20 15:24:04 fray cutting corners is the road to long term support hell Jun 20 15:24:14 Crofton|work -exactly- Jun 20 15:24:16 kanavin : Because the guy on top of me doesn't know anything about yocto and linux embedded systems Jun 20 15:24:26 * Crofton|work has personal experience with this Jun 20 15:24:29 kanavin : and the main role that I have Jun 20 15:24:33 I suspect most of us here do Jun 20 15:24:50 kanavin : is for exemple saying that the choose of a board is also linked to the choose of the software Jun 20 15:25:09 kanavin : to what imply the choose of that board, to what also imply the use of yocto Jun 20 15:25:23 ChrysD: I'd say you should choose the software first, and then the hardware :) Jun 20 15:25:28 ChrysD in your situation, I would start with the BSP from the board vendor and hte matching YP or OE version.. and get it all "building" (no local mods) Jun 20 15:25:43 then start to formulate a plan on what is necessary to upgrade to get where you want and what is ok to leave 'as-is' Jun 20 15:25:44 because my role isn't to have something to work Jun 20 15:25:49 but have a kind of study of what is it Jun 20 15:25:55 so before i came to the company Jun 20 15:25:59 they bought a board so that i can work in it Jun 20 15:26:05 ...then report back to the people who choose this.. that they may have choosen poorly and why Jun 20 15:26:14 Yeah Jun 20 15:26:22 (the why is really important in those types of intern reports) Jun 20 15:26:39 you have 100% understand Jun 20 15:26:55 i realized only 3 months after that the pipeline i wanted to make it working doesn't work Jun 20 15:27:00 for gstreamer Jun 20 15:27:13 Because i was using plugins made for linux mainline Jun 20 15:27:24 And i'm not using that kernel Jun 20 15:27:26 but a linux-imx Jun 20 15:27:34 I'm forgetting what is the magic packagegroup I need to get compiler, autoconf, automake and such onto the target image? Jun 20 15:27:48 fray: tools-sdk IMAGE_FEATURE? Jun 20 15:28:05 thanks Jun 20 15:28:16 for some reason I was thinking that was an SDK flag and not a target image feature Jun 20 15:28:31 I feel it very complicated when you are a big noobish in linux, in yocto and in hardware stuff Jun 20 15:28:43 and being able after some month to explain how the whole works Jun 20 15:30:00 Be enough relevent to see the implication of every step etc Jun 20 15:31:16 this is why I always recommend make minimal changes and just get the system working first.. THEN work on modifications step by step.. otherwise you'll end up with a pile of changes and a non-working system with no hope to make it work.. Jun 20 15:31:29 yeah Jun 20 15:31:39 and i realize 3-4months after Jun 20 15:31:50 that I should change the kernel version, the gstreamer version, lot of libs version etc Jun 20 15:33:54 I had support from the gstreamer channel to know why it doesn't works, and they was speaking to me with the asumption that i have the latest version of gstreamer and the linux mainline kernel, so I was more confused about why it doesn't works... I was trying to go deep in the gstreamer process to understand how they allocate memory etc. So I loose lot of time ahah Jun 20 15:37:12 I think what is the most confusing things is when you read lot of stuff in internet, and other people say it differently and then you are really confuse. You don't know who say the right, to which context is it for and as a non-expert.... Quite difficult to make a judgement ahah Jun 20 15:40:02 ChrysD: reading stuff on the internet is not recommended :) other people are just as confused as you are. Jun 20 15:40:48 kanavin : and people on IRC don't have the whole time to explain also everything as they need to work. So what's the best way to get information? Jun 20 15:41:06 ChrysD: what fray said + study the source code and its behavior yourself Jun 20 15:41:18 ChrysD: work with a plan and