**** BEGIN LOGGING AT Thu Apr 11 02:59:58 2019 Apr 11 05:51:23 New news from stackoverflow: How to setup a machine as a minor server for Yocto when fetching packages? Apr 11 08:31:45 when are the gits in $DL_DIR/git2 updated? I have some old copies there that are missing newer commits Apr 11 08:32:15 ernstp: when something references a commit which isn't there Apr 11 08:32:56 I just get fatal: reference is not a tree: 16413d71cc06b02a6d859c35a017cc49b88283f7 Apr 11 08:33:08 ERROR: ptest-runner-2.1.1+gitAUTOINC+16413d71cc-r0 do_unpack: Function failed: base_do_unpack Apr 11 08:33:50 ernstp: did do_fetch rerun? Apr 11 08:34:08 anything interesting in the do_fetch logs if so? Apr 11 08:35:28 https://www.irccloud.com/pastebin/9U5SRiOO/ Apr 11 08:36:02 ah, could have posted that inline, only two lines... Apr 11 08:37:11 hi, elfutils has dual GPLv2 and GPLv3 license for shared libraries, but poky recipe marks only LICENSE = "(GPLv3 & Elfutils-Exception)". Is this on purpose or are there doubts about the licensing? https://sourceware.org/git/?p=elfutils.git;a=blob;f=NEWS;h=5a06047f255e3c9a63828953759fd18a4ba9a3f3;hb=HEAD#l362 Apr 11 08:38:04 ernstp: I don't know, I'd probably clean it and try again, maybe save out a copy of the directory to debug it. That revision does look correct and is in the tree, its old, December 2017 Apr 11 08:39:03 mcfrisk: probably LICENSE just didn't get updated :( Apr 11 08:39:18 mcfrisk: patches welcome Apr 11 08:40:36 RP: thanks, I will propose a patch Apr 11 08:40:39 RP: probably related to corporate network stuff... can I force https:// instead of git:// ... ? Apr 11 08:41:53 ernstp: good question, that has an open bug :( Apr 11 08:41:58 I guess this is not kicking in git://git.yoctoproject.org/.* git://git.yoctoproject.org/git/PATH;protocol=https Apr 11 08:42:30 I was hoping something like that could work, if it doesn't we should find out how to tweak the fetcher to make it work Apr 11 08:42:53 right.. I can dig somemore Apr 11 08:43:07 RP: think you can find the bug #? Apr 11 09:00:06 ernstp: https://bugzilla.yoctoproject.org/show_bug.cgi?id=9738 Apr 11 09:00:07 Bug 9738: enhancement, Medium, 2.99, richard.purdie, NEW , [PATCH] fetcher: allow git+: syntax Apr 11 09:38:25 kanavin: I limited the python3 tests just to test_ssl and have confirmed it appears to hang when run under ptest-runner over ssh :/ Apr 11 09:38:49 kanavin: running it from the commandline on target over an interactive ssh session doesn't do this Apr 11 09:39:03 now the question becomes why Apr 11 10:00:11 hello, total noob here. i want to change the psplash logo, i am following this tutorial Apr 11 10:00:11 https://jumpnowtek.com/rpi/Raspberry-Pi-Systems-with-Yocto.html Apr 11 10:00:22 where do i need to add the bbappend file for psplash? Apr 11 10:22:52 lastaid: did you create your own layer for your customisations? Apr 11 10:23:28 RP: not yet. i will do that once i have this project "ready". then take the time to learn about yocto proper Apr 11 10:23:30 LetoThe2nd: sorry I wasn't around last night. Sounds like I need to learn more about twitch Apr 11 10:23:38 i am using the meta-rpi package from the example Apr 11 10:24:11 lastaid: usually you'd create your own layer to contain your customisations and you'd put the bbappend there Apr 11 10:25:03 what constitutes a layer Apr 11 10:25:46 or rather Apr 11 10:26:07 lastaid: directory with a conf/layer.conf file. Try "bitbake-layers create-layer" Apr 11 10:26:26 (as a starting point to google or command to play with) Apr 11 10:26:49 aaah. that means all the meta-x folders are layers Apr 11 10:27:00 lastaid: yes Apr 11 10:28:09 what should i choose for priority if i want my layer to overwrite the others? Apr 11 10:31:42 lastaid: I admit I can never remember which way those numbers sort :( Apr 11 10:33:02 larger value is higher priority Apr 11 10:33:28 ok, it worked