**** BEGIN LOGGING AT Wed Nov 21 02:59:58 2018
Nov 21 03:33:26 <cslcm> and i'm not getting any video output from omxplayer when i reverted to the earlier commit :(
Nov 21 05:22:47 <yocti> New news from stackoverflow: PXE Boot on Yocto Intel Hardware <https://stackoverflow.com/questions/53405641/pxe-boot-on-yocto-intel-hardware>
Nov 21 06:44:47 <khem> cslcm: wpe doesnt yet work with vc4graphics so thats not going to work for you
Nov 21 06:45:33 <khem> cslcm: are you using eglfs ?
Nov 21 07:43:59 <BlauskaerM> RP: Good morning. If you have some time today. Can ju apply this patch https://pastebin.com/Ln34ZbBF to commit 4b92d150322e07f098f7c5bd2300c058d953884d in poky-contrib and see if you get a conflict error?
Nov 21 07:45:03 <BlauskaerM> Its in the case when a root filesystem is created so I also added useradd-example-depends and useradd-example to an image
Nov 21 08:20:13 <tristanram> khem: I have seen that you proposed an adaption in breakpad.bbclass http://lists.openembedded.org/pipermail/openembedded-devel/2018-November/197514.html
Nov 21 08:21:35 <tristanram> khem: How do you think about adding this in a for loop like for "BREAKPAD_BIN in ${BREAKPAD_BINS}; do <patch> done" in order to be able to specify multiple files to be symbol-dumped?
Nov 21 08:23:28 <tristanram> khem: This is the way I currently use because I have one recipe building multiple small binaries and the symbols shall be generated for all of them
Nov 21 08:48:36 <cslcm> khem: Yes, i'm using eglfs
Nov 21 08:50:04 <cslcm> (and good morning)
Nov 21 08:58:29 <cslcm> @khem - FYI, i had to revert meta-raspberrypi to c1aabba056 and then manually edit the recipe to change the srcrev back to b8ff59dccd9307f10  before I could get omxplayer actually outputting video
Nov 21 09:15:53 <pepijndevos> How can I add a poky recipe to my new layer? I want to add gstreamer. Or are all those recipes included by default? If so, can I disable some later?
Nov 21 09:17:41 <LetoThe2nd> pepijndevos: you want to add gstreamer to an image you build, i guess?
Nov 21 09:17:55 <LetoThe2nd> in that case, just extend the IMAGE_INSTALL variable in your image recipe
Nov 21 09:30:11 <pepijndevos> uuuh, I'm just really confused by all the terminology. I'm at the point where I built core-image-base, and now I'm trying to add stuff. I read I should not modify the base layers. So I guess I need to make a new layer and then add a new image recipe?
Nov 21 09:30:46 <LetoThe2nd> pepijndevos: absolutely correct.
Nov 21 09:31:48 <LetoThe2nd> pepijndevos: for starting you can just copy over something like core-image-minimal or core-image-base and name it the way you want.
Nov 21 09:51:35 <pepijndevos> Awesome! Two lines and now bitbake is packaging dropbear and gstreamer.
Nov 21 09:54:05 <LetoThe2nd> congrats by the way - given your statement that you're confused about the terminology i assume that you're rather new to the openembedded technology, and you grasped the custom image thing about an order of magnitude faster than most people who ask here.
Nov 21 09:59:24 <RP> BlauskaerM: will have a try
Nov 21 10:00:08 <pepijndevos> Yea, I started looking into yocto yesterday. Maybe "grasp" is a big word, but a close enough approximation to mess around. Current mental model: A layer is a collection of recipes and an image is a selection of recipes.
Nov 21 10:00:45 <LetoThe2nd> pepijndevos: close. an image is a selection of packages + some special sauce. a recipe can provide several packages
Nov 21 10:02:23 <LetoThe2nd> pepijndevos: the other thing you really have to understand is that what ever you set in a recipe is local to that recipe. whatever you set in a .conf file is global to the build.
Nov 21 10:02:30 <LetoThe2nd> then you're mostly good to go :)
Nov 21 10:07:16 <rburton> pepijndevos: i'm happy with that mental model and will reuse it in the future
Nov 21 10:09:56 <LetoThe2nd> as a first explanation attempt its absolutely fine, yes.
Nov 21 10:11:38 <RP> rburton, bluelightning_: We now have test result collection for each build: https://autobuilder.yocto.io/pub/non-release/20181120-1/
Nov 21 10:11:54 <rburton> yay
Nov 21 10:12:17 <rburton> we need to remember to track skipped variation too
Nov 21 10:12:31 <rburton> in case everything suddenly decides to skip and we didnt notice
Nov 21 10:13:18 <pepijndevos> hey it boots!
Nov 21 10:13:28 <rburton> \o/
Nov 21 10:13:34 <pepijndevos> Next step: wifi
Nov 21 10:13:47 <LetoThe2nd> rburton: \m/ O \m/
Nov 21 10:14:41 <RP> rburton: idea is to have a baseline we compare against
Nov 21 10:15:11 <RP> rburton: next is to start running ptest, then they'll magically appear here
Nov 21 10:15:44 <pepijndevos> I do get INIT: Id "S0" respawning too fast: disabled for 5 minutes
Nov 21 10:16:08 <RP> BlauskaerM: That does fail for me, I can also make it work
Nov 21 10:16:47 <LetoThe2nd> pepijndevos: that happens when the image tries to use a serial console that is not there. which board is this?
Nov 21 10:17:08 <pepijndevos> Raspberry pi 3
Nov 21 10:17:31 <pepijndevos> At least 99% sure it's 3, as it should have wifi.