in small steps Jun 20 15:41:40 kanavin : but you can make working without understanding also Jun 20 15:41:46 Is there a way to run oe_runmake on a sbudir? Jun 20 15:41:49 subdirectory? * Jun 20 15:42:03 ChrysD: yes, until you need to make a change, and it breaks, and you are completely lost Jun 20 15:42:09 You need to ask questions.. but realize not all answers will work in your situation. IRC is the first place I go.. if people know they will help, if they don't usually silence.. Jun 20 15:42:35 then move on to mailing lists.. if what someone suggestions doesn't work.. try to capture the error messages and ask for help -- but don't be surprised if they don't know how to resolve the issue.. Jun 20 15:42:54 again, mailing lists may be better for that kind... and if all else fails you will have to just dig into the code.. Jun 20 15:43:08 the smaller the amount of change per iteration the easier it will be to back out a 'self-imposed' problem Jun 20 15:43:20 fray: I slightly disagree - digging into the code is always my first choice :) Jun 20 15:43:34 kanavin : only if you know what to dig Jun 20 15:43:40 kanavin I'm assuming he doesn't know where to dig at first Jun 20 15:43:44 thus IRC/mailing lists Jun 20 15:43:59 ChrysD: usually you have some kind of error message, so you start by figuring out where that comes from Jun 20 15:44:10 kanavin : usually yes Jun 20 15:44:34 kanavin : For exemple I have problem with gstreamer Jun 20 15:44:43 kanavin : I tried to go in deep also Jun 20 15:44:55 kanavin : But I needed to understand what DMA, CMA is etc etc Jun 20 15:45:17 kanavin : I was surprise of the performance but someone told me that there is no IOMMU on the board... I never heard of that Jun 20 15:46:22 kanavin : sometimes havign some knowledge background like workflow of processing etc help I guess Jun 20 15:47:59 kanavin : And you know, when you are not use on how the open source world works, you are a bit lsot at the beggining. When a company told me that the board should works to stream video. Jun 20 15:48:31 kanavin : I understand some weeks after that they means, should works "hardwarely" Jun 20 15:50:40 kanavin : it was way more difficult to make it really working ahah Jun 20 15:50:52 I do have to wonder if they thought they could get away with an inexperienced intern doing the work, instead of paying for a commercial support contract Jun 20 15:51:20 kanavin : the reason is they don't need a result for now Jun 20 15:51:33 kanavin : they just want have a first touch on it Jun 20 15:51:55 kanavin : because the point is that they want to grow in competences and be able to work on it Jun 20 15:52:04 kanavin : so kind of "exploration" step Jun 20 15:52:26 throw away the board then :) Jun 20 15:52:44 kanavin : but the best is also why we should throw away the board ahahah Jun 20 15:52:49 kanavin : to explain* Jun 20 15:52:56 and grow your competence with an intel NUC or similar walk in the park hardware Jun 20 15:53:19 kanavin : The main prupose is to have a GUI with Qt and Gstreamer Jun 20 15:53:30 kanavin : and use I.MX6 Jun 20 15:54:37 kanavin : And the guy at the top of me is more used in real time board and do his own SOM etc.... So when he choose the board he was more worried about question of pricing / delivering guarante / performance etc Jun 20 15:55:11 ChrysD: and he didn't consider software support at all? Jun 20 15:55:27 kanavin : The vendor provide software Jun 20 15:55:41 ChrysD: take your issues to the vendor then :) Jun 20 15:55:57 kanavin : You know when you have never been into open source world, never been into linux embedded and so on... you don't know what to do ahah Jun 20 15:56:00 kanavin -- this is a common problem I see.. customer asks for 'YP Support', and the vendor says sure.. they buy the board and find out the support is 4-5-6 versions old.. Jun 20 15:56:17 fray : is the feeling i get Jun 20 15:56:34 fray : and the feeling i get is mainly they do a kind of meta layer which they put everything on it even if the recipes exist on other meta layer Jun 20 15:56:46 fray: yeah, we should do something to have stricter quality control here Jun 20 15:56:56 the old world of board + BSP + userspace (moment in time frozen) is changing.. problem is customers, BSP vendors and such havn't caught up Jun 20 15:57:07 so you understand that i'm bet confused Jun 20 15:57:10 bit* Jun 20 15:57:19 kanavin, this goes back to the YP Compatible stuff.. the solution is that you need to list -which- version(s) you are compatible with.. Jun 20 15:57:25 I don't know if that was ever added to the requiremetns Jun 20 15:57:34 yeah but even Jun 20 15:57:39 you need to know what you are talking about Jun 20 15:57:45 Like in the website Jun 20 15:57:50 but should be part of the 'YP logo/mark' stuff.. and then it's education to the customers to ask that question Jun 20 15:57:55 it was said that they were using krogoth 6 month before Jun 20 15:58:15 and when you see demonstration of the baord with video working Jun 20 15:58:20 that there is gstreamer Jun 20 15:58:23 and Qt Jun 20 15:58:26 you think it will work Jun 20 15:58:35 fray: maybe it should not be allowed to say 'yocto compatible' at all Jun 20 15:58:42 But you need to have competenced in gstreamer, qt, linux etc to understand that it doesn't really work Jun 20 15:58:58 fray: 'yocto compatible' only Jun 20 15:59:07 Yocto Project Compatible is a trademark and is part of the 'value' of the YP.. Jun 20 15:59:13 But it was the case Jun 20 15:59:22 the company said that it was krogoth compatible Jun 20 15:59:26 the part that is missing is 'which version'.. and enforcing the trademark if people are saying they are 'and are not' Jun 20 15:59:54 ChrysD, which is why the second part is teaching people to understand what this means.. old software is much harder to support.. Jun 20 16:00:35 fray : exactly Jun 20 16:01:04 but i'm even more confuse now then before ahah Jun 20 16:01:20 because the way that they support you Jun 20 16:01:26 make you do some assumption inside Jun 20 16:01:32 as they support on the way they work on the software Jun 20 16:01:43 which maybe not follow the open source philsophy, the ycoto philosophy etc Jun 20 16:01:54 So that you go far from understanding how it works Jun 20 16:01:57 yes.. which is why education is more and more important.. Jun 20 16:02:15 for exemple Jun 20 16:02:18 what is the workflow Jun 20 16:02:22 when you want to add a software Jun 20 16:02:40 When you use gstreamer plugins, be aware with what iit is compatible with.... Jun 20 16:02:55 for Morty and newer.. devtool is the recommended workflow.. (it's in the product docs).. Jun 20 16:03:13 on the specific applicaton or framework (gstreamer, qt, etc) there are no specific general recommendations Jun 20 16:03:23 i realize "too late" that for exemple, the VPU inside the I.MX6 have two different drivers Jun 20 16:03:29 one for linux-imx kernel Jun 20 16:03:32 and one for linux-mainline Jun 20 16:03:46 and some plugisn works well with one and the other with the other one Jun 20 16:04:06 But if you are not aware of the architecture of how linux works and what you need to be aware at, it's not easy Jun 20 16:04:08 the linux-mainline BTW is avalable in new YP/OE.. if that works with the i.MX6 without compromising functionality then I'd use that instead.. Jun 20 16:04:27 fray : yeah and now how to put that ahah Jun 20 16:04:28 often though graphics drivers are proprietary (at least accelerated ones) and so you are forced to use an 'evil vendor tree' Jun 20 16:04:51 fray : for example to add more complexity to my project Jun 20 16:04:57 fray : i needed to do a GUI with Qt... Jun 20 16:05:04 fray : never done C++ / QML before, and not qt Jun 20 16:05:20 