**** BEGIN LOGGING AT Tue Apr 05 02:59:58 2016 Apr 05 08:18:05 good morning Apr 05 08:51:14 Hi, I am unable to solve a problem when launching my bitbake compilation : "ERROR: Taskhash mismatch ... verses ..." Do you have any proposition to solve this ? Apr 05 08:54:52 antb: maybe pastebin your details... Apr 05 08:55:00 sure Apr 05 08:56:54 http://pastebin.com/raw/Bzpzx5kS Apr 05 08:57:12 The message come from siggen.py Apr 05 08:57:35 I tried a cleanall without success Apr 05 08:57:55 I also tried to delete the sstate for this package Apr 05 08:58:02 but doesn't work either Apr 05 09:33:10 antb: I guess your problem is into meta-own Apr 05 09:33:41 antb: meta-own/recipes-own/skell/skell_svn.bb Apr 05 10:14:36 Hi does anybody know by chance if screen rotation with an imx6 works properly? when i'm rotating with 'xrandr -o "inverted"' i get a rotated image, but the screen is no longer redrawn Apr 05 10:52:56 mckoan: Pretty sure it doesn't depends on my recipe Apr 05 11:45:09 hi. have a problem. i have a python package that i've written a custom recipe for. in the package is a C extension. when building the extension it fails because is missing Apr 05 11:45:22 i have included bluez (tried with both 5 and 4) on the image Apr 05 11:45:48 and /usr/include/bluetooth/bluetooth.h exists in the bluez image working directory Apr 05 11:46:03 any idea? Apr 05 11:46:06 the recipe needs to DEPEND on bluez5 Apr 05 11:46:41 unless you're building the C extension on the target, in which case you need to install bluez5-dev to get the headers Apr 05 11:47:24 hm, i see i've only added bluez5 to RDEPENDS, not DEPEND. don't remember the difference...? Apr 05 11:47:58 oh, runtime dependency... i see.. thanks! Apr 05 14:44:14 Does anyone know whether it is possible to get the "official" 7'' touschreen for the Raspberry Pi to work with any of the recipes out there with yocto? If so I would appreciate any pointers. I got to the point where I could see console output to the screen after messing with rpi-update but I feel that's not the right way to do it. I was building using meta-raspberrypi. Apr 05 14:57:21 YPTM: Ready-Access Number: 8007302996 Access Code: 2705751 Apr 05 14:57:27 YPTM: Stephen Joined Apr 05 14:58:16 YPTM ross joined Apr 05 14:58:28 YPTM Saul is here Apr 05 15:01:12 YPTM: armin is on Apr 05 15:01:26 YPTM: RP joined Apr 05 15:04:31 https://wiki.yoctoproject.org/wiki/2.1_QA_Status Apr 05 15:16:07 nope Apr 05 15:17:03 Folks, at the moment bitbake git fetcher does some really weird stuff like cloning bare repository and clone from there real sources. Do you know any lightweight git fetcher which only clones latest sources with --depth=1? Apr 05 15:19:34 Kakadu: the nice people at github point out that shallow clones are not as useful as you think Apr 05 15:19:56 (but the answer is no) Apr 05 15:20:28 TT Apr 05 15:20:47 btw, do you have a like to 'nice people at github'? Apr 05 15:20:55 (there was an interesting post where they had to blacklist someone using github as a CDN Apr 05 15:22:04 https://github.com/CocoaPods/CocoaPods/issues/4989#issuecomment-193772935 Apr 05 15:26:31 thanks Apr 05 15:28:59 Maybe I should ask an original question... What should I specify in SRC_URI if I want to fetch, for example, from git@localhost:/opt/git/linux-da8xx.git ? Apr 05 15:31:46 git://localhost/opt/git/linux-dax888.git should work Apr 05 15:32:10 assuming you've got a local git daemon running, as that git uri suggests Apr 05 15:33:19 No, I didn't start any daemon Apr 05 15:33:29 and `git -c core.fsyncobjectfiles=0 ls-remote git://localhost/opt/git/linux-da8xx.git` Apr 05 15:33:43 but with git@localhost/... it works Apr 05 15:34:56 I have done something like this before with a local repo: SRC_URI = "git:////opt/git/linux-da8xx.git;protocol=file;" Apr 05 15:35:05 so git must be doing magic and seeing that whilst you're asking for git transport over tcp to localhost, it actually just shortcircuits to local files Apr 05 15:35:11 yeah, what janderson said Apr 05 15:35:16 if you want local access over files then ask for it Apr 05 15:35:47 i think you have one to many slashes there though, git:///opt should be sufficient Apr 05 15:36:02 Yeah, maybe copy paste error Apr 05 15:39:36 Yeah, I tried it but I get in ${WORKDIR} a directory /opt/linux-da8xx.git/ with contents of bare repository (probably beacuse /opt/git/linux-da8xx.git/ is a bare repository). And I decided