Nov 21 10:17:43 <LetoThe2nd> pepijndevos: so you're using the meta-raspberrypi layer from git.yoctoproject and its recommended setup as mentioned in the readme?
Nov 21 10:18:18 <pepijndevos> Yes. Unless I did something stupid...
Nov 21 10:18:37 <LetoThe2nd> nobody said that. that message did not show up earlier?
Nov 21 10:18:55 <pepijndevos> Just checked Raspberry Pi 3 Model B V1.2
Nov 21 10:19:01 <RP> BlauskaerM: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=ded15777473f98c2c08b04da68c6a90f4acc59eb makes it work
Nov 21 10:19:08 <pepijndevos> Well, this was the first boot.
Nov 21 10:19:33 <pepijndevos> I'll double-check the readme.
Nov 21 10:19:39 <LetoThe2nd> pepijndevos: you have set MACHINE accordingly?
Nov 21 10:20:39 <pepijndevos>  MACHINE ??= "raspberrypi3"
Nov 21 10:21:03 <LetoThe2nd> pepijndevos: and everything on master? e.g. you just did checkout the layers, but no specific branches?
Nov 21 10:21:59 <pepijndevos> poky is on the something 2.6 branch because the tutorial said so
Nov 21 10:22:11 <LetoThe2nd> which tutorial?
Nov 21 10:22:17 <LetoThe2nd> (link please)+
Nov 21 10:22:28 <pepijndevos> https://www.yoctoproject.org/docs/2.6/brief-yoctoprojectqs/brief-yoctoprojectqs.html
Nov 21 10:22:56 <pepijndevos> The code actually mentions 2.5 but the text 2.6, so I went for 2.6
Nov 21 10:23:35 <LetoThe2nd> should be ok. its essnential that all layers are on the same branch, so thats one thing to take care of
Nov 21 10:24:01 <LetoThe2nd> 2.6 is thud, so checkout that branch on meta-openembedded and meta-raspberrypi too
Nov 21 10:25:40 <pepijndevos> oh ok
Nov 21 10:26:00 <LetoThe2nd> (while its technically possible, i actually dont expect that to be the root cause)
Nov 21 10:28:50 <pepijndevos> There is no thud on meta-raspberrypi it seems
Nov 21 10:31:15 <pepijndevos> I changed oe to thud, and nothing needed to be rerun.
Nov 21 10:32:24 <LetoThe2nd> yeah sounds very well possible, thud just got released so there should be little discrepancy between the branches and master
Nov 21 10:33:21 <LetoThe2nd> pepijndevos: but actually thats probably your thing: https://lists.yoctoproject.org/pipermail/yocto/2018-August/042302.html
Nov 21 10:33:53 <LetoThe2nd> sorry, first mail in thread is https://lists.yoctoproject.org/pipermail/yocto/2018-August/042294.html
Nov 21 10:34:08 <LetoThe2nd> in short, the UART is not enabled by default on the rpi because of bluetooth.
Nov 21 10:34:10 <BlauskaerM> RP: My man, you are a god
Nov 21 10:34:13 * BlauskaerM cries a little bit
Nov 21 10:34:21 <BlauskaerM> Thank you very much
Nov 21 10:35:09 <LetoThe2nd> pepijndevos: so you'd either have to disable the serial console so it doesn't try and respawn, or enable the uart. see https://meta-raspberrypi.readthedocs.io/en/latest/extra-build-config.html#enable-uart
Nov 21 10:38:25 <RP> BlauskaerM: hope that lets you move forward :)
Nov 21 10:39:24 <pepijndevos> So how do I disable it? Do I just set ENABLE_UART = "0", or edit the machine file to change the UART thing?
Nov 21 10:40:11 <LetoThe2nd> to disable it AFAIK you need SERIAL_CONSOLES = ""
Nov 21 10:47:29 <pepijndevos> That worked!
Nov 21 10:47:49 <pepijndevos> Now I'm trying to understand what this paragraph means: https://meta-raspberrypi.readthedocs.io/en/latest/layer-contents.html#wifi-and-bluetooth-firmware
Nov 21 10:48:42 <pepijndevos> Because lsmod does not list and wifi stuff
Nov 21 10:50:08 <LetoThe2nd> that paragraph is merely a reminder that the firmmware blobs are acquired through a side channel.
Nov 21 10:50:47 <LetoThe2nd> and lsmod will only tell you what is loaded. are they also not there altogether? look in /lib/modules
Nov 21 10:53:57 <yocti> New news from stackoverflow: BeagleBone can't boot with systemd <https://stackoverflow.com/questions/53410186/beaglebone-cant-boot-with-systemd>
Nov 21 11:01:03 <pepijndevos> I can actually modprobe brcmfmac, which from my google fu seems to be the one that's needed
Nov 21 11:01:21 <LetoThe2nd> so you're only missing autoloading, maybe
Nov 21 11:01:43 <LetoThe2nd> try to make wifi work manually, and then you can incorporate the things you need into the build
Nov 21 11:02:15 <pepijndevos> Well, lsmod shows it now, but still no wlan in ifconfig. So... not sure
Nov 21 11:02:33 <pepijndevos> Maybe I should hop in #raspberrypi
Nov 21 11:03:47 <LetoThe2nd> that might help, yes. my wifi experience is extremely limited, but mayber there's jsut something you need additionally, in terms of packages
Nov 21 11:07:15 <pepijndevos> Linux in the year 2118: Runs on quantum computers, but wifi is still broken.
