**** BEGIN LOGGING AT Thu Mar 05 02:59:57 2020
Mar 05 04:57:48 <kergoth> Hmm, should probably submit vim-bitbake to vim-polyglot
Mar 05 04:58:07 <kergoth> and still need to re-align it and the bits in the bitbake repo.. probably move to an official vim-bitbake instead
Mar 05 05:06:03 <mischief> xyzzy42: what recipe does that?
Mar 05 05:11:15 <mischief> ah, i see.
Mar 05 07:23:50 <yocti> New news from stackoverflow: What OS distribution inside poky? <https://stackoverflow.com/questions/60540061/what-os-distribution-inside-poky> || Not able to find dependency for fop command in Yocto <https://stackoverflow.com/questions/60539839/not-able-to-find-dependency-for-fop-command-in-yocto> || How to install yocto on a real machine? <https://stackoverflow.com/questions/60539752/how-to-install-yocto-on-a-real-machine>
Mar 05 07:49:01 <mckoan> good morning
Mar 05 07:50:46 <erbo> mckoan: actually true this morning, sun is shining outside and kids are finally healthy and back at kindergarden
Mar 05 07:59:16 <kriive> Which is the standard way to have a Qt toolchain AND ext_sdk? Do I need to generate them separately or can I inherit populate_sdk_qt5?
Mar 05 08:09:45 <mckoan> erbo: I can't say the same due to SARS-CoV2, but let's keep on
Mar 05 08:13:32 <erbo> mckoan: :(  We have some cases where I live in Sweden (Gothenburg), but it seems to still be sort of contained and not publicly spread. Most cases are people who got it while travelling.
Mar 05 08:15:18 <kriive> here in italy everything is locked down schools and unis closed till mid march, a nightmare
Mar 05 08:20:04 <erbo> Yeah I can imagine that sucks. Let's hope things get better real soon.
Mar 05 08:24:03 <yocti> New news from stackoverflow: Any pre-built yocto hardware image to test? <https://stackoverflow.com/questions/60540909/any-pre-built-yocto-hardware-image-to-test> || Jetson Nano: Gstreamer inside docker container <https://stackoverflow.com/questions/60540684/jetson-nano-gstreamer-inside-docker-container>
Mar 05 08:54:09 <yocti> New news from stackoverflow: How to add ldd utility to bitbake image <https://stackoverflow.com/questions/54895637/how-to-add-ldd-utility-to-bitbake-image>
Mar 05 09:35:40 <Guest97881> Hello, Can we get inputs to install Yocto Eclipse Plugin? Followed steps given in https://community.nxp.com/docs/DOC-172840. It doesn't work(check comments of user:ghalib khan). Not getting response from NXP too.
Mar 05 09:37:32 <qschulz> Guest97881: I think it was dropped a few releases ago: https://github.com/crops/eclipse-yocto?
Mar 05 09:47:45 <paulbarker> Guest97881: The eclipse plugin is unsupported
Mar 05 09:56:21 <sno> is there a reasonable example how to install some vendor software in e.g. /opt/$vendor instead of /usr using Yocto?
Mar 05 10:01:19 <qschulz> sno: http://lists.openembedded.org/pipermail/openembedded-core/2020-February/293122.html worth reading
Mar 05 10:03:44 <qschulz> IIUC, if only one recipe needs this /opt dir, you can just add a do_install_append in a bbappend for your vendor recipe and move the things from /usr to /opt from there (surely needs an updated FILES_${PN} and aybe even a SYSROOT_DIRS if your vendor thing is used in a DEPENDS of another recipe. This all is my guess and would maybe be needed because /opt isn't in base-files? All of this to be triple checked
Mar 05 10:03:50 <qschulz> :)
Mar 05 10:05:11 <sno> qschulz: that's only about later stage of deployment - I have serious trouble using pkg-config and prefix=/opt/foo
Mar 05 10:05:53 <sno> PKG_CONFIG_SYSTEM_INCLUDE_PATH is set kind-of unexpected in bitbake.conf
Mar 05 10:50:09 <rburton> sno: why are you trying to install something you're building into /opt?