that I'm doing something wrong......... Apr 05 15:41:27 rburton: do you think do_packagedata[vardeps] = "PKGDATA_DIR" would make sense? would that dependency be detected automatically? (and no, changing PKGDATA_DIR wasn't my idea -- it's because of a shared nativesdk/target sysroot thingy we do for our SDK.) Apr 05 15:46:30 += might be safer, but mostly checking the concept. i have the issue that changing PKGDATA_DIR (for a fix) does not seem to regenerate package data in the new location. Apr 05 15:46:36 i'll do some more digging though Apr 05 15:57:13 do_packagedata doesn't seem to get rerun when PKGDATA_DIR changes even with do_packagedata[vardeps] += "PKGDATA_DIR". same for do_package. Apr 05 15:57:17 maybe i'm missing something... Apr 05 17:32:07 Hello, I've recently run into a snag while bitbaking openssl and hoping someone can help Apr 05 17:35:20 openssl-1.0.2d-r0 is hanging during the do_fetch task (as is openssl-native of the same version). The download of the openssl tarball seems to work fine; fetch log shows a number of "Searching for ... in paths" lines (for patches), and stops after the line "Searching for debian1.0.2/block_diginotar.patch in paths" and subsequent list of locations Apr 05 18:37:24 hi there Apr 05 18:39:50 here is the vague question about the performance of Yocto builds - is i7 5960x overclocked to ~4GHz (8core) with ~32GB RAM and 500GB SSD going to be much worse than regular Xeon (24-48cores)/60+GB RAM rig? Apr 05 18:40:24 standard Xeon core is 2.3GHz Apr 05 18:40:28 something like that Apr 05 18:41:29 tough to say, there's certainly diminishing returns on the cores, but it also probably depends on what you're building. i imagine RP & pidge and co and done a fair bit of benchmarking on their autobuilder, maybe they'd know Apr 05 18:41:45 adding qt5 basically doubles my local build times :'( Apr 05 18:43:00 the tradeoff is between getting a 1) local high-spec PC, 2) local workstation (loud!) 3) VM/remote workstation Apr 05 18:43:24 the 1) is the cheapest obviously and has other benefits of having a good spec local linux terminal Apr 05 18:43:47 yeah, that's a tough call. i pretty much gave up on local build machines for noise/power reasons and live with 3, despite the annoyances Apr 05 18:43:51 laggy vpn = pain Apr 05 18:44:57 gahh. *so* annoying when i go to fire up a quick image build to test something on my board and it decides to rebuild the universe from scratch and completely block my progress for a while Apr 05 18:45:44 kergoth: that's pain, agreed; I also hate how high ping affects usability of the machine - even stupid autocompletion in the shell goes wrong Apr 05 18:46:33 so I find myself assuming that autocompletion will fill the line and start typing new stuff, and then figure out it didn't autocomplete and everything gets wrong Apr 05 18:46:41 but maybe it's just me and my high-ping experiences with linux shell Apr 05 18:46:43 ugh, yes, i do that all the time too Apr 05 18:46:52 https://mosh.mit.edu/ might be of interest here. it does speculative local echo Apr 05 18:47:22 that's cool! Apr 05 18:47:25 sadly it doesn't do agent forwarding as is, but there's a branch that does Apr 05 18:52:05 also as far as I remember usually there are couple of bottlenecks during the build. When it starts Yocto will kick off as many threads as possible until e.g. gcc gets compiled. Then everything calms down and only gcc is compiling. After that we go again full throttle and calm down waiting for some X11, am I right here? Apr 05 18:53:11 yeah, there'll always be choke points where it can't really max the resource usage of a beastly system, though it can usually run off and do some fetching or unpacking while stuck on a compile Apr 05 18:53:28 I'm just trying to say that probably after 8 cores doubling them will not deliver build in 50% of time Apr 05 18:53:52 indeed Apr 05 18:58:57 the two bottlenecks I usually see are the compiler bootstrapping and libc for the target Apr 05 19:09:36 ok, two of them fair play Apr 05 19:10:34 now GHz-wise 8cores * 4GHz = 14 core * 2.3GHz Apr 05 19:10:53 so it looks for me until I had a workstation with more than 14 cores I should not see much of a difference here Apr 05 19:11:41 including the two bottlenecks where only 1 core is involved - we probably get around 8cores*4GHz = 18cores*2.3GHz Apr 05 19:11:43 does that make sense? Apr 05 19:12:33 or I'm missing some other