Nov 21 11:07:29 * LetoThe2nd would be totally fine with that.
Nov 21 11:10:05 <RP> rburton: why would all the connections to PRServ suddenly time out on the same machine (different PRServ, different builds)
Nov 21 11:10:21 <RP> rburton: its like a resouce limit issue
Nov 21 11:10:50 <rburton> sounds like fd limit reached
Nov 21 11:11:46 <RP> this is the second time its happened, both on opensue423
Nov 21 11:12:33 <rburton> check ulimits?
Nov 21 11:13:40 <RP> rburton: comparing with another builder, could be max user processes
Nov 21 11:20:01 <rburton> 4096 processes is still quite a lot
Nov 21 11:21:38 <RP> rburton: we've been here before, its not really spread over three builds on a 48 core machine
Nov 21 11:21:47 <cslcm> How can I select the "real" gnu wget rather than the POS that busybox provides?
Nov 21 11:21:52 <rburton> cslcm: install wget
Nov 21 11:22:24 <cslcm> it won't conflict or anything?
Nov 21 11:22:30 <rburton> cslcm: nope
Nov 21 11:22:35 <cslcm> cool
Nov 21 11:22:40 <RP> cslcm: the system can deal with it
Nov 21 11:23:01 <dheeraj> I was using yocto installed inside Eurotech Gateway device.
Nov 21 11:23:01 <dheeraj> Device's eth0 port was in static IP 172.16.0.1 mode.
Nov 21 11:23:01 <dheeraj> By mistake I changed it to DHCP IP mode.
Nov 21 11:23:01 <dheeraj> I want it to be again in static IP mode.
Nov 21 11:23:01 <dheeraj> I modified /etc/network.interfaces as below
Nov 21 11:23:01 <dheeraj> auto eth0
Nov 21 11:23:01 <dheeraj> iface eth0 inet static
Nov 21 11:23:02 <dheeraj> address 172.16.0.1
Nov 21 11:23:02 <dheeraj> netmask 255.255.255.0
Nov 21 11:23:03 <dheeraj> but ping 172.16.0.1
Nov 21 11:23:03 <dheeraj> Is not successful.
Nov 21 11:23:14 <dheeraj> Any help is appreciated.
Nov 21 11:23:16 <rburton> cslcm: ditto for anything else busybox has a limited version and you want the real one, just install it and it replaces the busybox one
Nov 21 11:23:33 <RP> rburton: 50*50*3 = 7500 (BB_NUMBER_THREADS * PARALLEL_MAKE * No Builds)
Nov 21 11:23:48 <rburton> RP: right
Nov 21 11:24:45 <cslcm> i dont even see the point of busybox's wget, it simply doesn't work for any TLS sites.. which is 99% of the internet now :P
Nov 21 11:26:53 <rburton> cslcm: CONFIG_FEATURE_WGET_HTTPS=y
Nov 21 11:26:56 <rburton> so it should
Nov 21 11:28:16 <rburton> albeit in the busybox style of not actually verifying anything
Nov 21 11:28:17 <RP> cslcm: missing ca-certificates ?
Nov 21 11:28:52 <rburton> /config:       Note: currently, TLS code only makes TLS I/O work, it
Nov 21 11:28:52 <rburton> /config:       does *not* check that the peer is who it claims to be, etc.
Nov 21 11:29:21 <cslcm> RP: no, just fails to handshake with most sites
Nov 21 11:29:25 <cslcm> i guess it has no tls 1.2
Nov 21 11:29:53 <dheeraj> @rburton
Nov 21 11:31:09 <RP> rburton: frightening
Nov 21 11:33:10 <rburton> cslcm: grep suggests it at least claims to have tls
Nov 21 11:33:14 <rburton> i guess its just a bit minimal
Nov 21 11:33:24 <rburton> this is what proper wget is for :)
Nov 21 11:49:40 <RP> rburton: -next ok? I dropped the musl upgrade
Nov 21 11:51:00 <rburton> funny, just pulled it
Nov 21 11:51:02 * rburton looks
Nov 21 11:52:05 <rburton> RP: yes
Nov 21 11:54:59 <RP> rburton: thanks
Nov 21 11:57:48 <RP> rburton: I'm idly wondering about having the AB generate a buildhistory report into that results dir
Nov 21 13:20:09 <rburton> RP: i did a thing!
Nov 21 13:20:09 <rburton> https://pastebin.com/5FpY9xyE
Nov 21 13:20:09 <rburton> busybox-specific, would be neat to make update-alternatives handle that
**** BEGIN LOGGING AT Wed Nov 21 13:20:27 2018
Nov 21 14:12:50 <yates> i can't see exactly where this autotools do_configure() is failing: https://paste.fedoraproject.org/paste/ZfQ4wUCJAeZ~xBzI7YutyA
Nov 21 14:14:11 <yates> is one of the warnings causing a subsequent failure?
Nov 21 14:14:21 <yates> sure, this: ERROR: autoreconf execution failed.
Nov 21 14:14:31 <yates> but WHY is it failing? i've no clue
Nov 21 14:25:24 <kergoth> yates: configure.in:2886: warning: macro 'AM_PATH_GTK' not found in library, configure.in:2841: warning: PKG_PROG_PKG_CONFIG is m4_require'd but not m4_defun'd
Nov 21 14:30:29 <JPEW> RP, rburton: Alright, so how do we get wine on the autobuilders?