Mar 05 10:51:07 <sno> rburton: it's kind-of 3rd party software which is likely better separated from /usr
Mar 05 10:51:09 <rburton> the entire point of /opt is that someone who isn't the operating system can install software easily in eg /opt/foo.  if you're building software in bitbake then you've got packages and so on
Mar 05 10:51:35 <sno> rburton: a lot of NIH legacy code ...
Mar 05 10:52:36 <LetoThe2nd> rburton: forwarded the pointers to memory resident bitbake and kergoths bb to the vscode plugin guy, lets see what happens!
Mar 05 10:53:06 <sno> rburton: I'm currently at the beginning of the voyage building this software - but I found some legacy versions of OSS libs, which are required in their old versions (at least for now)
Mar 05 10:53:59 <sno> rburton: another reason for using /opt is ability to split off parts into dedicated filesystems at a later stage
Mar 05 11:04:25 <rburton> the easiest way is to just hand-code your configure call or whatever to not use $bindir etc
Mar 05 11:04:54 <rburton> you don't want to change the system paths, ie changing prefix will change where it looks up other libraries
Mar 05 11:05:08 <rburton> so just pass --prefix=/opt/foo to configure or whatever, and you're done
Mar 05 11:05:55 <sno> rburton: so you say, overriding --prefix=... over the --prefix setting, bitbake does on it's own?
Mar 05 11:06:11 <rburton> the class you are using does it
Mar 05 11:06:16 <rburton> so look there as to how to override it
Mar 05 11:07:05 <sno> rburton: I just wanted to understand where to jump in (for now) - because just modifying prefix variable in recipe fails
Mar 05 11:07:59 <rburton> just read the class that you're using to do the install to find where it sets prefix
Mar 05 11:08:12 <rburton> as you say --prefix i'm guessing autotools.bbclass
Mar 05 11:08:34 <rburton> arguably the class should split the variables up a bit so some pieces are easier to change
Mar 05 11:11:18 <sno> rburton: for short fix this seems to be reasonable
Mar 05 11:12:21 <sno> rburton: but this requires doing the package-split on my own (-debug, -src, -dev, ...)
Mar 05 11:12:46 <sno> for long-term a more generic approach would be nice (I'm happy to send patches)
Mar 05 11:23:48 <rburton> sno: i guess you could have install_prefix?=$(prefix) and use those uniformally
Mar 05 11:24:34 <luckywho> Hi, while building python3_3.5.6.bb, i'm getting error in os.path.join. https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/python/python3_3.5.6.bb?h=thud#n265
Mar 05 11:25:03 <sno> rburton: I mean this could be supported from with-in yocto
Mar 05 11:25:17 <luckywho> filename = ${@os.path.dirname(d.getVar('FILE', True))}/python3/python3-manifest.json
Mar 05 11:25:43 <rburton> sno: right, i mean bitbake.conf could have install_prefix and friends, and use those everywhere relevant
Mar 05 11:26:06 <rburton> of course autotools assumes that if you pass --libdir then that is where libraries are
Mar 05 11:26:11 <rburton> *and* where to put them
Mar 05 11:26:18 <rburton> so you might not actually solve anything in the real world
Mar 05 11:26:46 <sno> rburton: I asked to get a vision (to finally develop a patch) - hacking wasn't the problem :D
Mar 05 11:28:13 <sno> I'll take a look how to add an install_prefix to bitbake.conf and friends (autotools.bbclass)
Mar 05 11:57:40 <RP> sno: isn't that $prefix and $base_prefix ?
Mar 05 11:58:08 <rburton> base_prefix is something else isn't it
Mar 05 11:58:21 <sno> RP: nearly half entire poky assumes prefix=/usr
Mar 05 11:58:41 <sno> changing prefix breaks a lot of classes in poky/meta/classes
Mar 05 11:59:29 <sno> changing prefix was my first shot - and it fails heavily :D
Mar 05 11:59:48 <rburton> sno: changing prefix at a distro level should just work
Mar 05 11:59:59 <rburton> but you can't just change it per-recipe
Mar 05 12:00:11 <sno> rburton: but I really want prefix for system libs being in /usr :)
Mar 05 12:00:38 <sno> rburton: that's why I agree on something like install_prefix= ...