factors Apr 05 19:15:59 Xz: the compiling of the bottlenecks is done in parallel Apr 05 19:16:08 just the configure phase, for example, isn't Apr 05 19:16:44 In our experience, higher clock rate was better than more cores with lower clock rate Apr 05 19:25:28 I got kicked off by the web client Apr 05 19:25:38 I don't have a normal client configure at this moment :\ Apr 05 19:25:45 did I miss much of a conversation? Apr 05 19:26:07 I think the web client kicks-off everytime somebody uses my nick in the message Apr 05 19:26:10 that's messed up! Apr 05 19:28:10 again my concern is whether the workstation is going to be 2x/4x/8x faster than a high-spec PC Apr 05 19:30:28 and I'm not talking high-spec workstation, that would be around $2k-$2.5k workstation from a *cheaper* source, so read around $3.5-$4k retail price Apr 05 19:32:08 I think you got the three lines I wrote earlier? Apr 05 19:32:26 I got when you said compiler gets compiled in parallel Apr 05 19:32:32 only configure step is 1 core Apr 05 19:32:51 "In our experience, higher clock rate was better than more cores with lower clock rate" Apr 05 19:33:22 that's in favor of high-spec PC Apr 05 19:33:46 I wonder if somebody overclock Xeons also... :) Apr 05 20:48:43 hi, do you know guys if there is a clean way to apply patches contained in a repo ? Apr 05 20:48:44 something like Apr 05 20:48:44 SRC_URI = "git://myrepo.git" Apr 05 20:48:44 SRC_URI += "file://${S}/git/patches/mypatch.patch" Apr 05 20:50:51 note : the above fails because do_fetch task complains that mypatch.patch doesn't exist, which is true because do_unpack didn't occur yet Apr 05 20:51:10 I think you'll have to do this manually, e.g. in do_unpack_append() Apr 05 20:51:25 do_patch_prepend or do_patch[prefuncs] would be better Apr 05 20:51:46 actually, this might work: SRC_URI_append_task-patch = " file://${S}/git/patches/mypatch.patch" Apr 05 20:51:54 only adds it for the do_patch task Apr 05 20:52:36 kergoth: oh that seems to be exactly what I'm looking for Apr 05 20:52:44 untested, of course, but should work Apr 05 20:52:49 kergoth: I try it right now, thanks ! Apr 05 20:53:03 np Apr 05 20:59:14 Xz: It really depends on how parallel your workload is after the compiler. There is the bottleneck, then the cores help. RAM is also a good boost for general speed Apr 05 20:59:31 Xz: I have a dual Xeon system and love it, as long as its in a different room to me ;-) Apr 05 20:59:35 jbrianceau: perhaps a silly question - but why not simply have those patches on a different branch? Apr 05 20:59:40 * RP hates laggy remote systems Apr 05 21:02:49 bluelightning: good question :) it's because we use open source software that embeds patches in its own git repo Apr 05 21:02:59 and I don't want to replicate the patches in the recipe Apr 05 21:03:27 ah ok, so the question would go to the maintainers of said open source software I guess Apr 05 21:03:57 Richard, the comparison here is between 4+ GHz 8 cores with 32GB of RAM vs workstation with let's say 24 cores of Xeon 2.3GHz with let's say 64GB RAM Apr 05 21:04:21 and I'm looking for a machine I can stash under my desk Apr 05 21:04:37 but I could work on the workstation to make it quiet *I hope* if needed Apr 05 21:05:36 bluelightning: correct :) Apr 05 21:05:50 smallest Xeon on demo-depot is 14 cores, so both of them will be minimum 28 cores... Apr 05 21:07:10 kergoth: it works perfectly ! thanks again, you saved my evening Apr 05 21:07:41 np, glad to hear it Apr 05 21:19:40 Xz: I suspect the Xeon will be fairly loud Apr 05 21:20:28 Xz: I was impressed with the Xeon system I have being quiet at idle. When its wound up its like a jet turbine though Apr 05 21:40:31 RP: thats because yours is mental Apr 05 21:40:59 mine is quiet at idle, and only a bit noisy at full load - and only webkit can get it to actually turn the fans on full blast Apr 05 21:45:13 interesting, I don't hear mine even when doing two compilations simultaneously. Apr 05 21:45:13 I guess that means my office is too load. Apr 05 21:45:31 loud. Apr 05 21:53:57 looks like everybody here uses workstations Apr 05 22:01:17 Oh, and make sure you get decent SSDs. I had the "pleasure" to build on a Samsung 840 EVO until recently, that wasn't the best setup. Apr 05 22:11:47 is the EVO a SSD on PCIe? Apr 05 22:12:13 with all of the bells and whistles? like internal RAID matrix **** ENDING LOGGING AT Wed Apr 06 02:59:58 2016