Apr 11 10:35:54 now i have my layer, is this layer automatically included in my next build? Apr 11 10:37:50 lastaid: you can check conf/bblayers.conf, I think create-layer adds it Apr 11 10:38:51 did not, but i can fix that Apr 11 10:39:07 lastaid: bitbake-layers add-layer then :) Apr 11 10:39:43 if i just edit the file, will this break anything? Apr 11 10:39:48 lastaid: no Apr 11 10:39:54 ok Apr 11 10:39:56 well, depends how you edit it I guess :) Apr 11 10:40:07 but its fine to do that Apr 11 10:41:24 is there a command to add a recipe? Apr 11 10:42:08 lastaid: recipetool --help Apr 11 11:02:55 RP: no problem, Crofton and fray were very helpful Apr 11 11:11:33 RP: i mean, of course it would be helpful if you can create an account and let me know so i can add you as a moderator for YP Apr 11 11:11:42 but no hurry needed Apr 11 11:13:02 LetoThe2nd: I suspect fray is a good person to have for this! :) Apr 11 11:13:05 LetoThe2nd: will do Apr 11 11:15:23 great Apr 11 11:22:22 New news from stackoverflow: Yocto failing to fetch URL not understood Apr 11 11:31:54 RP: right, you got there first! Anything I could help with right now? Apr 11 11:39:39 RP: can i just paste this http://git.yoctoproject.org/cgit.cgi/poky/tree/meta/recipes-core/psplash/psplash_git.bb into my recipes and modify it? Apr 11 11:41:28 or just overwrite the existing one with bbappend? Apr 11 11:46:35 https://github.com/Guacamayo/meta-guacamayo/blob/master/meta-guacamayo/recipes-core/psplash/psplash_git.bbappend this is basically what i am looking for Apr 11 12:06:23 RP: I got test_ssl to hang in a direct qemu session. Just testing if it is in fact the same 5.0 kernel issue. Apr 11 12:07:15 kanavin: how did you do that out of interest? For me it seems to be a problem with the qemu remote execution Apr 11 12:07:50 RP: runqemu kvm nographic, then python3 -m tests -W from the command promopt Apr 11 12:09:02 RP, test_session_handling (test.test_ssl.ThreadedTests) ... ^CWarning -- threading_cleanup() failed to cleanup 2 threads (count: 2, dangling: 3) Apr 11 12:09:02 Dangling thread: Apr 11 12:09:03 Dangling thread: Apr 11 12:09:03 Dangling thread: <_MainThread(MainThread, started 140391056508352)> Apr 11 12:09:03 ^CWarning -- threading._dangling was modified by test_ssl Apr 11 12:09:04 Before: <_weakrefset.WeakSet object at 0x7faf552eaef0> Apr 11 12:09:06 After: <_weakrefset.WeakSet object at 0x7faf552f12e8> Apr 11 12:09:12 it looks exactly like the previous one Apr 11 12:09:19 just confirming with an earlier kernel Apr 11 12:19:58 RP: yes it is. like the previous one, succeeds sporadically for me, maybe once out of ten runs. Apr 11 12:20:16 so the existing patch needs to be expanded Apr 11 12:20:50 Hi, i want to give my unit a static IP from build. So i created an .bbappend recipe with my settings (recipes-core/init-ifupdown/init-ifupdown_1.0.bbappend). The settings gets ported i can see them in /etc/network/interfaces but it is not used. i tried to bring eth0 down and up again but its still using DHCP Apr 11 12:22:04 kanavin: I'm not seeing that behaviour. It would be interesting to disable it and see what happens on the autobuilder I guess Apr 11 12:22:27 RP: just confirmed that test_session_handling is the only one in test_ssl that hangs for me Apr 11 12:22:54 RP: I will run the full suite now to see if any hangs remain Apr 11 12:45:05 okay so i was a bit fast on jumping to conclussion, if i bring down eth0 and up again it works, but is it possible to get static IP without having to re-configure eth0 after each reboot? Apr 11 12:54:39 lol, something litterly named conman was overriding my settings Apr 11 13:22:22 Apr 11 13:22:46 New news from stackoverflow: How to setup a machine as a mirror server for Yocto when fetching packages? Apr 11 13:41:30 | failed to load "./psplash-bar.png": Couldn't recognize the image file format for file './psplash-bar.png' Apr 11 13:41:39 i get this error while trying to build psplash Apr 11 13:42:09 apparently libpng or something needs updating, how do i update the version in my recipe? Apr 11 13:55:49 otavio JaMa: hey, are your working on meta-qt5? Apr 11 13:56:06 yes Apr 11 13:57:17 does a crossbuild of qtwebengine/warrior (hosted on arm) compile for you? Apr 11 13:57:26 yes Apr 11 13:57:33 I see that the buildtask for gn fails due to missing libstdc++ :-/ Apr 11 13:58:15 T_UNIX: https://github.com/meta-qt5/meta-qt5/issues/159 Apr 11 13:58:48 I saw that. But I donot even get that far :-D Apr 11 13:58:56 gn fails to build Apr 11 13:59:50 I can confirm this comment though: https://github.com/meta-qt5/meta-qt5/issues/159#issuecomment-477895738 Apr 11 14:01:50 but since it affects the build system tools, I assume it's the same matter, right? Apr 11 14:02:27 T_UNIX: most likely yes Apr 11 14:09:22 so for PREFERRED_VERSION I have to do "1.2.3+git%" but I can do foo_git.bbappend , right? Apr 11 14:17:56 JaMa: thanks :) Apr 11 14:19:03 does "yocto" support building 32bit so the recipe could depend in `gcc-runtime-32` or whatever? Apr 11 14:21:46 zeddii, do you know off hand if we have the kfrags for ltp ? Apr 11 14:23:15 T_UNIX: not for host Apr 11 15:25:17 LetoThe2nd, https://twitter.com/sdrobertw/status/1116361369786277888 Apr 11 15:34:16 RP: am I right that stephen usually records the triage meetings? Apr 11 15:35:01 moto-timo: don't know Apr 11 15:35:36 RP: ok Apr 11 15:36:15 record as in zoom? or email sent out with ARs? Apr 11 15:38:16 * armpit hehe we should give Stephen AR to show up ; ) Apr 11 15:38:41 armpit: record as in zoom Apr 11 15:39:06 for anybody doing data mining: https://github.com/chaoss/grimoirelab-perceval Apr 11 15:39:13 the back end for bitergia Apr 11 15:41:44 moto-timo, I think there is a special user "Yocto" that is present when stephen joins and that might be the account that does the recording.. Micheal might know Apr 11 15:47:47 whatever i do, connman daemon still keeps showing up? I assumed DISTRO_FEATURES_remove = "connman" would be enough? Apr 11 15:48:08 armpit: right. as we finished the call I was noticing that special user missing. I sent stephen email. Apr 11 15:48:37 armpit: not a big deal, just wondered if that part of the organically grown process should be followed and captured ;) Apr 11 15:48:48 moto-timo, I am sending on the MH.. think he set it up Apr 11 15:49:19 armpit: great. we'll know for sure then :) Apr 11 15:50:16 willie2: i think you're confused about what distro features is. Apr 11 15:50:36 willie2: it has nothing to do with picking packages for your image. distro features controls policy, which usually affets how multiple recipes build. Apr 11 15:50:46 armpit: watch this video and dream about automatic backporting https://www.youtube.com/watch?v=KWholDv6-Rw Apr 11 15:51:23 kergoth, probably. first i tried to remove from my image, then conf/local.conf and finaly in the distro Apr 11 15:51:34 guessing at things blindly isn't going to help Apr 11 15:51:50 your image is either installing connman, or it's installing another package that depends on connman Apr 11 15:52:52 kergoth: any interest/bandwidth in upgrading recipetool to support python3? https://bugzilla.yoctoproject.org/show_bug.cgi?id=13264 Apr 11 15:52:53 Bug 13264: enhancement, Medium+, 2.8 M1, unassigned, NEW , Refactor recipetool/devtool to support python3 (only) Apr 11 15:53:18 kergoth: Well not completely blindly, I read that remove it by distro was the way to go for testing https://yocto.yoctoproject.narkive.com/zQ11Cl29/best-way-to-remove-distro-features Apr 11 15:53:28 again, packages are not distro features Apr 11 15:53:43 a distro feature is something like wayland, x11, systemd, etc Apr 11 15:53:55 they change how other recipes build, what optional dependencies they include, etc Apr 11 15:54:14 