Mar 05 12:02:01 <sno> rburton: as you're right - one cannot fix the world, but sometime one might need an old perl, an old python, a very dirty openssl, or a bundle of several of those and more :)
Mar 05 12:02:57 <RP> sno: I don't see how install_prefix will buy anything that prefix and base_prefix don';t
Mar 05 12:04:56 <rburton> base_prefix is the prefix for files that are typically in /, so sysconfidir default is base_prefix/etc
Mar 05 12:05:12 <rburton> so you can set base_prefix to $prefix is you want /usr/etc
Mar 05 12:05:32 <rburton> what sno wants is the ability to drop a specific recipe into /opt/foo
Mar 05 12:05:40 <rburton> and setting prefix=/opt/foo in the recipe doesn't work
Mar 05 12:06:55 <rburton> example 1: PKG_CONFIG_PATH uses $prefix
Mar 05 12:28:16 <RP> I don't think that will end well with the way autotools works, sadly
Mar 05 12:30:57 <rburton> yeah that was my hunch too
Mar 05 12:31:09 <rburton> it assumes that --libdir is both where to find libraries and where to install into
Mar 05 12:49:17 <sno> RP: I have serious reasonable success using dedicated --prefix for some packages - mostly when building for data center some solutions
Mar 05 12:49:27 <sno> RP: that typically ends very well :)
Mar 05 12:50:04 <sno> RP: rburton: I will give that a shot and then share the results - bad or good
Mar 05 12:50:14 <sno> not today but soonish
Mar 05 13:15:40 <moustafa> I'd like to know any info or brief how I can run some app on systemsd  ?
Mar 05 13:16:15 <rburton> moustafa: the systemd documentation is good for that
Mar 05 13:16:49 <LetoThe2nd> basically its just dropping in a service file, and we have lots of examples in the layers.
Mar 05 13:23:22 <qschulz> any way to specify the ssh port for a git repo in a Yocto recipe (>= krogoth ideally but thud is fine) other than setting the port in .ssh/config?
Mar 05 13:29:07 <fullstop> smurray: I did get the updated kernel headers working yesterday.  I put them in a separate directory with sysroot_stage_all_append.
Mar 05 13:29:32 <smurray> fullstop: cool
Mar 05 13:37:59 <zeddii> which reminds me, I found my patch for that as well. I’ll attach it to my bugzilla and propose it for the next dev cycle.
Mar 05 13:39:54 <fullstop> thanks, zeddii!
Mar 05 13:55:07 <yocti> New news from stackoverflow: Gstreamer with Opencv in Yocto Linux <https://stackoverflow.com/questions/60546811/gstreamer-with-opencv-in-yocto-linux> || How to specify older gcc version in yocto <https://stackoverflow.com/questions/60546800/how-to-specify-older-gcc-version-in-yocto>
Mar 05 14:35:12 <kriive> If I have a package that throws Illegal instruction
Mar 05 14:35:48 <kriive> and the proposed fix is to set OPTTARGET=i686, should I apply it?
Mar 05 14:36:10 <rburton> kriive: you need to give us far more context if you want a decent answer
Mar 05 14:36:14 <kriive> Sorry
Mar 05 14:36:33 <kriive> https://github.com/P-H-C/phc-winner-argon2 on an Intel Atom throws illegal instruction
Mar 05 14:36:55 <kriive> Both library and binary
Mar 05 14:37:25 <rburton> so the makefile does -march=$(OPTTARGET)
Mar 05 14:37:29 <rburton> which defaults to 'native'
Mar 05 14:37:39 <kriive> Intel(R) Atom(TM) x5-Z8350
Mar 05 14:38:05 <rburton> which has interesting semantics in a cross compiler.  khem, does march=native in a cross compiler do entirely the wrong thing?
Mar 05 14:38:12 <rburton> i'd have said it should abort
Mar 05 14:39:40 <rburton> if the value for that is passed directly to -march then read the gcc man page and find the best march for your target. ideally, the recipe will patch all that out and let the BSP pass -march -mtune as needed
Mar 05 14:41:28 <rburton> like if you use meta-intel corei7-64 machine the compiler flags will include -march=nehalem -mtune=generic -mfpmath=sse -msse4.2
Mar 05 14:41:47 <rburton> but that makefile probably ignores those because people who think writing a bare makefile is sensible don't think about everyone else
Mar 05 14:44:28 <kriive> A workaround would be to set OPTTEST to 1, am I right rburton?