Nov 21 14:30:41 <rburton> JPEW: ask halstead nicely
Nov 21 14:31:02 <rburton> yates: that means 'inherit pkgconfig'
Nov 21 14:33:44 <rburton> RP: did you test the wine patch? i didn't get around to it yesterday
Nov 21 14:36:22 * rburton installs wine
Nov 21 14:44:06 <RP> JPEW: I'm thinking we add it to a specific one and then code that into the configuration. Not sure I want to try all the different distro variants of this
Nov 21 14:44:24 <rburton> RP: can the new buildbot pick specific builders?
Nov 21 14:46:59 <rokm> Hey hi hello
Nov 21 14:47:08 <RP> rburton: we can have specific targets on specific builders
Nov 21 14:47:14 <rburton> nice
Nov 21 14:47:28 <rburton> is that some sort of tagging system, or literally builder name
Nov 21 14:47:44 <rokm> Can someone help me to remove psplash from image (and why it is so fu#$%#$% hard to do this) ?
Nov 21 14:47:51 <RP> rburton: you just register specific targets with specific builders
Nov 21 14:47:57 <rburton> RP: gitlab ci does tags so you can say 'this job needs the tag 'wine'" and have several hosts have that tag
Nov 21 14:48:19 <rburton> rokm: remove 'splash' from IMAGE_FEATURES
Nov 21 14:48:22 <RP> rburton: this is configured in the python config of the autobuilder
Nov 21 14:48:25 <rokm> doesnt work
Nov 21 14:48:40 <rburton> rokm: then something else in your image is pulling in  psplash directly
Nov 21 14:49:03 <rokm> I added IMAGE_FEATURES_remove = "splash" in local conf
Nov 21 14:49:22 <rokm> also tried to to change this in image recipe
Nov 21 14:50:14 <rokm> but still wants to fetch sources
Nov 21 14:50:24 <rokm> and it fails because of missing patch
Nov 21 14:51:09 <rburton> rokm: bitbake -e <your image name> and verify that IMAGE_FEATURES doesn't have splash in
Nov 21 14:52:24 <rokm> it is in FEATURE_PACKAGES_splash
Nov 21 14:52:40 <rburton> thats fine, you care about IMAGE_FEATURES
Nov 21 14:53:37 <rokm> it is
Nov 21 14:53:46 <T_UNIX> is there a recommended way to select files on an image basis while keeping the changes local to the affected recipes? i.e. use a different `fstab` (`base-files`) depending on whether the IMAGE_FSTYPES contains `wic` or not?
Nov 21 14:53:51 <rokm> in .bb
Nov 21 14:55:27 <rokm> there is 12 operations, original recipe appends splash but later on my .bbappend removes it
Nov 21 14:55:51 <rokm> so at the end
Nov 21 14:56:13 <rokm> it is not in IMAGE_FEATURES
Nov 21 14:56:24 <rokm> maybe I should clean build
Nov 21 14:57:03 <rburton> rokm: maybe the image explicitly adds psplash via a dependency? (shouldn't be doing that)
Nov 21 14:57:26 <rburton> rokm: what release are you using?
Nov 21 14:57:28 <rokm> I tried to avoid clean
Nov 21 14:57:36 <rokm> rocko
Nov 21 14:57:45 <rokm> 2.4
Nov 21 14:57:57 <rburton> bitbake -g [your image] to generate a .dot file
Nov 21 14:58:28 <rburton> then if you have it, try oe-depends-dot -k psplash task-depends.dot (iirc, that's the right name)
Nov 21 14:58:35 <rburton> i think oe-depends-dot was in 2.4
Nov 21 14:58:46 <rburton> if not then just search the dot file for psplash and see what depends on it
Nov 21 14:58:50 <rokm> ok will try
Nov 21 14:58:57 <rokm> thanks
Nov 21 15:09:03 <T_UNIX> currently I use `FILESEXTRAPATH_prepend` in conjunction w/  `bb.utils.contains(...)` within a `base-files_%.bbappend` but, it seems, it's not reevaluated even if the conditional changes.
Nov 21 15:09:19 <pepijndevos> I'm trying to write a recipe for an unversioned binary library, so I found this: https://wiki.yoctoproject.org/wiki/TipsAndTricks/Packaging_Prebuilt_Libraries#Example
Nov 21 15:09:27 <pepijndevos> Question: what about the headers?
Nov 21 15:10:30 <T_UNIX> pepijndevos: are they provided with the library?
Nov 21 15:10:36 <pepijndevos> I read there is some magic that splits it up in a -dev package, which sounds good... but how? Also, what is ${D} and ${S}?
Nov 21 15:10:43 <pepijndevos> T_UNIX, yea
Nov 21 15:11:12 <T_UNIX> ${D} is Destination (Target installation directory) ${S} is the source dir.
Nov 21 15:11:31 <pepijndevos> Oh, the versioned example does 	install -m 0755 ${S}/*.h ${D}${includedir}, so I guess that's fine. Will that end up in the -dev package then?
Nov 21 15:11:34 <T_UNIX> pepijndevos: is it makefile based?
Nov 21 15:12:03 <pepijndevos> No, it's... "here is a bunch of headers and .so file"-based
Nov 21 15:15:42 <T_UNIX> have a look at poky/documentation/ref-manual/ref-classes.xml
Nov 21 15:16:09 <T_UNIX> oops, meant `poky/meta/recipes-extended/libnsl/libnsl2_git.bb`
Nov 21 15:18:11 <georgem> Hmmm "Your membership in the mailing list yocto has been disabled due to excessive bounces". Seems odd that my gmail account would be bouncing.
Nov 21 15:22:12 * zeddii_home gets that all the time for my various list subscriptions
Nov 21 15:26:43 <mcfrisk> sigh, linux-libc-headers say that do not fork for custom kernel but in the end that seems to be the only way...
