**** BEGIN LOGGING AT Mon Sep 02 02:59:56 2019 Sep 02 03:17:44 can an yocto image be build for x86 or simulated x 86 on qemu Sep 02 05:38:06 I just watched the live coding video which talked about package splitting. In the FILES_xxx variable, it listed files as per the install path, i.e. "/usr/bin/ask". Is that referring to the path where the package would install its files in a normal installation? i.e. the "local" sysroot where the package gets built inside bitbake? Sep 02 05:39:23 iceaway2, yes I think so Sep 02 05:40:00 if I understood you q correct Sep 02 05:40:37 LetoThe2nd: Maybe you got feedback on this already, but in the libanswer recipe file, you used PACKAGES =+ instead of what I am used to seeing +=. DOes that yield the same result or is there any difference between += and =+? Sep 02 05:41:44 kroon: As I typed the question it made more and more sense to be that would be the case. How otherwise would bitbake know how to find the files? Sep 02 05:45:04 iceaway2, += appends, =+ prepends Sep 02 05:55:52 kroon: thanks for the clarification! Sep 02 05:57:48 iceaway2, wether or not that makes a difference to how files are packaged, I dont know.. maybe order plays a role where files go Sep 02 06:43:18 good morning Sep 02 06:54:33 iceaway2: order matters for packages. The same file cannot be in two different packages. The first package whose FILES_[package] contains said file will get it, all the others, not. This gets "tricky" when you have files that can be matched by multiple FILES_[package] patterns. Sep 02 06:54:47 in PACKAGES, left most is done first Sep 02 07:05:02 qschulz: I suppose by same file the path must be the same as well? I mean /bin/example and /usr/bin/example could co-exist and be pulled in be different recipes? Sep 02 07:06:03 Will bitbake give a warning if multiple packages share the same filename? Sep 02 07:14:22 iceaway2: yes, by same file I meant same full path Sep 02 07:16:03 iceaway2: I don't think so, tha's already used for the -dbg package which is packaged before the "normal" package. The "normal" package has the paths which would include the debug files (with .debug in the path) IIRC and you don't see warnings for those. I don't even think there is a NOTE for it. Sep 02 07:16:40 but if at one point you don't understand why your package does not have this or that file or that Yocto complains about your package being empty, it may have something to do with that Sep 02 07:28:11 Keeping that as a mental note, should it come up in the future. Sep 02 07:49:44 note: its a great feeling when you arrive in the morning and the questions for yourself have already been answered by other helpful folks :) Sep 02 07:54:17 LetoThe2nd: LOL Sep 02 07:54:53 mckoan: makes me feel famous! Sep 02 08:09:18 LetoThe2nd: you are! Sep 02 08:10:54 throw money! THROW MONEY! Sep 02 08:11:02 (no coins, please.) Sep 02 08:12:19 none of these nice heavy one pound coins? ;) Sep 02 08:13:19 hopefully not. Sep 02 08:14:07 there are even better ones :) Sep 02 08:14:17 florian: https://www.youtube.com/watch?v=9gjvU913aa4 Sep 02 08:14:45 *g* Sep 02 08:14:54 so, no coins please. Sep 02 08:15:12 hehe Sep 02 08:15:48 also relevant: https://www.youtube.com/watch?v=viDL2W0HcJw Sep 02 08:16:33 LetoThe2nd: I think the heaviest one I have access tho would be one of these: https://en.numista.com/catalogue/pieces13370.html Sep 02 08:18:03 :) Sep 02 08:19:55 florian: i actually think i have a 250 or 500 schilling somewhere. Sep 02 08:28:48 :-) Sep 02 08:31:52 So I've this package depending on openssl10 and other packages. But those packages are pulling openssl (not openssl10) as dependencies. Then obviously do_prepare_sysroot isn't really happy about deploying static libraries and headers at the same place Sep 02 08:32:15 is there anyway to hack around this thing? (we don't have the sources for this package requiring the use of openssl10) Sep 02 08:35:07 qschulz: can those "other packages" use openssl110? in that case, maybe you can use it as the rpvoider Sep 02 08:35:10 provider Sep 02 08:43:23 I wouldn't, because in the end it's going to be the whole system using openssl10 Sep 02 08:43:52 exactly. Sep 02 08:44:17 (i don't have much experience with openssl, just tossing ideas around) Sep 02 