Mar 05 14:44:33 <kriive> could*
Mar 05 14:45:07 <rburton> yeah, try that and then look at the actual compiler calls
Mar 05 14:45:17 <rburton> see if you get the BSP -mtune used
Mar 05 14:47:28 <kriive> OPTTEST is set with := unfortunately, I thought it was ?=
Mar 05 14:47:49 <kriive> Make doesn't give a damn, I guess I have to patch it
Mar 05 14:48:52 <rburton> you can override that
Mar 05 14:49:18 <rburton> EXTRA_OEMAKE will pass variables to make for you
Mar 05 14:49:51 <rburton> kriive: ^
Mar 05 14:50:43 <kriive> Ah ok, let's try
Mar 05 14:50:55 <tzsal> I've got an error related to the incorrect compiler flags. I want to see and override flags for the specific package. Is there any opportunity to do this?
Mar 05 14:51:06 <rburton> theres a difference between FOO=bar make and make FOO=bar
Mar 05 14:51:24 <rburton> the former is environment and make won't care, the latter is overriding the value in the makefile
Mar 05 14:51:42 <rburton> tzsal: bitbake -e [recipe] | less
Mar 05 15:17:24 <kriive> rburton, EXTRA_OEMAKE = "OPTTEST=1" seems to trigger the correct BSP flags, trying right now in the actual device
Mar 05 15:25:27 <yocti> New news from stackoverflow: Yocto: create recipe for SciKit-learn <https://stackoverflow.com/questions/59065058/yocto-create-recipe-for-scikit-learn>
Mar 05 16:14:44 <khem> rburton: march=native should not be used in cross compile context, it might pick wrong instruction set and latencies and infact generate worse code
Mar 05 16:17:45 <khem> I think any march settings from Makefile can be removed since we already set an appropriate one via TUNE_CCARGS
Mar 05 16:55:08 <rburton_> khem: what happens in gcc if march=native is used in a cross build though?
Mar 05 16:55:30 <rburton_> i guess if you're targetting the same machine it can use the host to determine, which is pretty evil
Mar 05 16:55:37 <rburton_> i'd say that should be a fatal error
Mar 05 17:24:51 <skrc> my defconfig configuration which is passed using linux-imx_%.bbappend is  not reflected in .config file
Mar 05 17:29:45 <qschulz> skrc: 1) how did you create your defconfig? 2) check which paths are searched before the defconfig is found (bitbake -e linux-imx | grep -e "^FILESPATH"; or look into WORKDIR/temp/log.do_fetch (I think?)
Mar 05 17:32:02 <skrc> i copied v7_defconfig and added my custom config
Mar 05 17:32:03 <qschulz> You might be missing a subdir for your defconfig that is probably named more or less after your machine
Mar 05 17:32:21 <qschulz> define "added my custom config"? manually or by using menuconfig?
Mar 05 17:32:57 <skrc> i added manually and didn't create fragments
Mar 05 17:33:26 <qschulz> Use menuconfig. defconfigs/.config aren't meant to be edited manually
Mar 05 17:33:40 <qschulz> That's the first thing. Then check 2)
Mar 05 17:37:27 <skrc> ok
Mar 05 17:38:27 <skrc> i tried generating fragments (fragment.cfg) and added it in my linux-imx_%.bbappend
Mar 05 17:38:33 <skrc> also added the following statement
Mar 05 17:38:34 <skrc> do_configure_append() {
Mar 05 17:39:04 <skrc> now the my custom changes are appearing in .config
Mar 05 17:39:30 <skrc> I am not sure whether this is the right way
Mar 05 17:40:18 <tgamblin> RP: JPEW: reproducibility works fine for me in poky. Looks like it's unique to the manual oe-core + bitbake setup
Mar 05 17:40:28 * tgamblin updates the issue
Mar 05 17:41:29 <RP> tgamblin: good to narrow it down!