Nov 21 15:37:35 <rburton> georgem: something bad happened this morning, lots of people on gmail got kicked
Nov 21 15:37:43 <rburton> georgem: as a list moderator i got 500 mails about it
Nov 21 15:38:33 <rburton> (waiting on halstead to wake up to get his expert opinion)
Nov 21 15:39:16 <georgem> rburton: I clicked the reenable button so probably no harm done as long as that works
Nov 21 15:51:39 <T_UNIX> would `FILESEXTRAPATHS[vardeps
Nov 21 15:52:07 <T_UNIX> would `FILESEXTRAPATHS[vardeps] += "IMAGE_FSTYPES"` be appropriate?
Nov 21 15:57:01 <pepijndevos> T_UNIX, what was there about libsnl2 that I should look at? Seems to compile from source, rather than install binaries like I'm trying to do.
Nov 21 15:57:44 <rburton> pepijndevos: just copy the headers to ${D}${installdir} and they'll be packaged into PN-dev for you
Nov 21 15:58:31 <pepijndevos> Right. Cool. Wait... installdir or includedir?
Nov 21 15:59:06 <T_UNIX> pepijndevos: just wanted to point at a recipe that packages a lib and a corresponding `-dev` package
Nov 21 16:00:05 <pepijndevos> ok
Nov 21 16:03:31 <JPEW> Which mailing list are patches to meta-poky submitted?
Nov 21 16:04:44 <kanavin_home> https://lists.yoctoproject.org/listinfo/poky I think
Nov 21 16:08:13 <rburton> JPEW: poky@
Nov 21 16:08:22 <JPEW> Thanks
Nov 21 16:08:24 <rburton> JPEW: you're missing HOSTTOOLS+="wine" in the mingw layer.conf
Nov 21 16:08:39 <rburton> JPEW: also turns out debian doesn't ship wine32 but the sdk i just built is all 32-bit
Nov 21 16:09:14 <rburton> oh maybe that's just wine moaning
Nov 21 16:09:17 * rburton tries again
Nov 21 16:10:04 <JPEW> rburton: Ah I have a bad line in the HOSTTOOLS.... it's looking for the wrong class
Nov 21 16:11:51 <JPEW> I'll try the 32-bit SDK again. It was working (I'm running fedora....) but I've made some changes since then.
Nov 21 16:13:17 <rburton> JPEW: what sdk config are you testing?
Nov 21 16:13:22 <rburton> i'll do that first :)
Nov 21 16:13:44 <JPEW> x86_64-mingw32
Nov 21 16:14:09 <rburton> fwiw here with a i686 sdk i get
Nov 21 16:14:10 <rburton> Exception: oeqa.utils.subprocesstweak.OETestCalledProcessError: Command '['wine', 'winepath', '-w', '/data/poky-tmp/master-mingw/work/intel_corei7_64-poky-linux/core-image-minimal/1.0-r0/testimage-sdk/']' returned non-zero exit status 1
Nov 21 16:14:33 <rburton> running that outside of OE works fine
Nov 21 16:14:53 <rburton> so maybe thats hosttools breaking it?
Nov 21 16:15:36 <JPEW> possibly, I will push up the fix in a minute
Nov 21 16:15:44 <rburton> hm maybe, wine does immediately exec /usr/lib/wine/wine64 here
Nov 21 16:16:19 <rburton> flipped to 64-bit, rebuilding
Nov 21 16:19:37 <yates> ok i added pkgconfig and binconfig to the inherit, and that fixed some things, but the autoreconf is failing and i don't know why: https://paste.fedoraproject.org/paste/~yQChkb-Olsg88Pat1WFDA
Nov 21 16:20:12 <yates> is it the AM_PATH_GTK macro? does the gtk+ package provide that?
Nov 21 16:20:29 <yates> here's my .bb so far: https://paste.fedoraproject.org/paste/i3tFdh3yFQy2W7gOYAnhow
Nov 21 16:23:56 <rburton> yates: i'd use += on EXTRA_AUTORECONF as there's a default
Nov 21 16:24:23 <yates> ok, trying that..
Nov 21 16:24:42 <rburton> the gtk+ package should provide that macro yes, but it looks like autoreconf isnt looking in the sysroot
Nov 21 16:25:17 <kergoth> build/aclocal should be checked to make sure it only contains local macros, not ones that sould be coming from other recipes
Nov 21 16:25:46 <rburton> oh dear god they've got a hacked fork of general.m4 functions.m4 etc
Nov 21 16:25:58 <rburton> https://github.com/wxWidgets/wxWidgets/tree/master/build/autoconf_prepend-include/autoconf
Nov 21 16:26:23 <kergoth> oh good god
Nov 21 16:26:34 <rburton> well their autogen.sh says to pass -B build/autoconf_prepend-include
Nov 21 16:26:35 <yates> this doesn't sound good.