08:48:44 qschulz: we had to create a binary only openssl10 recipe without any header files etc. The pure .so files don't conflict with openssl 1.1.x. Sep 02 08:50:38 alternative, one could remove all but .so output from openssl10 recipe. then it would not conflict with openssl. But do check at runtime that processes are not loading both openssl10 and openssl (1.1.x) shared libraries. I'm not sure but there may be symbol resolution conflicts between the two. Sep 02 08:52:59 mcfrisk: we need the header files of openssl10 as well IIRC. It's a 50/50 recipe with 50% having to be compiled and the other half being pre-compiled. Then we link the part we are building against the pre-compiled libraries, which incidentally were compiled with openssl10 Sep 02 09:01:23 New news from stackoverflow: Compilation database for BitBake (Yocto) builds Sep 02 09:03:54 qschulz: then you need to convert the full dependency chain to use openssl10. but I would rather try to switch to openssl 1.1.x completely. going half way may not work at runtime. I know it's hard with binaries from suppliers but still.. Sep 02 09:04:15 * LetoThe2nd advocates threatening suppliers. Sep 02 09:04:58 in the end, it's all about the money and contract details.. Sep 02 09:12:10 mcfrisk: right now, it's more about the delay. We REALLY want to upgrade to thud but the supplier's going to take weeks at least, maybe months to recompile Sep 02 09:13:01 mcfrisk: but it turns out, maybe I was wrong and we're not needing headers after all, so we'll test your suggestion of pure binary recipe until we get the new library from the supplier Sep 02 09:14:21 qschulz: good luck :) Sep 02 09:47:16 whats the currently recommended best practise to install cronjobs into an image? RTFM pointers appreciated! Sep 02 09:49:00 overwriting the cronie crontab by append most certainly works, but feels clumsy Sep 02 09:56:35 Hi Sep 02 10:00:12 A question about recipe location. It seems if there are several recipes for the same package (but with different version), they have to be located at the same relative path of their layer. Is it true and why is it needed ? Sep 02 10:00:58 ykrons: that is a best practice to have a clean directory tree Sep 02 10:01:47 ykrons: BTW you could put them wherever you want if you specify the proper wildchars pattern in layer.conf though Sep 02 10:02:15 mckoan, thanks. Ok so I have probably an issue with my layer.conf. Sep 02 10:02:47 I have checked already but it seems depending on my location the recipe is sometime not seen Sep 02 10:09:42 how can i remove a compiler flag in a bb script ? Sep 02 10:09:55 *bbappend script Sep 02 10:14:34 RP: thanks, seems like the eglinfo recipe removal was overlooked or somehow missed? it's here if you want to cherry-pick http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=akanavin/package-version-updates Sep 02 10:19:59 MarcWe1: probably depends on the build method used by the package, but you can override do_compile or play with EXTRA_OEMAKE or CFLAGS Sep 02 10:20:52 kanavin_: that patch somehow caused problems (mailing list corrupted?) and I didn't have time to figure out why over the weekend Sep 02 10:29:09 mckoan: Tnx going to look into that Sep 02 10:42:13 RP: that's why it was sent in a pull request with a link to git repo, you probably need to cherry-pick from there. The patch removes a file that is a mixed ascii python code/binary blob, so it was likely corrupted by smtp. Sep 02 11:06:10 mc Sep 02 11:11:56 how could i parse out all the package names from all the layers without building the actual recipes? Sep 02 11:13:40 weltling: first thought: bitbake -g depexp world, and then parse the .dot files. warning, might be huge! Sep 02 11:16:08 LetoThe2nd, thanks, gotta give it a try Sep 02 11:16:37 LetoThe2nd, my goal is to build all those and pump them into a repo, sure should be huge Sep 02 11:17:15 weltling: you can do bitbake world, it's going to compile everything Sep 02 11:17:25 weltling: ah its only bitbake -g. no depexp needed. Sep 02 11:18:05 BUT, beware that some layers have bbappend or change the default PACKAGECONFIG or settings, etc. Sep 02 11:18:27 weltling: is there anything specific you have in mind that requires you to do that? Sep 02 11:19:14 weltling: it depends on