Mar 05 17:42:19 <qschulz> You didn't give us enough info. what is there in do_configure_append()? Are you using fragments now or full defconfig? What is "this" in "I am not sure whether this is the right way"?
Mar 05 17:42:23 <qschulz> skrc: ^
Mar 05 17:44:09 <skrc> since my defconfig is not getting copied to .config, i generated fragment.cfg using the below steps.
Mar 05 17:44:15 <skrc> bitbake -c menuconfig linux-imx
Mar 05 17:46:45 <skrc> now i added this in linux-imx_%.bbappend
Mar 05 17:47:00 <skrc> SRC_URI += "file://fragment.cfg"
Mar 05 17:47:13 <skrc> do_configure_append() {
Mar 05 17:50:11 <qschulz> skrc: menuconfig creates a .config. I think you need to use diffconfig afterwards to create a valid config fragment.
Mar 05 17:51:50 <skrc> ok
Mar 05 17:54:26 <qschulz> what do you have in do_configure_append()?
Mar 05 18:14:12 <smurray> for linux-imx, you should just be able to have a bbappend that does FILEEXTRAPATHS_prepend := "${THISDIR}/files" (or wherever) and drop your tweaked defconfig in it
Mar 05 18:14:19 <smurray> skrc: ^
Mar 05 18:15:22 <qschulz> smurray: depends if there is already a defconfig in a subdir named after one of the entries in OVERRIDES (hence the check of FILESPATH or log.do_fetch)
Mar 05 18:15:32 <qschulz> in which case, there is some priority going on
Mar 05 18:15:40 <smurray> qschulz: hence the prepend?
Mar 05 18:16:15 <qschulz> smurray: not enough IIRC
Mar 05 18:16:20 <smurray> qschulz: though, yeah, I've not encountered that exact case
Mar 05 18:17:26 <smurray> I think it depends on which branch in meta-freescale, they used to just have one, now I think they have a directory for the mx8 override
Mar 05 18:18:00 <smurray> the linux-imx recipe doesn't use linux-yocto, so no fragment support unless you add it yourself
Mar 05 18:19:10 <robbawebba> Hello - I have a question about adding a custom package to the host SDK sysroot. My recipe provides two packages that are both installed on the target. I would also like to install one of those packages in the Host SDK. How might I go about selecting only one package that also gets a nativesdk variant?
Mar 05 18:19:54 <robbawebba> I understand that I must add BBCLASSEXTEND = "native nativesdk" to my recipe, but I'm wondering if there's anything else I need to do to only restrict nativesdk support to one of two packages in the recipe
Mar 05 18:21:10 <smurray> skrc: if you do try that, I missed the required ':' before the ending quote.
Mar 05 18:23:10 <smurray> robbawebba: you'd probably need to do some tweaking of do_install and the FILES_${PN}-package definitions based on _class-target and _class-nativesdk, etc. overrides
Mar 05 18:51:33 <robbawebba> smurray: okay, thanks! Now that I'm fonfronted by installed vs shipped errors in bitbake I see what you mean. I'll start tweaking.
Mar 05 18:53:12 <smurray> robbawebba: might be as easy as a do_install_append_class-nativesdk to rm the files
Mar 05 19:18:42 <sheelba> I suspect this is an easy question but it's Thursday so I'll ask it anyway.  Is there a way to list all available image types within the defined layers?  Happy to keep using grep if that's the way, just wondering if there's something in bitbake for that.
Mar 05 19:19:18 <sheelba> s/types/recipes
Mar 05 19:21:23 <rburton_> bitbake -e and grepping for IMAGE_CMD_ will give you a hint
Mar 05 19:21:41 <rburton_> but some classes need to be inherited first, so that isn't complete
Mar 05 19:28:33 <sheelba> Thanks, interesting.  On this poky / freescale project I get zero results grepping for IMAGE_CMD_ from bitbake -e but 'ls meta*/recipes*/images/*.bb' shows a few, and bitbake-layers show-recipes | grep 'image' shows even more.  That's all pretty clumsy but answers my question, thank you.
Mar 05 19:33:27 <sheelba> ...and I do recognize the power of having recipes be functionally agnostic.  Giving up some convenience for flexibility there, I can live with that.