Nov 21 16:26:35 <kergoth> thats horrifying
Nov 21 16:26:39 <rburton> and you're not doing that
Nov 21 16:27:21 <rburton> also looks like they support gtk+3 so you can swap that in instead of gtk2
Nov 21 16:27:38 <rburton> also did you read https://github.com/wxWidgets/wxWidgets/blob/master/README-GIT.md
Nov 21 16:27:53 <yates> rburton: some of it
Nov 21 16:28:11 <yates> oh, not that. no
Nov 21 16:28:43 <rburton> remember what i was saying about tarballs just working more often? :)
Nov 21 16:29:09 <yates> rburton: did you say that? :)
Nov 21 16:29:16 <rburton> they're using submodules so you'll need to use the gitsm fetcher
Nov 21 16:29:25 <rburton> (or the 3.0.4 tarball)
Nov 21 16:29:44 <rburton> assuming the tarball will autoreconf, given the existing recipe and the state of this tree, that's not a given
Nov 21 16:29:45 <yates> so that's the ony change: use gitsm:// instead of git:// ?\
Nov 21 16:29:59 <rburton> i'll delegate that to kergoth as he wrote it
Nov 21 16:41:08 <yates> suggestion: put date/time at the top of your log files
Nov 21 16:41:27 <yates> i gues you could get it from the filesystem
Nov 21 16:42:07 <rburton> hell i've argued for timestamps per line
Nov 21 16:42:25 <rburton> useful to spot slow sections
Nov 21 16:43:22 <yates> yes, i don't see much negative to that and a lot of potential positive
Nov 21 16:43:50 <yates> are these harmless? Unescaped left brace in regex is deprecated, passed through in regex; marked by <-- HERE in m/\${ <-- HERE ([^ \t=:+{}]+)}/ at /Storage/production-hardware-revision-A-1.0/sources/poky/build-hw-test-image/tmp/sysroots/x86_64-linux/usr/bin/automake line 3939.
Nov 21 16:44:06 <rburton> iirc yeah, new perl moans more
Nov 21 16:47:37 <yates> ok, this looks better, but still failing: https://paste.fedoraproject.org/paste/Ew9P1AWsw1WKDhuPmQzaJg
Nov 21 16:47:50 <yates> is it the autoheader: warning: missing template: XYZ?
Nov 21 16:48:01 <kergoth> it'd be nice to be able to exert a bit more control over how our task functions are run in general, i.e. customize the interpreter, or pipe the output, then adding timestamps would be trivial by passing the output through a script that adds it
Nov 21 16:48:48 <kergoth> possibly, or skip autoheader, not all projects use it at all. check autogen.sh to see what they run
Nov 21 16:49:08 <rburton> they *just* run autoconf
Nov 21 16:49:24 <yates> "they"? wx project?
Nov 21 16:49:31 <rburton> hm but they do have an aclocal
Nov 21 16:49:54 <kergoth> iirc autoheader generates config.h.in from the configure script and macros, which is then written to config.h, but some projects have hand-constructed config.h.in, and autoheader blowig that away causes headaches
Nov 21 16:50:08 <rburton> no config.h.in in the repo
Nov 21 16:50:33 <rburton> aghghghhgh https://github.com/wxWidgets/wxWidgets/blob/master/build/aclocal/gtk-3.0.m4 they also fork the gtk macros
Nov 21 16:50:35 <kergoth> ah, good, so just old content. the configure.in gives that away already, of course
Nov 21 16:50:53 <kergoth> wow, that's crazy, it's a miracle it builds at all for anyone if all of that gets out of sync..
Nov 21 16:51:20 <yates> i picked a shitty first-autotools recipe project, eh?
Nov 21 16:51:25 <rburton> hell yes
Nov 21 16:51:36 <yates> lol
Nov 21 16:51:41 <rburton> this is figure out what sort of crazy they're up to and replicate it
Nov 21 16:51:56 <kanavin_home> I am surprised wxwidgets is still alive :)
Nov 21 16:52:13 <rburton> the old recipe in meta-ros probably has the right idea: don't regenerate, cross your fingers, run configure directly
Nov 21 16:52:18 <rburton> and hope for the best
Nov 21 16:52:47 <kergoth> if it uses libtool that could be a world of pain
Nov 21 16:52:48 <rburton> yates: try adding --exclude=autoheader to EXTRA_AUTORECONF
Nov 21 16:52:49 <yates> i wanted to use qt but my boss didn't like the license
Nov 21 16:53:06 <rburton> yates: just use gtk3 directly?
Nov 21 16:53:15 <kanavin_home> soon, gtk4
Nov 21 16:53:37 <rburton> i mean, wx is just a gtk wrapper on linux, so what's the point
Nov 21 16:54:33 <rburton> JPEW: GccCompileTest class: SDK doesn't contain make
Nov 21 16:54:45 <yates> i am not sure about the low level. i know qt was wonderful (for the most part) when i used it a couple years back. i espeically liked their signals and slots.
Nov 21 16:54:52 <rburton> JPEW: bit surprised the core-image-minimal sdk doesn't have make in
Nov 21 16:55:04 <rburton> yates: gtk+ is basically the same, but written in C
Nov 21 16:55:11 <rburton> (with a pure oss license)
Nov 21 16:55:42 <JPEW> rburton: It suprised me too. I have tested with it added, and it works. I assume that means you got the tests to start?
Nov 21 16:55:43 <yates> i should play around with it sometime
Nov 21 16:55:51 <rburton> JPEW: yeah the rest passed!
Nov 21 16:56:04 <kanavin_home> yates: https://developer.gnome.org/gobject/stable/signal.html
Nov 21 16:56:13 <JPEW> \o/
Nov 21 16:56:30 <kanavin_home> signals in gtk :)
Nov 21 16:57:04 <RP> rburton: you've beaten me too it, been focusing on autobuilder changes first :)
Nov 21 16:57:56 <rburton> JPEW: bonus points: use the elf inspection code to verify the generated binaries match the expected arch
Nov 21 16:58:00 * rburton finds example
Nov 21 16:58:30 <yates> kanavin_home: interesting. i thought that was a qt invention.