QT itself could be a 3-month project to learn and setup.. Jun 20 16:05:29 fray : yeah.... Jun 20 16:05:45 fray : But for now Jun 20 16:05:55 fray : i'm using QtMultimedia Jun 20 16:05:59 fray : and it doesnt works well Jun 20 16:06:15 fray : so you need to dig into how does qt manage to work on embedded systems Jun 20 16:06:22 also the version of QT included in say meta-qt5 is often more useful for a demo situation then a 'product' situation.. this is partially because the Qt Company (or whatever they're called these days) has their own sales model and integration/support model that they want their commercial users to use Jun 20 16:06:24 fray : why it works with gstreamer but not qt xD Jun 20 16:06:58 fray: coincidenrally, I'm just now fighting with meta-boot2qt :-) Jun 20 16:07:08 fray: trying to get it to make me an image I could boot Jun 20 16:07:38 fray: meta-boot2qt is their openembedded layer that is at the center of their qt for device creation offering Jun 20 16:07:42 I tend to stay away from anything graphics related.. it just complicates the base-os... once the base-os works, I can hand it off Jun 20 16:07:43 fray : and now my problem is i'm using QtMultimedia, which is something based into Gstreamer, i have the image flickering and I don't know why without any errors.... Jun 20 16:08:07 flickering usually indicates a lack of optimization.. Jun 20 16:08:19 fray : so what I have do to try some test Jun 20 16:08:21 either the CPU is too slow or the graphics drivers are not optimized.. that kind of thing Jun 20 16:08:25 fray : is making a script which launch a process Jun 20 16:08:37 fray : and i ahve launch my gstreamer pipeline directly Jun 20 16:08:48 fray : which it works without using directly the QtMultimedia Jun 20 16:09:05 I don't know is this case.. sorry.. just know the symptom is usually a driver issue.. (but not always) Jun 20 16:09:32 fray : So it's not CPU issue Jun 20 16:09:46 fray : and the whole problem i get is because i don'tk now the architecture, the process Jun 20 16:10:15 fray : and that's are always difficult information to get Jun 20 16:10:47 I was more used to play with microcontrollers and C code, so to debug is really more easy ahahahah Jun 20 16:11:04 so when you come into linux embedded systems, with multiple softwares layers and so on Jun 20 16:11:18 if you want multimedia to work nicely you should just approach the vendor of the BSP and find out what they recommend for best performance with gstreamer Jun 20 16:11:22 a lot of ARM systems (i.MX included) have specific gstreamer, QT/graphics drivers, etc.. if you don't use them performance is aweful.. Jun 20 16:11:41 however, to get the drivers often you need to go directly to the vendor or chip maker and sign a bunch of NDAs.. Jun 20 16:11:46 rburton : and the recommendation isn't good aahah Jun 20 16:12:07 so I learn about how linux works Jun 20 16:12:08 well, start with what they claim works.. and then move from there to what you want.. Jun 20 16:12:24 fray : that's the kanavin step : throw away the board ahahah Jun 20 16:12:39 fray : that's works Jun 20 16:12:41 the embedded world is very differetn then the desktop PC... in that the vendors don't share as much, and often NDAs are required in order to get software (even for Linux) Jun 20 16:13:19 when a customer wants to buy $10m in parts, the vendors are very responsive.. when you are prototyping on one board.. they ignore you.. it's unfortunately but reality in embedded Jun 20 16:13:21 fray : you know what was the assumption of the guy on top of me? Jun 20 16:13:27 ChrysD: you don't necessarily have to throw away the board, but you could start with a board that is not such a pain in the butt Jun 20 16:13:57 ChrysD: get your stuff to work perfectly on it (yes, intel NUC or even qemu - no hardware at all :), then move it to the real thing Jun 20 16:14:13 