Mar 05 19:41:22 <rburton_> hmm you might want to bitbake -e [image]
Mar 05 19:41:27 <rburton_> that might be more useful
Mar 05 19:46:46 <sheelba> But if I'm looking for that initial set of image targets, how could I bitbake -e [image]?
Mar 05 19:50:51 <LetoThe2nd> sheelba: there are a couple of ways that might work if everybody sticks to the canonical way. one is "bitbake-layers show-recipes | grep image", as it is common practise to have image in the name
Mar 05 19:52:47 <LetoThe2nd> sheelba: you can also grep the layers for "inherit image" respectively "inherit core-image" and work your way from there.
Mar 05 19:53:19 <sheelba> OK, cool, thanks Leto-- that first method did show me the most of what I was looking for.  I've transgressed a bit on my own image names but I'll remember that going forward.
Mar 05 19:54:00 <LetoThe2nd> sheelba: but there is no guarantee that you will catch everything, because at least *technically* it is possible to have a recipe that is called "jeopardize" and use evil magic to avoid the image class, yet provide an image blob.
Mar 05 19:54:18 <LetoThe2nd> (i do hope though that no such thing exists)
Mar 05 19:54:21 <sheelba> fun
Mar 05 19:57:01 <LetoThe2nd> with great freedom comes great fun.. or what is the saying?
Mar 05 19:57:19 <smurray> this is an attempt to list based on IMAGE_CMD: git grep ^IMAGE_CMD | sed 's/.*IMAGE_CMD_\([^ ([{}]*\).*/\1/'|uniq
Mar 05 19:57:35 <smurray> output is sane-ish in e.g. poky
Mar 05 19:58:17 <smurray> it doesn't weed out _append, though
Mar 05 19:58:53 <LetoThe2nd> smurray: that lists image types, not images recipes (the initial question was revised just a line later)
Mar 05 19:59:29 <smurray> LetoThe2nd: oh well
Mar 05 20:00:14 <LetoThe2nd> smurray: apart from that you are totally right!
Mar 05 20:00:43 <smurray> LetoThe2nd: ;)
Mar 05 20:04:19 <sheelba> I was probably right with my wording the first time-- after all, isn't an image type just a recipe?
Mar 05 20:05:22 <sheelba> I was looking for the set of initial image types, i.e. "fsl-image-multimedia-full"; "core-image-rt" and the like...  How to drill down into an image's recipes-- that I get.
Mar 05 20:05:51 <smurray> sheelba: it's complicated ;)  In the sense I suspect you mean, yes.  But image "type" means something in OE, it's how to convert the rootfs into some output format
Mar 05 20:06:10 <smurray> sheelba: generally referring to what's listed in IMAGE_FSTYPES
Mar 05 20:06:18 <sheelba> Ah, gotcha.  But at any rate it's answered; doing the bitbake-layers show-recipes | grep image does the trick.
Mar 05 20:06:34 <sheelba> Most of the trick, that is-- enough of what *I* need.  :)
Mar 05 20:07:51 <sheelba> But yeah, image 'type' should be that set of sdimg, wic, tar.gz, that sort of thing if I understand correctly.
Mar 05 20:08:12 <LetoThe2nd> sheelba: exactly, thats what an image type is in OE
Mar 05 23:41:17 <mrc3> jonmason, where can i send patches to for meta-arm?
Mar 05 23:49:46 <smurray> mrc3: there's a mailing list for it hosted on lists.yoctoproject.org, probably there
Mar 05 23:50:01 <mrc3> thanks!
Mar 06 00:27:30 <khem> 2
Mar 06 00:57:49 <xyzzy42> What does one change to stop bitbake from auto-cleaning a package build directory when finished?  I've searched for every instance of the work "clean" in the docs and can't find it.
Mar 06 01:09:38 <khem> xyzzy42: do you have rm_work enabled
Mar 06 01:18:00 <xyzzy42> khem, yes thanks!  I didn't know the name.  I thought since the do_clean step deletes the work area, that the word "clean" would appear in the docs.
Mar 06 01:24:44 <xyzzy42> still doesn't work, after building everthing is still removed except the logs (which don't actually have logs of the build output).
**** ENDING LOGGING AT Fri Mar 06 03:00:02 2020