kergoth: yes, so if I remove something from it, should it not complain if I'm trying to add it? Apr 11 15:54:34 i don't understand the question Apr 11 15:54:45 removing something that isn't in the variable to begin with is going to do n othing at all Apr 11 15:54:53 connman wasn't in your distro features to begin with Apr 11 15:55:18 _remove just removes a word from a variable Apr 11 15:55:31 it's a no-op if whatever it is wasn't already in the variable Apr 11 15:55:41 and as i said like 3 times now, connman isn't a distro feature Apr 11 15:56:11 the *image* controls what packages get installed. the packagegroups installed by the image are affected by distro features, i.e. "do we support wifi" or "do we support bluetooth", but that's all Apr 11 15:56:42 if you want to remove something from an image, you need to either remove it from IMAGE_INSTALL, if it was tehre, or figure out what other package is pulling it in, most likely by examining buildhistory/ Apr 11 15:57:03 you can't just blindlye xclude a package from the build, because something else that's being installed may well require it Apr 11 15:57:44 kergoth: would that not be a very good way to see why it is getting included? Apr 11 16:00:57 willie2: Use `bitbake -g ` to get a dependency graph and then look for what's pulling in connman Apr 11 16:01:14 moto-timo: i was just looking at that this morning, i wouldn't mind working on it, but need to discuss it at work to see about resource availability Apr 11 16:01:48 kergoth: same here Apr 11 16:02:07 kergoth: perhaps we can tag team it Apr 11 16:02:38 * moto-timo dreams of meta-python3 (with zero python2 support) Apr 11 16:02:53 or similar Apr 11 16:02:56 RP: I made a fix for python3s test_locale. The test_asyncio failure is the same 5.0 issue. For test_ssl and test_httplib, you have a patch in master-next. So I think we're good, except upstream isn't reacting. Apr 11 16:03:11 I wonder if we should eventually add support for non-setup.py buildsystems for python projects and describing build dependencies. flit, pipenv, poetry, etc. would be a later enhancement, of course Apr 11 16:03:42 I have wondered the same Apr 11 16:04:34 the plethora of python build systems is...um...a challenge? (I'd like to say unnecessary ;) Apr 11 16:04:47 personally I use pipenv quite a bit Apr 11 16:04:58 agreed. the lack of consensus is problematic. even if things move to e.g. pyproject.toml that doesn't help, sicne every tool gets its own section of the damn file :) Apr 11 16:05:07 at least it's easier to parse than setup.py ast crap Apr 11 16:05:35 yeah. ast. It always makes my head hurt. Apr 11 16:06:38 ok. we can either add discussion on that bug or start a thread on ML. either way it would be fantastic to give recipetool/devtool a refresh on python support Apr 11 16:07:09 * moto-timo goes afk for a bit Apr 11 16:09:11 moto-timo, why not just do those in java ; ) Apr 11 16:55:31 kanavin: do I understand correctly that 5.0 makes an async io test fail? Apr 11 17:30:36 I built an image with postgreql. I see postgres in /usr/bin, but can't find psql. An ideas where that is? Apr 11 17:34:58 I have a quick Q about organization. Do you usually keep all your layers inside the poky directory? Seems like it might be good to keep poky unmodified. Thoughts? Apr 11 18:08:56 It seems like most projects are setup with their layers inside the poky directory. Apr 11 18:41:30 rabbit9911: it depends how you want the setup there are tools to do it either way Apr 11 18:46:37 most projects I'm involved with don't use poky Apr 11 18:51:00 I use foo/build foo/bitbake foo/oe-core foo/meta-openemebdded etc Apr 11 18:51:26 so I do not need to modify any git repos, unless I need to make changes I need to send upstream Apr 11 19:08:36 armpit: nah. smalltalk and lisp Apr 11 19:11:44 Tartarus, is meta-soho on github or something Apr 11 19:12:08 