Nov 21 16:58:31 * RP tries to figure out a better name for "nightly"
Nov 21 16:58:58 <yates> carnally?
Nov 21 16:59:00 <JPEW> rburton: I stole that from the other SDK tests (but added the checks to the GCC tests).
Nov 21 16:59:30 <JPEW> rburton: The assimp (cmake) SDK test in oe-core does it
Nov 21 17:00:05 <RP> JPEW: guess who wrote the assimp test :)
Nov 21 17:00:48 <rburton> yates: signals are pretty old tech really
Nov 21 17:00:59 * zeddii_home is clearly being dense in trying to define a new FSTYPE.
Nov 21 17:01:08 <zeddii_home> time to wander off before I smash something
Nov 21 17:01:28 <JPEW> RP: Ah... indeed.
Nov 21 17:01:34 <rburton> JPEW: ah yes there it was.  of course it was cmake i couldn't trust to do the right thing :)
Nov 21 17:02:03 <yates> rburton: exclude=autoheader worked! at least i'm getting past it to the next error, which is from binconfig_package_preprocess
Nov 21 17:02:06 <yates> thank you!
Nov 21 17:03:18 <kergoth> heh, there are 9 recipes in oe-core excluding autoheader right now :)
Nov 21 17:03:55 <rburton> yeah people who don't use it tend not to bother using the full AC_DEFINE syntax and autoheader gets upset
Nov 21 17:05:04 <JPEW> rburton: I may have misunderstood... I am actually verifying the elf files in the GCC and binutils test cases already (grep for assertIsTargetElf)
Nov 21 17:05:20 <rburton> JPEW: oh yes, nice
Nov 21 17:05:42 <yates> does do_compile happen before do_package? i would think of couse
Nov 21 17:05:44 <yates> course
Nov 21 17:05:46 <rburton> yes
Nov 21 17:06:02 <JPEW> rbutrton: Also the tests just passed for me using i686-mingw32
Nov 21 17:06:13 <rburton> nice
Nov 21 17:06:38 <yates> so the project is getting through the "make" stage. (of ./configure make, make install..)
Nov 21 17:06:45 <yates> the recipe, rzather
Nov 21 17:08:51 <JPEW> RP: nightly -> regression ?
Nov 21 17:09:03 <rburton> yates: if binconfig is failing then they moved the scripts around that it is looking for (binconfig neuters and otherwise edits scripts)
Nov 21 17:10:53 <rburton> JPEW: before merging to master, squash that hosttools fix into the right place
Nov 21 17:11:02 <yates> log.do_package: https://paste.fedoraproject.org/paste/aXc0lYsVnODQk2yPTNRT-w
Nov 21 17:11:06 <JPEW> rburton: Will do
Nov 21 17:11:49 <yates> we must have our wx-config now
Nov 21 17:11:53 <rburton> yates: pretty much what it said. that file got renamed or something?
Nov 21 17:12:00 <yates> looking..
Nov 21 17:12:25 <rburton> JPEW: also just added nativesdk-make to the packagegroup here
Nov 21 17:13:10 <yates> it IS there. it's a symlink though
Nov 21 17:13:27 <yates> oh.. symlink to a non-existent target...
Nov 21 17:13:38 <yates>   lrwxrwxrwx 1 YoctoBuild users     58 Nov 21 12:00 wx-config -> /usr/lib/wx/config/arm-fslc-linux-gnueabi-gtk2-unicode-3.0
Nov 21 17:13:51 <yates> of course it's not there!
Nov 21 17:15:26 <RP> rburton, JPEW: quick and full?
Nov 21 17:16:38 <yates> where's the sysroot during such a build?
Nov 21 17:17:28 <zeddii_home> If I’m defining a new IMAGE_FSTYPE in a layer, I’m gleaning from my google and grepping that I shouldn’t try and globally add it to IMAGE_FSTYPE  in local.conf … can anyone confirm or deny that fact ? It’ll direct my grepping a bit further :D
Nov 21 17:17:52 <RP> zeddii_home: correct as then it would end up applied to non images
Nov 21 17:18:10 <RP> zeddii_home: also tends to break things like initramfs
Nov 21 17:18:12 <zeddii_home> ok. so I should test by just adding it in an image recipe.
Nov 21 17:18:26 <RP> zeddii_home: yes
Nov 21 17:18:35 <zeddii_home> ok. cool. that’ll get me going again.
Nov 21 17:19:08 <zeddii_home> I’m trying to finish up the oci class and my integration is showing gaps in my knowledge in this area :D
Nov 21 17:20:03 <RP> zeddii_home: the image classes stuff needs more work to make it nicer. We got so far with it
Nov 21 17:20:56 <zeddii_home> np. I may ask a few more questions .. but I’ll try and figure most of it out myself. time will tell. I’d like to update the 2.7 planning document with a basically finished image type and class as reference :D
Nov 21 17:46:26 <yates> where is the do_install script for autotool"
Nov 21 18:01:23 <rburton> yates: in autotools.bbclass
Nov 21 18:01:28 <rburton> RP: works for me
Nov 21 18:01:48 <rburton> JPEW: added nativesdk-make to the sdk and now all the tests pass \o/
Nov 21 18:05:44 <rburton> JPEW: i'd say next steps are 1) add my make patch to next 2) squash that hosttools fix 3) get wine on an AB worker 4) extend mingw ab run to build both 32- and 64-bit sdks, and test them both 5) ... 6) profit!
