**** BEGIN LOGGING AT Fri Dec 07 03:00:01 2018 Dec 07 08:40:00 hi al, already asked, but maybe today is a better moment: i have 2 apps that needs to use (to link with) 2 different versions of a same library. How would be convenient to manage this in ycoto ? Dec 07 09:46:32 hi Dec 07 09:47:45 has anybody experienced freezes of bitbake invocation? I.e. `bitbake core-image` takes forever and won't proceed? Dec 07 10:06:33 T_UNIX, and eventually times out? Dec 07 10:08:25 if you're using the same build directory for multiple machine configuration, check the size of the files under ./cache/ Dec 07 10:51:39 * RP is surprised but pleased to see sumo-next worked Dec 07 10:57:01 Didn't quite log everything but I've found those missing patches Dec 07 11:39:10 mihai: single directory per build. Shared cache and download though. No timeout (not even eventually) :-/ Dec 07 11:40:08 or maybe it's just takes exceptionally long (compared to all the other parallel builds). Dec 07 12:21:23 Hi, folks! Dec 07 12:21:39 I have an another newbie question :) Dec 07 12:23:09 Can I set PREFERRED_VERSION inside a recipe file? Or I have to use it only in configs? Dec 07 12:58:33 acrap: setting it inside the recipe wouldn't achieve anything as the selection of versions happens outside of recipes Dec 07 12:58:48 got it Dec 07 12:58:58 you're welcome to set it but nothing will change ;) Dec 07 13:00:32 :) Thanks! Dec 07 13:39:21 New news from stackoverflow: Adding net-snmp package to poky yocto fails Dec 07 14:09:26 New news from stackoverflow: am335x evmsk screen init:id"00" respawning too fast disabled for 5 minutes yocto sumo [closed] Dec 07 14:35:23 is there quick way i can see if a specific PN is in my project without actually runing bitbake PN? Dec 07 14:37:29 s/runing/running/ Dec 07 14:43:34 define project Dec 07 14:47:18 yates: bitbake -s | grep PN ? Dec 07 14:56:01 rburton: an endless pit into which money, time, and resources are thrown... Dec 07 14:56:27 RP: close enough - thanks! Dec 07 15:12:42 * kergoth yawns Dec 07 15:58:07 How unhelpful: https://autobuilder.yoctoproject.org/typhoon/#/builders/37/builds/68/steps/7/logs/step1c :( Dec 07 15:58:31 I'm half tempted to ban assertTrue and friends in our QA code Dec 07 16:02:41 everyone who introduces a bad assert immediately gets the bug when it fails Dec 07 16:17:18 what I igured out is that assertTrue is only useful if included with a message explinign WHAT you checked Dec 07 16:17:41 in the layerindex checks, I added that stuff to the bitbake tests.. cause after a few weeks I forgot what i was checking Dec 07 16:18:43 agreed Dec 07 16:18:47 that's one thing i like about pytest, it automatically includes context. i.e. the basic assert example at https://docs.pytest.org/en/latest/ Dec 07 16:19:40 So assertTrue is fine.. but some kind of a check that a message is included is vital Dec 07 16:34:39 fray: right, that is the real problem. The code is full of it though :( Dec 07 16:35:51 (for those of you following along, but don't know what we're talking aobut.. it's the difference between) Dec 07 16:35:58 elf.assertFalse(self.layerindex.is_empty()) Dec 07 16:35:58 and Dec 07 16:36:04 self.assertFalse(self.layerindex.is_empty(), msg="Layerindex is empty") Dec 07 16:36:20 in the former case you just get an 'assert failed' message, in the later, you get a message that says WHY it failed Dec 07 16:36:27 rburton: something in -next is causing this, its reproducing :/ Dec 07 16:37:18 fray: Its actually worse as people do assertEqual(result.exitcode, 0) and then you never see result.output Dec 07 16:37:33 and its output that would tell you why its failing Dec 07 16:37:37 RP: next not master? Dec 07 16:38:26 i hope it doesn't nee libart-lgpl! Dec 07 16:40:43 need to see the build output Dec 07 16:40:51 have i mentioned how much i hate that bit of oeqa? Dec 07 16:40:57 so much that i literally ignored it when writing a new sdk compile test Dec 07 16:42:02 RP yes, I expected that was what you were talking about.. but it's really the same problem.. an assert without ANY info Dec 07 16:42:25 RP: remember we have from oeqa.utils.subprocesstweak import errors_have_output Dec 07 16:42:56 we should just monkeypatch