Crofton: no, as it's intended to be done while walking through and you put in your actual details Apr 11 19:19:07 I'm having a more stupid problem, not getting hostapd configed Apr 11 19:19:19 it does broadcast though Apr 11 19:24:15 I did wave my hands on configuring hostapd itself as there's guides on that :) Apr 11 19:24:47 and it matters which hw it is, for some options Apr 11 19:30:01 there seems to be a ink getting the ip aaddr set for wlan0 Apr 11 19:30:09 I can make it work by setting manuallly Apr 11 19:52:58 kanavin: success, python3 seems much happier Apr 11 19:53:09 kanavin: now just an intermittent e2fsprogs Apr 11 19:53:32 kanavin: I ran master-next and warrior-next ptest builds and got the same results between them which was nice Apr 11 19:54:55 Crofton: OK, so, hold on :) Do you mean the bridge you're putting wlan0 on? Apr 11 19:56:18 the guide I ran into said to give wlan0 a static ip Apr 11 19:56:42 fo rmy app, I want to use a phone to connect to the devce as a screen, no other conenctivity Apr 11 19:56:56 Crofton: you trust the internet? Apr 11 19:57:03 No Apr 11 19:57:26 Crofton: I was going to say, you of all people... ;-) Apr 11 19:58:33 * RP can't do stablisation and release at the same time, this making my head hurt :( Apr 11 20:15:06 of course I don't mean that, which just goes to show how fried I am :/ Apr 11 20:17:24 Everyone desires a stable release, but sometimes it is little appreciated as to the amount effort required. Apr 11 20:18:14 jwessel: everyone seems to be going full steam on version upgrades and things I can't take into warrior. That is what is making my head hurt as I have two lots of patches to keep track of Apr 11 20:19:29 Which is why I figured you'd be releasing sooner than later. So you can "just fork it" as they say... Apr 11 20:20:33 jwessel: it gets released when the release blocking issues get fixed Apr 11 20:39:03 RP, just what I have to go one, endless sea of embedded nonsense hacks. I discard all the ones that start with, install poky :) Apr 11 20:39:41 Crofton: bit below the belt? ;-) Apr 11 20:51:47 RP we put the upgrade tool into the masses hands.. what you expect? Apr 11 20:56:10 armpit: heh :) Apr 11 20:56:28 * RP ponders the e2fsprogs timeout. I think that really is just a slow test Apr 11 20:58:20 have you timed in qemu ? Apr 11 20:58:33 armpit: guess what is running in the terminal Apr 11 20:59:30 is it the ptest or runtime ? Apr 11 20:59:46 armpit: its the e2fsprogs ptest. I think there is a slow individual test Apr 11 21:00:30 Crofton: Ah, if you aren't bridging then yeah, ok, that makes sense. And then yes, hostapd won't give wlan0 an IP, that's not its job. systemd/whatever can do that :) Apr 11 21:00:46 I have slower systems than the AB so I can fire on a simple test at home Apr 11 21:01:10 Tartarus, and it fials miserable at its job Apr 11 21:01:23 bot do it by hand and poof, works Apr 11 21:01:31 Crofton: systemd? Apr 11 21:01:38 RP, I am almost done with runtime ltp with json results Apr 11 21:01:40 what do your network units look like? Apr 11 21:01:47 armpit: I'm just trying /usr/lib/e2fsprogs/ptest/run-ptest | while IFS= read -r line; do printf '[%s] %s\n' "$(date '+%Y-%m-%d %H:%M:%S')" "$line"; done Apr 11 21:02:01 armpit: that sounds good, is that into ptest format? Apr 11 21:03:11 in the scene each LTP is listed separately with pass/fail.. there is no timing info Apr 11 21:03:39 https://paste.debian.net/1077192/ Apr 11 21:04:04 armpit: are they listed as separate tests or separate ptest results? Apr 11 21:04:35 armpit: timing info is nice but not essential Apr 11 21:04:52 the syscalls in their individual test fork etc.. the standard ltp human readable format Apr 11 21:05:10 its bloody long Apr 11 21:05:22 also qemu will need more space Apr 11 21:05:24 armpit: I don't think I follow what you mean :( Apr 11 21:07:24 Crofton: OK, I