your use case, but if you want a repo server then the world target is probably worth thinking about. as well as a pr server Sep 02 11:20:42 qschulz, yep, keep the image small, but be able to customize for a specific case, that's the approach Sep 02 11:21:13 qschulz, for the initial filling in the repo, single package updates would then require sstate cache probably Sep 02 11:21:43 LetoThe2nd, what is a pr server, if i may ask? :) Sep 02 11:21:48 weltling: soudns very much like you want to read the chapters on packge feeds in the dev manual, in depth. Sep 02 11:22:22 weltling: pr server is the magic that makes package management from a feed work, basically. Sep 02 11:22:39 LetoThe2nd, currently using aptly, the package feed seems a very simplistic approach :( Sep 02 11:23:08 weltling: for reference: https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/package-manager-white-paper.pdf Sep 02 11:23:15 kanavin_: right, I just didn't have the time to figure that out :) Sep 02 11:24:07 weltling: you can totally run a full blown distribution grade package feed using yocto. but it requires a little more than just "building" Sep 02 11:26:11 LetoThe2nd, fe i couldn't find how the package feed can be authenticated with a repo key, or the way it puts packages there without ability to group them, etc. Sep 02 11:26:41 LetoThe2nd, it also depends on teh requirement of my org to be able to push into a separate repo, using a custom api, etc. Sep 02 11:27:04 LetoThe2nd, thanks for the link, good read Sep 02 11:27:34 weltling: in the end its only software, so everything is possible. yet i can't help with details, as i do not use runtime PM. Sep 02 11:28:53 LetoThe2nd, yeah, the most uses are immutable image builds, i know Sep 02 11:29:19 LetoThe2nd, many thanks for the leads, seems i've got my daily bread today to dig further :) Sep 02 11:29:31 weltling: have fun Sep 02 11:38:14 arg, im getting unrecognized command line option '-fstack-clash-protection' sins armhf is not able to compile it with this option. Sep 02 11:38:37 but i cant get disable it Sep 02 11:40:13 * can figure out the thing that is setting it Sep 02 11:40:27 *cant Sep 02 11:53:40 bitbake -e may be able to help you with figuring out which class or inc or bb or bbappend is setting it? Sep 02 11:55:43 tnx Sep 02 11:55:45 RP: I managed to get the binutils tests as a recipe independent from the binutils-cross buildir. Only issue is the gold tests and the libiberty tests. However for the gold tests I realised it was not even running them properly in the -cross recipe anyways. For libiberty, its not the same one that is shipped on target or -native, so the value provided by those results were limited anyways. Sep 02 12:01:02 RP: i also have some questions about how to handle the selftest tagging, when you are free Sep 02 12:07:27 nrossi: I'm around Sep 02 12:07:58 libiberty is an odd one as we build it in several places. Tempted not to worry about that atm Sep 02 12:09:18 RP: so firstly, the existing decorator code in the core stuff does not handle decorating classes at all, and the test filter codepaths appear to be unused. Should i still try to get this tag decorate working using those pieces as the base? or just start fresh to make it simpler? Sep 02 12:09:57 nrossi: I like simpler, if that code is unused we should probably remove it too Sep 02 12:10:16 nrossi: unused as in we don't use the commandline options or that the functions are unused? Sep 02 12:10:36 RP: for the filtering/execution side, should it be similar to "-r", e.g. "-t " only runs the tests with the tag (or multiple tags), or should it be a filter, e.g. "-a -t " runs all the tests except Sep 02 12:10:59 RP: unused in the sense that the classes are never created outside oeqa/core/ Sep 02 12:11:39 nrossi: I'd be tempted to have a "run all matching" option and a "run all not matching" option Sep 02 12:13:04 RP: should it handle multiple tags, e.g. "run all matching foo or bar", "run all not matching foo or bar"? Sep 02 12:13:49 RP: just trying to figure out the logic for when a test case has multiple tags. and how to run that Sep 02 12:20:33 nrossi: I think it should, yes Sep 02 12:20:42 and "or" makes more sense than "and" Sep 02 12:21:10 RP: ok, i wil try to keep it