fray : he was use to do bsp on boards. And he do by hand. So for him a vendor is just only an hardware support and not a software support. So when I say that some stuff doesn't work he was more like to say how to know to make it happen. Jun 20 16:14:22 a lot of people will prototype on a PC.. get the steps and actions right (use the same versiono fthe YP) and then run on the board.. and go to the vendor and ask for help.. that is the typical real-world commercial scenerio Jun 20 16:14:41 ya, and that is not the modern embedded world. Jun 20 16:15:00 modern embedded world is that you (as a customer) become an integrator of software and hardware components.. as well as an application provider Jun 20 16:15:02 for him , the fact that they provides the meta, is just a "+" Jun 20 16:15:14 the cream of the cream Jun 20 16:15:23 but they are not paid for supprting that Jun 20 16:15:43 well you can always ignore that stuff and do your own kernel port.. followed by your own graphics drivers port.. followed by your own codec support.. Jun 20 16:15:49 but that is stupid, it wastes time, money and effort.. Jun 20 16:15:54 yeah Jun 20 16:15:59 the old model of embedded where you do everything yourself is dead.. Jun 20 16:16:05 But you know he came from the RT embedded worlds Jun 20 16:16:16 Linux is not the right answer for everyone.. but it's a better answer then keep doing it the same way it's always been done Jun 20 16:16:22 so mostly you do the thing son your own Jun 20 16:16:23 RT doesn't mean performance.. Jun 20 16:16:32 for multimedia you are asking for performance.. Jun 20 16:16:34 I didn't speak about performance Jun 20 16:16:46 Yeah and for now he have a gui working on a 640 x 480 screen Jun 20 16:17:05 but the problem is for a 800 x 600 screen need more performance as you said Jun 20 16:17:09 so more powerfull microprocessor Jun 20 16:17:19 and come the idea of linux Jun 20 16:17:47 long term.. software is far more expensive then the hardware.. on a re-occuring basis though bean-counters look at the hardware costs as a reoccurring expensive while software -- that's just someone's paycheck.. Jun 20 16:18:08 if you actually put the costs together properly, software is often 2x (or more) then the hardware costs on a re-occuring basis.. Jun 20 16:18:22 but yet people ignore software when they make hardware decisions.. Jun 20 16:18:29 while before intern team works in the software Jun 20 16:18:39 but now is to much complicated Jun 20 16:18:55 Doing gstreamer plugins is a work by itself Jun 20 16:19:13 graphics, multimedia, board support and OS are 4 separate jobs.. Jun 20 16:19:30 Where before the guy on top of me was doing this 4 jobs ahah Jun 20 16:19:32 writing a gst plugin is definitely not trivial for non-experts, the vendor should definitely be writing those for their hardware. Jun 20 16:19:44 so explain it is a little difficult Jun 20 16:20:41 When you start from doing your own bsp on light microprocessor with RTOS doing GUI / Board Support / Multimedia etc... And you want a little more performance but you see that you need to multiply by 3 the cost because now software become to much complicate... it sometimes difficult to make people aware of that. Jun 20 16:21:31 rburton BTW I can't build tools-sdk in master.. :P Jun 20 16:21:41 | WARNING: /home/mhatle/git/oss/oe-core/build-prelink/tmp-glibc/work/core2-64-oe-linux/python3-pycairo/1.10.0-r2/temp/run.do_compile.23425:1 exit 1 from './waf build -j 144' Jun 20 16:21:44 And when you are an intern " that know nothing in embedded" and trying to explain to people that worked on more thant 10 to 20 years in embedded ( but not linux ), my position is quite complicate to convince things that I'm not sure i understant ahahah Jun 20 16:21:48 apparently that is required and doesn't currently work Jun 20 16:23:37 * fray lowers the -j for that recipe Jun 