Nov 21 18:10:34 * armpit was hoping to get beer on the AB working
Nov 21 18:15:51 <rburton> yates: (its just make install)
Nov 21 18:25:55 <no_such_user> Hiya
Nov 21 18:37:50 <la_croix> How would I install a load of python packages onto an image? I've managed to get python3 into it, as well as pip3, so I can install packages once the image is booted, but I'm not sure how to do it when building the image
Nov 21 18:50:41 <JPEW> rburton: #1 and #2 done.
Nov 21 18:57:17 <yates> your stickmen are better than mine: \o/ vs. \/
Nov 21 18:58:01 <yates> i'll do a bbappend to that recipe...
Nov 21 19:02:55 <yates> i'm seeing the wx-config file in both of the following locations:
Nov 21 19:03:00 <yates> ./armv7at2hf-neon-fslc-linux-gnueabi/wxwidgets/3.0.4-r0/sysroot-destdir/usr/bin/crossscripts/wx-config
Nov 21 19:03:06 <yates> ./armv7at2hf-neon-fslc-linux-gnueabi/wxwidgets/3.0.4-r0/build/wx-config
Nov 21 19:03:25 <yates> what's the difference? which one is the output of the do_install?
Nov 21 19:03:35 <kergoth> neither
Nov 21 19:04:07 <kergoth> build is wher eit was built in do_compie, sysroot-destdir is the population of do_Install output from D for populate_sysroot
Nov 21 19:04:13 <kergoth> you want 'image', which is D
Nov 21 19:08:30 <Dvorkin> I have a problem applying patches to yocto-linux kernel tree. I'm getting "Patch needs to be refreshed ... .git/rebase-apply/resolve_rejects". I know the patch is nice
Nov 21 19:09:21 <yates> kergoth: do build is the INPUT to do_install? (that's what i meant to ask)
Nov 21 19:09:30 <yates> s/do build/so build/
Nov 21 19:09:53 <kergoth> most likely build is B, which is where do_compile is run, so yes
Nov 21 19:10:51 <aehs29> la_croix: pip is a runtime tool, it needs to be used once the image is booted, if you want to add something at build time, there needs to be a recipe for it, and you add it just like you added python3 or pip3, creating a recipe for python packages is fairly easy, using the pypi class, you can look at the python3-async recipe for example
Nov 21 19:11:33 <la_croix> aehs29 Great, I'll take a look, thank you
Nov 21 20:09:00 <la_croix> aehs29 Sorry, I'm sure I'm being really stupid, but in the recipe you mentioned: http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-devtools/python/python-async.inc where is SRC_URI defined? Or more generally, from where does it download the code?
Nov 21 20:12:16 <bluelightning> la_croix: the pypi class sets SRC_URI based on the recipe name: http://cgit.openembedded.org/openembedded-core/tree/meta/classes/pypi.bbclass
Nov 21 20:15:11 <la_croix> bluelightning Ah, ok, thank you
Nov 21 20:27:04 <yates> i've traced this erroneous symlink into the following portion of the wx makefile: https://paste.fedoraproject.org/paste/3lEBywz7-w6PLqaZcpUg6g
Nov 21 20:27:30 <yates> but i can't see where things like $(INSTALL_PROGRAM) are defined
Nov 21 20:27:48 <yates> is that something inherited somehow from autotools?
Nov 21 20:41:47 <rburton> yates: configure finds them
Nov 21 20:41:56 <rburton> the makefile is possibly broken
Nov 21 20:44:41 <rburton> ah, hm
Nov 21 20:44:42 <rburton> no
Nov 21 20:44:51 <rburton> well done you found a bug in binconfig
Nov 21 20:44:57 <rburton> well, sort of
Nov 21 20:45:12 <rburton> it can't follow the symlink because its a target symlink
Nov 21 20:45:15 <rburton> won't work locally
Nov 21 20:45:29 <rburton> binconfig needs to skip symlinks when looking
Nov 21 20:46:06 <rburton> set BINCONFIG_GLOB to match the actual filename instead of the symlink
Nov 21 20:46:27 <rburton> it defaults to *-config so finds the symlink
Nov 21 20:46:42 <rburton> set it to whatever the value of the link target is
Nov 21 20:48:58 <rburton> yates: ^
Nov 21 21:48:36 <yates> ha.
Nov 21 21:53:42 <yates> rburton: is the sysroot-destdir directory the target of the populate_sysroot? and populate_sysroot is done after the do_install (or at the end of it)?
Nov 21 21:56:31 <yates> did you mean to set BINCONFIG_GLOB to match the symlink target filename?
Nov 21 21:57:54 <yates> so inplace-arm-fslc-linux-gnueabi-gtk2-unicode-3.0 since "./armv7at2hf-neon-fslc-linux-gnueabi/wxwidgets/3.0.4-r0/build/wx-config -> lib/wx/config/inplace-arm-fslc-linux-gnueabi-gtk2-unicode-3.0" ?
Nov 21 21:58:21 <yates> BINCONFIG_GLOB = inplace-arm-fslc-linux-gnueabi-gtk2-unicode-3.0
Nov 21 21:59:20 <yates> doh. yes, you said "the value of the link target" ..
Nov 21 22:02:39 <yates> why are the bakefiles being installed?!? https://paste.fedoraproject.org/paste/Cnaxr0aTXpMeZZGM9O4ZDQ
Nov 21 22:03:11 <yates> isn't this an install error?
Nov 21 22:07:46 <yates> time to go. i'll leave myself logged in so i can see any responses Monday.
Nov 21 22:07:53 <yates> Happy Thanksgiving everyone!
**** ENDING LOGGING AT Thu Nov 22 02:59:58 2018