simple. It can also be reworked to fit changing requirements later Sep 02 12:21:34 nrossi: sounds good. Some of the QA code does need cleanup, particularly in the filtering area :/ Sep 02 12:21:53 RP: oh and something i keep forgetting to query. dejagnu recipe is currently in meta-oe should it be moved to oe-core? Sep 02 12:22:03 nrossi: if we depend on that, yes Sep 02 12:22:30 RP: ok, will add that to the v2. I am just working on the tag stuff and once done will send out a v2 Sep 02 12:22:39 nrossi: sounds good, thanks Sep 02 12:22:58 * RP is going to try and sort the hash equiv bits now the kernel is mostly done Sep 02 12:23:05 need to debug qemumips first though Sep 02 12:30:37 i can finde any output of -fstack-clash-protection being set in the output of bitbake -e (geoclue) Sep 02 12:30:48 * i can't finde any output of -fstack-clash-protection being set in the output of bitbake -e (geoclue) Sep 02 12:37:57 MarcWe: bitbake -e geoclue | less Sep 02 12:38:08 MarcWe: then /clash :) Sep 02 12:42:44 tnx. but pattern not found :( Sep 02 12:43:14 MarcWe: maybe the makefile is adding the flag itself? Sep 02 12:43:27 MarcWe: Where do you expect it to come from? http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/distro/include/security_flags.inc doesn't specify -fstack-clash-protection? Sep 02 12:43:48 Ah, sorry, just now read your message above. Sep 02 12:44:09 MarcWe: in that yase, you rather have to look at the build process of the package if it is being added there Sep 02 12:44:37 grep -r in geoclue git did not bring it up Sep 02 12:46:22 neverpanic: its not in that securit_flags.inc Sep 02 12:49:55 MarcWe: are you struggling with the upstream recipe at http://cgit.openembedded.org/meta-openembedded/tree/meta-oe/recipes-navigation/geoclue/geoclue_2.5.3.bb?h=master ? Sep 02 12:52:06 no im on rocko Sep 02 12:52:15 thats the strange part Sep 02 12:53:21 .... my sleepy head still had oe-core on rocko-next Sep 02 12:53:50 that did not seem to fix the problem to reset it to rocko Sep 02 12:59:00 MarcWe: what other layers are in use? maybe a custom distro or machine config pulling it in? Sep 02 13:01:40 im using arago ( and a ti arm processor one) Sep 02 13:02:18 i personally would try digging there. at least from my POV there's nothing suggesting that flag in poky nor meta-oe Sep 02 13:02:42 k Sep 02 13:04:14 it seems that it is fixed in gcc for arm 7.5+ so an upgrade to thud also seems interesting Sep 02 13:05:10 but the stange part is that i can run the os only geoclue won't compile Sep 02 13:08:32 stateless-rootfs IMAGE_FEATURE is missing some documentation in poky. just noticed I had to enable it to recreate old rootfs images with new poky version. Sep 02 13:36:20 * alessioigor waves all Sep 02 13:36:22 I have just published my first layers so I would like to thank RP, mckoan, LetoThe2nd and rburton for all help and support! Thanks guys! Sep 02 13:38:43 time to celebrate! https://www.youtube.com/watch?v=roRQ2mNwMMQ Sep 02 13:48:12 alessioigor: great news! congrats! Sep 02 13:48:24 does anybody have a quick example for placing a symbolic link in /etc/cron.daily or comparable? I'm just being mondayishly confused. Sep 02 13:48:50 do_install() { ln -s .... } Sep 02 13:49:05 rburton: thats the obvious part. Sep 02 13:49:14 remember the target should be of the form you want on the target, so don't put $D in the target Sep 02 13:49:24 erm, too many uses of 'target' Sep 02 13:49:27 but you know what i mean right Sep 02 13:50:09 rburton: check Sep 02 13:53:04 rburton: *sigh* seems like ln was being angry because /etc/cron.daily wasn't already present. Sep 02 13:53:42 yeah remember $D is empty in do_install, mkdir first Sep 02 13:53:48 have i already mentioned how much i hate doing custom do_installs? i always mess them up. Sep 02 13:54:00 each and every time. Sep 02 13:54:08 alessioigor: cool :) Sep 02 13:54:46 :) Sep 02 14:20:51 ok another nice one. i'm trying to run a rather simple cifs share on an arm platform. testbed is qemuarm at the moment (krogoth unfortunately). when I'm doing systemctl start nmb, it locks up for 60s, then returns. the nmb log says that it happily did start, but then systemd seems to decide it timed out and kills it. any pointers? Sep 02 14:22:45 (at the moment