20 16:24:10 | WARNING: /home/mhatle/git/oss/oe-core/build-prelink/tmp-glibc/work/core2-64-oe-linux/python3-pycairo/1.10.0-r2/temp/run.do_compile.24571:1 exit 1 from './waf build -j 1' Jun 20 16:24:13 same problem damn Jun 20 16:24:23 AH-HA! Jun 20 16:24:24 | ../src/cairomodule.c:21:10: fatal error: Python.h: No such file or directory Jun 20 16:24:24 | #include Jun 20 16:24:24 | ^~~~~~~~~~ Jun 20 17:49:37 Having trouble finding a recipe for python-contextlib Jun 20 18:32:33 drewbert: contextlib is part of the standard library, so the recipe is python or python3, the same recipe that provides python itself. Jun 20 18:33:18 Hi kergoth, thank you for your reply. I should have posted sooner that I found that contextlib was added to the standard lib sometime after we cloned it: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=7f58a85337a0712802bdb604a58985a8800b6477 Jun 20 18:33:21 drewbert: that said, if you really need the backport for some reason (i.e. quite old python recipe for an old yocto release), search the layer index, as you would for any recipe Jun 20 18:33:22 https://layers.openembedded.org/layerindex/branch/master/recipes/ Jun 20 18:33:53 which finds http://layers.openembedded.org/layerindex/recipe/45303/http://layers.openembedded.org/layerindex/recipe/45303/ Jun 20 18:33:57 http://layers.openembedded.org/layerindex/recipe/45303/ rather Jun 20 18:34:24 so either update poky, pull in meta-openstack, or copy the recipe into your own layer Jun 20 18:35:24 contextlib2 doesn't provide contextlib (although maybe it should?). Either way, I've got a path forward. Thank you for your help. My changes are building now. *fingers crossed* Jun 20 18:37:04 drewbert: the best bet would be to just apply the changes from https://git.yoctoproject.org/cgit.cgi/poky/commit/?id=7f58a85337a0712802bdb604a58985a8800b6477 yourself via a bbappend Jun 20 18:37:12 then you wouldnt' ahve to alter the package to also support contextlib2 Jun 20 18:37:51 PACKAGES +=, RDEPENDS_${PN}-modules +=, and the new vars Jun 20 18:37:59 oh interesting. I'll read about bbappend. I just applied the package directly. Jun 20 18:38:36 s/package/patch/ Jun 20 18:45:17 rburton, btw thanks again for engagement last days and pointing me to right direction. I see there is much RTFM for me ;) Have a nice day! Jun 20 18:48:01 kergoth: just sent a patch to the bitbake list, fiddling with the cache code so your expert opinion would be good Jun 20 18:48:12 k Jun 20 18:48:52 the classic one-liner that may have disastrous consequences :) Jun 20 20:03:35 khem: for some odd reason your gcc/pie changes make the core-image-sato-sdk tests fail as klogd can't find the kernel maps file... Jun 20 20:03:39 veery odd Jun 20 20:31:45 Hi =) I have a little problem. I'm using yocto and I see recipe from the BSP vendor with the name of "linux-mainline" Jun 20 20:31:53 But when I use uname -a , i see linux-imx ? Jun 20 20:44:41 ChrysD: what do you mean by 'see the recipe'? where? just because you happen to see something on your disk doesnt' mean it's going to be used. the MACHINE controls what kernel recipe is used, either via PREFERRED_PROVIDER in the machine .conf or via COMPATIBLE_MACHINE in the kernel recipe files Jun 20 20:47:48 kergoth : I know that ahah Jun 20 20:47:54 kergoth : but in the bsp vendor meta layer Jun 20 20:48:05 kergoth : only provides linux-mainline recipes, Jun 20 20:48:18 kergoth : no linux-imx recipes Jun 20 20:48:28 kergoth : and when I do uname -a its written linux-imx Jun 20 20:48:46 the fact that you can't find the recipe means nothing Jun 20 20:48:57 run bitbake -e virtual/kernel | grep '^FILE=', that's the recipe that'll be built Jun 21 02:45:13 rburton_: is the klogd from busybox? Jun 21 02:46:38 rburton_: we also need to make sure that kernel is compiled with -fno-PIE -no-pie **** ENDING LOGGING AT Wed Jun 21 03:00:03 2017