that into selftest startup Dec 07 16:44:51 RP: building a sdk here, i've a hunch i broke this Dec 07 16:48:41 hi Dec 07 16:50:35 aehs29: around? Dec 07 16:51:36 I've built an esdk for my project, but I'm getting a Taskhash mismatch anytime I try and use devtool Dec 07 16:52:04 aehs29: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=df17b1240fbcdae199d944673728e9df965d1002 - If I rerun create_manifest it simply removes that :/ Dec 07 16:53:23 It's not causing a problem except that every devtool operation parse recipes, which really slows down the build cycle. Dec 07 16:53:40 rburton: why on earth do we need that class, its horrific :( Dec 07 16:54:46 RP: the assimp test i wrote just ignores it, rewriting the galculator test to be the same now. side effect is it just throws an exception with the output in on failure Dec 07 16:55:15 rburton: Surely check_output or similar already does what that does? Dec 07 16:55:44 RP: no, it doesn't display the output. you need that subprocesstweak to get that output too Dec 07 16:56:19 rburton: I'd swear it does if you have things setup right Dec 07 16:57:20 all the functions just raise the exception Dec 07 16:57:23 and that is def __str__(self): Dec 07 16:57:23 return "Command '%s' returned non-zero exit status %d" % (self.cmd, self.returncode) Dec 07 16:58:13 kergoth: remember long ago we were talking about automatic transformation of .in files to replace strings like {prefix} Dec 07 16:58:47 just discovered string.Template which is a cut-down formatting tool that uses $ to markup Dec 07 16:59:05 but lets you escape easily Dec 07 16:59:09 so that's almost a better thing to use Dec 07 16:59:30 especicially as you can change $ to eg @ Dec 07 17:02:21 rburton: ah, see bb.utils._print_exception() Dec 07 17:02:25 rburton: we hacked it Dec 07 17:03:07 yay bitbake Dec 07 17:04:48 rburton: I was sure we displayed it, I'd forgotten that was us :) Dec 07 17:09:33 having a problem with my eSDK. It parses recipes ever devtool operation. I assume this is because of the taskhash mismatch error it keeps throwing for do_patch. How do I find the source of this? Using pyro (bitbake 1.34.0) Dec 07 17:14:07 https://pastebin.com/DN2LX7kU Dec 07 17:14:21 RP: ok why does data_smart's __getitem__ pass false to getVar Dec 07 17:14:36 that's really annoying for what i was planning to hack up Dec 07 17:23:48 not getting any mismatches when bitbaking or building the SDK only after it's deployed. Have done a full clean build not too long ago... Dec 07 17:45:07 rburton: as the expand parameter? We always used to default not to expand Dec 07 17:49:30 JPEW: I merged the persistdb patches as everything seemed fine, then we saw: https://autobuilder.yoctoproject.org/typhoon/#/builders/84/builds/61 :/ Dec 07 17:50:27 RP: sure, but we changed getVar. can we change the dict accessors too Dec 07 17:51:06 rburton: do we actually use them much? Dec 07 17:51:29 RP: arguably its a nicer way to get the data, and i want to hack up a class that needs dict accessors to the data store Dec 07 17:51:29 rburton: I suspect the answer is we can change but I'd like to understand if/where they're getting used Dec 07 17:51:40 i can write an adaptor object obviously Dec 07 17:52:18 huh maybe i should have reverted that binutils change before building a sdk Dec 07 17:52:20 rburton: any luck finding the cause of the sdk failure? Dec 07 17:52:25 still building! Dec 07 17:52:33 98% done Dec 07 17:52:39 and i've rewritten the test case too Dec 07 17:52:44 rburton: I reran a failed build and it passed but there were other failures Dec 07 17:53:14 i wonder if it was the tmpdir move Dec 07 17:53:14 halstead: sorry the current build is running into the maint window. It can be stopped if that helps Dec 07 17:53:19 maybe its deleting stuff too early Dec 07 17:56:17 RP, It looks like it should finish soon enough. And I have the new hardware to keep me busy. Dec 07 17:58:40 RP: so something broke rm_work Dec 07 17:59:00 in that its not rm_working Dec 07 18:02:18 ok fine it is sometimes Dec 07 18:02:46 rburton: -c populate_sdk by any chance? Dec 07 18:05:12 anyone have any tips on tracking down the source of a taskhash mismatch? Dec 07 18:07:13 