setting up a warrior build to cross-check there) Sep 02 14:23:24 LetoThe2nd: "TimeoutSec" on the .service unit? unless you want to fix the cifs issue :P Sep 02 14:25:15 LetoThe2nd: you can do it with "/etc/systemd/system/nmb.service.d/.conf" drop ins Sep 02 14:25:29 nrossi: doesn't change anything. it rather seems to be some problem with systemd giving a false positive on timeout. Sep 02 14:26:04 LetoThe2nd: is nmb systemd aware? or just a simple daemon? Sep 02 14:26:20 nrossi: good question. Sep 02 14:27:03 nrossi: in fact i'd expect it to be, the recipe already happily ships with unit files and all Sep 02 14:29:02 the recipe ships its own service files, or it uses upstream ones Sep 02 14:29:13 never underestimate the ability for people to submit untested patches Sep 02 14:29:25 LetoThe2nd: yep, looks like a proper systemd aware service at least looking at upstream samba source Sep 02 14:30:14 rburton: should be upstream AFAICS Sep 02 14:30:53 LetoThe2nd: have you tried TimeoutStartSec=? Sep 02 14:34:43 nrossi: not yet. just kicked it off Sep 02 14:36:54 -> https://hastebin.com/inoyifiyun.coffeescript Sep 02 14:38:13 thats what i mean by "starts up happily, then gets killed." Sep 02 14:39:23 LetoThe2nd: I assume that you have checked that the value is being applied with systemctl cat? Sep 02 14:40:29 nrossi: to my understanding it is: https://hastebin.com/malanuzeba.makefile Sep 02 14:41:49 just waiting for the warrior qemu build to finish Sep 02 14:41:58 zeddii: should poky.conf be setting 5.2 instead of 5.0? Sep 02 14:42:39 LetoThe2nd: what about the value in "systemctl show nmb"? Sep 02 14:43:35 nrossi: TimeoutStartUSec=1min Sep 02 14:44:28 LetoThe2nd: try "infinity" to disable the check, see if that lets it work Sep 02 14:45:08 nrossi: so TimeoutStartSec=infinity? Sep 02 14:45:24 LetoThe2nd: yer Sep 02 14:46:58 nrossi: kicked it off, so far systemctl start is blocking Sep 02 14:51:20 blocks foreve, just as expected. Sep 02 14:51:39 LetoThe2nd: sure, but it hasn't killed the service right? Sep 02 14:52:07 LetoThe2nd: you should be fine to ctrl+c the start command Sep 02 14:52:11 nrossi: doesn't look like it Sep 02 14:53:41 LetoThe2nd: if the service is still running, try setting the "TimeoutStartSec=600" or similar 60s is probably too short Sep 02 14:57:21 nrossi: i already tried that, but after all that time it will be killed anyways. Sep 02 14:57:39 so the point seems to be: why does systemd think the thing failed? Sep 02 14:58:37 LetoThe2nd: ah ok, then its the service itself not telling systemd it has started. If you systemctl status it, it will probably still say starting Sep 02 14:59:20 nrossi: hum no? Loaded: loaded, Active: inactive: dead Sep 02 14:59:50 LetoThe2nd: no when you have the timeout value set to infinity Sep 02 15:02:10 nrossi: sorry i got confusef over the several things i tried. let me give it a clean restart Sep 02 15:02:39 LetoThe2nd: also this commit on samba looks interesting ... https://github.com/samba-team/samba/commit/440ddf8470b11a46066d282bf8945201d547c192 Sep 02 15:03:11 nrossi: very much so, indeed Sep 02 15:04:00 nrossi: anyways its hopefully only a couple of moments until warrior is ready, and if it works there it hopefully just comes down to some backporting Sep 02 15:23:07 nrossi: on warrior it works almost out of the box. so seems to be an already-fixed thing and backporting might save me. thnks for your input! Sep 02 15:24:07 LetoThe2nd: good to know :), np was just curious myself Sep 02 15:24:45 nrossi: i'll let you know if i find something else interesting or am finished, whichever happens first :) Sep 02 15:32:36 New news from stackoverflow: Yocto Bitbake doesn't include kernel config fragment in build Sep 02 15:41:37 RP, want me to bug the arm64 build issue for " devtool build v4l2loopback-driver" Sep 02 16:04:11 nm, just opened 2 **** BEGIN LOGGING AT Mon Sep 02 16:05:25 2019 Sep 02 16:11:23 armpit: thanks, is zeddii ccd? Sep 02 16:19:31 zeddii: multiple issues if we make 5.2 the default, kernel module and stap related :/ Sep 02 16:42:36 <__angelo> hi. is it possible to apply a patch in a package only on a specific kernel verison ? Sep 02 17:56:17 __angelo, do you me if kernel version is 4.19 apply patch? if that is what you