think /24 on the end is valid, but I don't usually do that. And I don't know if we build systemd such that the build-in dhcp server works either, honestly. Apr 11 21:07:32 RP, do you mean the test naming convention in the json file? Apr 11 21:07:41 then yes Apr 11 21:07:50 the raw logs are like: Apr 11 21:07:54 Testcase Result Exit Value Apr 11 21:07:54 -------- ------ ---------- Apr 11 21:07:54 abs01 PASS 0 Apr 11 21:07:54 atof01 PASS 0 Apr 11 21:07:54 float_bessel FAIL 134 Apr 11 21:07:55 float_exp_log FAIL 134 Apr 11 21:08:23 what one can do on long flights Apr 11 21:08:36 armpit: so are you turning this into a ptest or something else? Apr 11 21:08:37 the built server seems to work, just can;t get it to set address, with or without th e24 Apr 11 21:08:41 next week Apr 11 21:08:49 cow field weekend Apr 11 21:10:34 RP, sort of.. On main executor that loops through a list of test group.. I did not launch the test all Apr 11 21:11:06 armpit: ok, so we could trigger this through ptest-runner? Apr 11 21:11:19 so each group with run and then I grab the log. Apr 11 21:11:20 no Apr 11 21:11:48 I launch "runltp" Apr 11 21:12:17 armpit: from where? :) Apr 11 21:12:26 armpit: this is a new runtime test then? Apr 11 21:12:27 ssh Apr 11 21:12:36 yes, a new runtime Apr 11 21:12:50 armpit: I'm wondering how this integrates into our automated testing Apr 11 21:13:24 armpit: so a new lib/oeqa/runtime/cases/ltp.py ? Apr 11 21:13:29 correct Apr 11 21:13:40 armpit: out of interest why that and not a ptest? Apr 11 21:14:17 its not a ptest Apr 11 21:14:27 its a system test Apr 11 21:15:01 armpit: My worry is more how we process the results Apr 11 21:15:42 armpit: I'd have thought ltp would fit our ptest model of "pass/fail/skip" stats and compare for regressions rather than our runtime model of pass/fail Apr 11 21:15:43 oh, well then that would be the problem for the person sending the patch Apr 11 21:16:18 armpit: you're not planning to send a patch? Apr 11 21:16:36 ltp is still running 5 hours so far.. you really want that bundled into ptest? Apr 11 21:17:00 yes, I am .. so I would have to make sure the test out it in compliance Apr 11 21:17:33 armpit: I think the results processing model matches ptest. How we trigger slower tests, good question Apr 11 21:17:44 "ltpresult.math.abs01": { Apr 11 21:17:44 "status": "PASSED" Apr 11 21:17:45 }, Apr 11 21:17:45 "ltpresult.math.atof01": { Apr 11 21:17:45 "status": "PASSED" Apr 11 21:17:52 json format Apr 11 21:18:19 I think we may need to tweak that to look like a ptest result, then the ptest result processing code would handle it Apr 11 21:18:35 (it doesn't matter if really was a ptest or not) Apr 11 21:18:57 its a detail we can figure out Apr 11 21:19:15 I'm just thinking out loud, I would love any step forward in getting ltp results into our system Apr 11 21:19:30 * RP really isn't trying to be negative Apr 11 21:19:31 RP, that is what I modeled it after Apr 11 21:20:19 armpit: there is magic in resulttool which understand ptest specifically. We either teach result tool about ltp|ptest or put it in that format Apr 11 21:23:59 that is the part I am going to look into next. I think I might need to have "section" defined.. Apr 11 21:24:29 armpit: possibly, yes Apr 11 21:24:53 if we get ltp figured out then the manual compliance test will follow Apr 11 21:25:30 its neat to see this Apr 11 21:25:34 "ltpresult.cve.cve-2011-0999": { Apr 11 21:25:34 "status": "PASSED" Apr 11 21:25:34 }, Apr 11 21:26:41 yes! :) Apr 11 21:27:01 but not this Apr 11 21:27:04 "ltpresult.cve.cve-2017-5754": { Apr 11 21:27:04 "status": "FAILED" Apr 11 21:27:04 }, Apr 11 21:30:59 so I am hoping to have a section in the result tool like "There was no ptest data" if ltp is not included Apr 11 21:31:38 armpit: I still think ltp is a special case ptest Apr 11 21:32:31 oh, you mean like have it included under the