RP: yeah i just had that chain of thought. it wasn't doing any cleaning of stuff that had been built Dec 07 18:07:25 presumably because its all hooking on do_build Dec 07 18:07:55 rburton: right Dec 07 18:08:29 aehs29: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=edef36b0246c5c6196a104fcab80ac2e302f67d6 fixes my problem... Dec 07 18:08:40 apart from annoying trailing whitespace in the json file Dec 07 18:10:38 your whitespace obsession is an ongoing concern! Dec 07 18:12:19 someone said something about whitespace? :) Dec 07 18:12:26 heh Dec 07 18:12:29 * Crofton trolls Dec 07 18:12:51 good day for channel role reversals Dec 07 18:12:52 * JaMa hides behind wall of tabs (and few spaces on top or before that) Dec 07 18:13:20 I should propose a talk on white sapce for the LF leadership thing Dec 07 18:13:59 * JaMa wonders what Trump thinks about tabs Dec 07 18:14:06 OMG Dec 07 18:14:15 * Crofton needs a Trump tweet generator Dec 07 18:26:16 Crofton: "yocto changes whitespaces" is that good one ? Dec 07 18:41:28 RP: autodiscovery of tests is great and all but i think we need a way of saying 'core-image-sato sdk test should be passing these tests' Dec 07 18:41:44 to catch stuff like galculator suddenly skipping because the sdk changed Dec 07 19:00:06 RP: That's a rather random failure case... Dec 07 19:05:49 RP: Ah nevermind, I see the error now. Dec 07 19:06:15 rburton, RP: I like dictionary access too, but I don't think the signature generation accounts for dictionary access. you'd have to add everything to vardeps Dec 07 19:25:11 fixable surely Dec 07 19:29:41 * zeddii_home wonders if there really hasn’t been any oe-core email for 6 or 7 hours. or if my gmail is screwing with me. Dec 07 19:30:13 we can hardcode the assumption that 'd' is a datasmart instance, i suppose. or add alot of false positives, since not everything accessed like a dictionary is a datasmart, and we can't determine that from the ast Dec 07 19:30:19 well, not easily Dec 07 20:21:57 zeddii_home: still have sense of wonder in this social media world... I would suggest go ahead and tweet your frustation out and tag #gmail and #everything-sucks .. come on be one of us.. in the end its not your fault you know Dec 07 20:22:58 I am sick and I am trying to comeup with blaming something/someone for it Dec 07 20:24:00 thumb2 codegen sucks on llvm found another wrong-codegen today Dec 07 20:24:19 hehe :D Dec 07 20:24:53 I’m just going to enjoy my oe-core mailing list silence, and check again tomorrow to see if it is empty. I checked the mailing list admin site, and I am still subscribed! Dec 07 20:25:39 I got muted on arm-linux ml because gmail was sending them bounces on my behalf Dec 07 20:25:58 so I unsubscribed completely from arm-linux ml Dec 07 20:26:12 * zeddii_home approves! Dec 07 20:26:24 I just mark it all read periodically .. I should do the same. Dec 07 20:26:42 less stuff to delete at the end of day - I have zero mails inbox policy Dec 07 20:27:22 I have a very short inbox policy as well. it’s my TODO list. and I completely fail on those lists .. so you are right, just unsub and worry about them again if they become a focus. Dec 07 20:28:23 yes, I was talking to Alpine maintainer and it seems all distro folks have to subscribe to so many MLs he was laughing out when I said I subsribe to so many mls Dec 07 20:28:44 there should be more offline webby way to communicate Dec 07 20:29:11 I can always search for topics if I need them. no need for a gmail copy most of the time. Dec 07 20:29:51 gmane Dec 07 20:30:27 we need gmane-ng Dec 07 20:30:38 :D Dec 07 20:31:13 https://www.mail-archive.com/ is not as good, I could reply to threads from gmane without subscribing which is ideal for one of discussion Dec 07 20:35:10 Can someone donate CPUs to oe build servers plzzz.. ka6sox was mentioning its going to be < 500 USD Dec 07 20:36:54 Took 21 hr on abaco .. Took 13 hr on kwaj Dec 07 20:37:11 kwaj has new CPUs we need to upgrade other builder Dec 07 20:37:49 I am building on my own machine which is faster than these machines Dec 07 20:44:26 They are getting pretty old Dec 07 20:51:57 raspbian still uses thumb1 ISA wow Dec 07 21:08:15 I use a very complex