mean, try bbappend for the affected kernel version Sep 02 17:57:32 <__angelo> armpit, well, in recipes-kernel i have a separate module recipe, so it's a different recipe Sep 02 17:58:38 <__angelo> wondering if i can check "PREFERRED_VERSION_linux" someway Sep 02 18:06:10 __angelo, uh... interesting uses case.. you may want to hit the mailing list. Sep 02 18:27:26 __angelo: if you're patching a kernel module then the module code can check the version. if you're patching an app then check the running version for whatever reason you need to know the version for Sep 02 18:28:30 <__angelo> rburton, oh thanks. I am patching a kernel module, but it's done on a separte recipe. How could i check the kernel version ? Sep 02 21:18:37 * armpit knew someone had the answere Sep 02 21:25:06 RP: yes. I have a patch to poky to make 5.2 the default, and another to remove 5.0. What’s the kernel module issue ? I can build them on all the arches without issues. stap was working for me .. but are you seeing a failure on all arches, or just a particular one ? Sep 02 21:29:24 zeddii, see 13497 and 13498.. Sep 02 21:29:39 I hope I put enough info in the bugs Sep 02 21:30:00 I’m not actually back to the office until Wednesday or Maybe thursday, so I won’t actually get to work on anything until then. Sep 02 21:30:12 not sure which RP is referring to Sep 02 21:31:11 zeddii, k.. I am not feeling the pressure ; ) Sep 02 21:32:06 for 13498 .. there was a request to ditch the kernel module it was using. but I don’t use devtool, so I have no idea on that one. hello world works fine from the skeleton dir :D Sep 02 21:32:42 armpit: but FWIW, if there’s an individual test case that can be pointed at for those bugs, it is better. Sep 02 21:32:57 since I don’t run the test images, I have to track down the case, and figure out how to test it manually. Sep 02 21:33:03 but from those logs, I’m not sure I could find them. Sep 02 21:33:25 looks like sdkext, I guess that’s enough. Sep 02 21:34:35 zeddii, let me look Sep 02 21:36:25 * armpit googles for uk # Sep 02 21:37:02 the kprobes one looks harder to figure out the individual case. but a git grep would fine it. I’ll see about poking at them a bit later. Sep 02 21:37:11 * armpit too many channels Sep 02 21:39:19 * armpit laughs ... just on swat this week... log bugs I am told.. no pressure on me Sep 02 21:40:17 * armpit tries to reproduce.. Sep 02 21:40:47 armpit: you know how I feel :/ Sep 02 21:41:05 yep.. Sep 02 21:41:29 zeddii: there is also a kernel stability problem for qemumips. The kernel is locking when we try and to the "make scripts prepare" test on building kernel modules Sep 02 21:41:40 zeddii: Reproduced with 5.0 and 5.2 :( Sep 02 21:42:02 5.0 ??? wow Sep 02 21:42:06 zeddii: the symptom is a hang in the kernel testimage and oddly seems less likely to do it on the commandline Sep 02 21:42:20 zeddii: not sure when this crept in Sep 02 21:42:36 zeddii: annoyingly its not 100% reproducible Sep 02 21:43:35 do we have something that is more reproducible. ? Sep 02 21:43:47 armpit: no. Been trying all day on that Sep 02 21:43:52 nothing to show for it Sep 02 21:44:33 https://autobuilder.yoctoproject.org/typhoon/#/builders/60/builds/981 is the latest example. 5.2 kernel there, died after 7 hours building Sep 02 21:44:43 hmm, I should tty sometthing at homw Sep 02 21:44:48 the qemu emulation just seems to lockup Sep 02 21:45:11 I also had it do it locally, secondary ssh into the image went dead too Sep 02 21:45:43 * armpit scratches head... Sep 02 21:46:22 armpit: we probably have to trace which recent commits brought it in Sep 02 21:48:38 RP, I had an issue on some boards in 5.2 on mmc on a BSP I maintain and it went away in later updates.. let me try 5.2 on qemu.. something seems odd Sep 02 22:05:41 RP: but not reproducible for me. I ran those commands about 20 tmes during my 5.2 testing. Sep 02 22:06:00 oh. you said “not” reprodicible. Sep 02 22:21:09 zeddii: seems to happen with the load from testimage but not the actual commands run manually Sep 02 22:21:59 I wonder if qemu 4.0 -> 4.1 did it Sep 02 22:24:29 zeddii: I'll retest with marka's devtool change to the kernel module devtool test and a revert of the qemu upgrade **** ENDING LOGGING AT Tue Sep 03 02:59:57 2019