heading of "ptest" Apr 11 21:32:39 armpit: yes Apr 11 21:33:23 ah... Ill take a look that would make resulttool integration simpler??? Apr 11 21:33:30 very Apr 11 21:33:57 it would likely "just work" Apr 11 21:34:19 yeah.. wonder if I can just append to that section Apr 11 21:34:29 * RP doesn't see why not Apr 11 21:35:26 I even have the failures save off like ptest does so the get printed at the end Apr 11 21:36:16 testseries | result_id : oeqa | runtime_core-image-minimal_qemux86_20190411141132 Apr 11 21:36:16 ltpresult.math.float_power Apr 11 21:36:59 armpit: if you put them in as ptest results it would just do that too Apr 11 21:37:15 armpit: do you inject the failure logs? Apr 11 21:38:43 I parse them separately like ptest does and call self.fail() Apr 11 21:39:19 armpit: injecting logs is a later step Apr 11 21:40:10 well then its magic.. I don't do anything special Apr 11 21:40:44 I need to see what happens if I have a ptest and ltp both included Apr 11 21:41:14 armpit: right, you may need to teach them to be nice to one another Apr 11 21:42:06 did you mean the log file under logs/oeqa/ptest* ? Apr 11 21:42:43 I place ltp in its own Apr 11 21:43:41 armpit: I mean the extras['ptestresult.rawlogs'] = {'log': output} Apr 11 21:44:23 oh, Apr 11 21:44:53 hmm, I do capture the raw logs but the format is completely different Apr 11 21:45:16 I let ltp generate the human readable format Apr 11 21:45:22 armpit: We just inject the logs into the json output so it can be debugged later Apr 11 21:45:41 k, got that info Apr 11 21:45:49 armpit: there is code somewhere which splits the logs per ptest Apr 11 21:46:03 armpit: can't quite spot it atm :/ Apr 11 21:46:04 the logparser Apr 11 21:46:13 armpit: yes! Apr 11 21:46:14 core/utils/ ? Apr 11 21:46:32 I have add a class there for ltp Apr 11 21:47:10 ok, makes sense Apr 11 21:47:24 I did plan to drop the rawlogs once I was convinced the parser worked Apr 11 21:48:46 * RP notes that skipping the slow e2fsprogs tests makes things a lot happier Apr 11 21:49:08 the ltp raw logs don't mark "FAIL or ERROR" so they are useless to parse Apr 11 21:49:38 armpit: right, something to mention to upstream perhaps Apr 11 21:50:00 they have a human readable flag Apr 11 21:50:20 it puts them in the listing i posted earlier Apr 11 21:50:23 armpit: ok, fair enough Apr 11 21:51:27 the raw log shows the test # within the test, ie fork10 may have 6 testcases Apr 11 21:52:03 so fork10-1 fork10-2 etc Apr 11 21:52:26 right, makes sense Apr 11 21:52:49 as long as its something which can be traced back to a test within ltp its fine as a name :) Apr 11 21:53:58 * RP thinks we may now have all the release blocking bits in warrior-next Apr 11 21:56:21 * armpit 6 hours and still running Apr 11 21:56:26 Whats a good method to say collect and upload symbols for select packages? Thinking maybe making bbclass and adding to USER_CLASSES. Maybe add variable to append package names to which I want to collected debug symbols for. Apr 11 21:57:33 armpit: assuming a 10min/mile pace thats well over a marathon distance, you should probably stop ;-) Apr 11 21:58:12 rabbit9911: that could certainly work Apr 11 22:12:58 * armpit ltp is a special case ptest which itself is a special case runtime test Apr 11 22:31:50 armpit: special cases all around! Apr 11 23:13:55 * armpit ah.. more work needed.. Apr 12 00:09:42 i still have memories of LTP test failures Apr 12 01:23:53 seebs, thats why its "Linux Test Problems" ; ) Apr 12 02:11:21 i think that was the test suite that had the floating point test that depended on guesses about what 1/400 is as a binary float. Apr 12 02:23:02 I'm trying to add meta-python layer to my yocto image and getting a 'error parsing layer configuration': Apr 12 02:23:04 https://www.irccloud.com/pastebin/9iJkzJn1/ Apr 12 02:23:07 any ideas? **** ENDING LOGGING AT Fri Apr 12 02:59:57 2019