set of gmail filters... it works most of the time Dec 07 21:12:11 khem, binary compat Dec 07 21:20:30 Crofton: yeah but it performs really well Dec 07 21:20:44 so I wonder if arm should have just stuck to thumb1 Dec 07 21:26:47 RP: left testsdk running for an hour in a loop, no failures. will throw my rewrite at the AB, if it does fail it has better logging. Dec 07 21:29:47 khem: Having to switch to ARM mode to do something as simple as disabled interrupts is *so* annoying. Dec 07 21:34:35 RP: ok let me check Dec 07 21:55:48 RP: I sent something that I think will fix that persist_data problem... although I sent it to the wrong mailing list. Doh. Dec 07 21:59:42 khem: abaco might be a VM on the same HW as bonaire VM where I'm running LuneOS builds, yes both (all all 3) are rather slow Dec 07 22:00:30 and who knows what else is running on those boxes.. Dec 07 22:39:01 JaMa: I dont think its a VM but who knows Dec 07 23:46:11 RP: patch coming up for the python issue Dec 07 23:46:31 JPEW: just saw the patch, thanks! Dec 07 23:46:41 aehs29: thanks. I'm curious how far out I was :) Dec 07 23:47:46 RP: pretty far haha Dec 07 23:48:01 RP: this is happening because we use directories on our manifest Dec 07 23:48:40 RP: and its hitting one of those corners where the strings matched where the shouldnt have Dec 07 23:48:55 aehs29: I wondered about that Dec 07 23:49:54 aehs29: I ran out of time to understand what was happening Dec 07 23:50:00 RP: plus sadly I havent had time to maintain the py2 version as much as the py3 Dec 07 23:50:11 RP: yeah the code on py2 is harder to understand Dec 08 00:05:50 Hmm, the mystery sdk bug turns out to be one of my patches :( Dec 08 00:12:06 RP: phew at least it wasn'tone of mine :) Dec 08 00:14:19 JPEW: your bitbake patch isn't against master? :/ Dec 08 00:14:42 RP: found an interestnig multiconfig bug yesterday Dec 08 00:14:59 I will file a bug later Dec 08 00:15:55 aehs29: I'm not entirely surprised there are some... Dec 08 00:16:50 you need to build two separate DISTROs, one that INHERITS buildhistory for example, or something that has a post parsing event, you need to declare DISTRO on your main local.conf to that one that has the extra INHERIT, and one of your other confs should build the DISTRO that doesnt INHERIT that Dec 08 00:17:36 aehs29: they get the wrong events? Dec 08 00:17:38 and the INHERIT NEEDS to be on the distro.conf, if it is in local.conf it works fine Dec 08 00:18:07 aehs29: I can imagine why that could break Dec 08 00:18:08 RP: so it errors out becuase it tries to read variables that dont exist, e.g. BUILDHISTORY Dec 08 00:18:36 RP: the interesting thing for me was that I was testing the INHERIT inside the local.conf and it didnt error out Dec 08 00:18:44 it needed to be inside the distro.conf Dec 08 00:19:26 aehs29: interesting. The event handlers aren't namespaced per multiconfig so can probably get confused :/ Dec 08 00:19:27 otherwise it was parsed as well by the other multiconfs and the BUILDHISTORY variable existed Dec 08 00:19:55 RP: yeah I will file a bug later today and take it myself Dec 08 00:20:34 RP: also I have a patch that modifies the debug info, to show for which mc its doing something for, because that was making things even more confusing Dec 08 00:20:51 when multiconfig isnt being used it'll just show the default I suppose Dec 08 00:21:37 another thing I meant to ask, was that when is the variable expansion happening?, can we keep the parsed files and reuse them for another multiconfig?, because at this point were parsing everything several times Dec 08 00:22:02 basically, can we keep the parsed files, and expand them depending on each multiconfig Dec 08 00:24:12 aehs29: we can keep the ast for the files and I think we may already do so. There are things like immediate expansion during parsing though so its not that much of a win Dec 08 00:24:24 aehs29: our flexible structure is a nightmare for optimisation/caching Dec 08 00:24:32 * RP -> Zzzz Dec 08 00:24:52 RP: alright Dec 08 00:24:56 RP: night Dec 08 01:44:49 RP: sent you the python fixes for both py2 and py3 **** BEGIN LOGGING AT Sat Dec 08 01:45:08 2018 **** ENDING LOGGING AT Sat Dec 08 03:00:00 2018