**** BEGIN LOGGING AT Wed Jun 24 02:59:57 2020 Jun 24 08:15:04 Trying to build my recipe with origin/dunfell got strange error on the very early step (parsing recipes): https://pastebin.com/1ky1Lda0 my recipe is: https://pastebin.com/3hHnes9B Jun 24 08:25:32 dv|2: Indentation error? Just put a tab before your rm lines Jun 24 08:25:34 dv|2, use proper indentation (4 spaces or tab) for two lines in your do_configure_append(), Jun 24 08:26:05 thats what i thought too, but as the functions are shell it should be permitted IMHO Jun 24 08:26:29 Letothe2nd, only that it might not be shell :) Jun 24 08:27:12 mihai-: and it isn't, indeed Jun 24 08:27:20 http://cgit.openembedded.org/openembedded-core/tree/meta/classes/npm.bbclass?h=master#n62 Jun 24 08:27:40 nice Jun 24 08:27:52 dang Jun 24 08:28:38 dv|2: either use the python way to delete your files or add a task to be run right after do_configure ([postfuncs] I think it's called) which is using shell. Try and let us know :) Jun 24 08:39:02 xtron, with 4 spaces the result is the same. Jun 24 08:40:00 qschulz, it is shell... Jun 24 08:41:38 dv|2, the existing do_configure inherited from npm bbclass is python, you're appending bash commands to a python function Jun 24 08:42:28 ..yocto and its secret tricks.. :) Jun 24 08:42:55 qschulz, oh! how can I remove the file in python correctly then? Before /dunfell it work OK... Jun 24 08:45:46 dv|2, hm, indeed that used to be bash before dunfell Jun 24 08:45:48 nice catch :) Jun 24 08:46:46 dv|2: https://docs.python.org/3/library/shutil.html#shutil.rmtree Jun 24 08:47:02 dv|2, I guess you can use os.remove Jun 24 08:47:36 or what qschulz said Jun 24 08:52:16 a postfunc would be best Jun 24 08:55:50 may I force the bash somehow there? Jun 24 09:00:33 dv|2, why don't you use the python way ? It looks pretty straightfoward.. no /? Jun 24 09:00:34 "force the bash" sounds pretty brutal. Jun 24 09:09:17 now you have to reasearch if it is OK to write in bash in this function or not... Better to have a way to say "here is my bash script" Jun 24 09:15:17 postfunc should take care of that Jun 24 09:23:31 ok. but looks like needto find a way to specify the scripting language in recipes Jun 24 09:24:49 just write a two line shell postfunc Jun 24 09:25:00 you can't force do_configure to be another language, because its already been written Jun 24 09:30:18 dv|2: by default, tasks/funcs are shell. If it's prefixed by `python`, it's... python :) So you should be good if you're using do_configure[postfunc] += "myfunc" or something and have myfunc() { rm whatever } Jun 24 09:31:16 also.. is this rm that big of a deal? Does it bring anything to the table? because if you're losing a few hours just for cosmetic stuff, it's not worth it :) Jun 24 09:31:29 (though you've learned a few new things already) Jun 24 09:32:39 wholesome words.. Jun 24 09:46:20 JPEW, hmm, wasn't there a plan at some point to get the hash-equiv data output put under version control in buildhistory ? Jun 24 11:12:08 Anyone here recruiting embedded engineers ? Jun 24 11:28:18 Yocda: nobody officially/publicly :) Jun 24 11:49:13 hi, BitBake does not seem to like git addresses starting with "git@". is there anything I can do about it? Jun 24 11:54:25 polaris-: use the fetcher parameters for the username Jun 24 11:56:55 polaris-: git://git@... works here. Jun 24 12:02:27 sstiller: I'm trying that. I get "fatal: Unable to look up git@MYURL (port 9418) (Name or service not known)". git clone git@MYURL directly works alright though. Jun 24 12:04:43 polaris-: perhaps you need to append ;protocol=ssh Jun 24 12:08:34 sstiller: ah that's it :) thanks! Jun 24 13:02:14 kroon: It wasn't on my radar Jun 24 13:18:44 Hello everyone, I'm wondering how the mega manual was setup? Reason I'm asking: I could not find an alternative to the mega manual with sane split of chapters. I find the mega manual super annoying to work with (single straight go-to navigation, no splitting into subsection which leads to really tiresome and endless scrolling, etc.). So if the Jun 24 13:18:44 manual source is somewhere available I'd be happy to attempt a more easy to work through layout... Jun 24 13:19:26 emrius: the mega-manual is just a concatenation of all single manuals in the poky/doc directory, AFAIK Jun 24 13:19:39 I see Jun 24 13:19:50 Thanks for the hint. I'll start from there Jun 24 13:20:49 :) Jun 24 13:26:25 hm, under what conditions would do_patch not create a quilt series with what it applied ? Jun 24 13:30:44 emrius: FWIW, I find this actually awesome. I **know** the info will be there, I just need a good word for Ctrl+F and/or scroll in some parts that are interesting to me. Jun 24 13:31:03 emrius: https://www.yoctoproject.org/docs/ Jun 24 13:31:13 everything you need is there Jun 24 13:36:55 qschulz still structure helps. A lot! Also, navigation side bars that highlight where you are is a huge pro for orientation and where to find things (again). replace Ctrl+F with a search field and you have the same, plus search results ordered by chapter. Jun 24 13:37:42 I mean there is a reason why single page documentations are hard to find on the web. They are a pain in the ass. Jun 24 13:39:41 emrius: feel free to suggest improvements. historically the documentation focused on a printout-style, hence the structure as far as i understand it. that certainly differs quite a bit from todays web expectations. Jun 24 13:41:28 I totally get that. Sorry for the swearing! That was the pain in my worn out scrolling finger that made me say that. Jun 24 13:41:54 all is fine. Jun 24 13:42:07 I'll give it a shot if I find the time Jun 24 13:42:28 we usually resort to the "tastefully chosen" ctrl-f invocation, but that doesn't mean we can't improve. Jun 24 13:45:31 emrius: that is an opinion you should share publicly on the docs mailing list honestly. The timing is excellent since there'll be quite some changes in the docs (well, the behind the scenes mostly) Jun 24 13:46:56 qschulz good to know! thanks for the hint! Jun 24 14:13:19 emrius: https://people.linaro.org/~nicolas.dechesne/yp-docs/html/ that's what in WIP so it should match more or less your request AFAICT Jun 24 14:14:15 qschulz That matches exactly my request :) Nice! Jun 24 14:15:57 emrius: it's a PoC though, so probably broken in many places. Maybe you can contribute to this effort :) Jun 24 14:16:55 qschulz I definitely will! Even if it's just using it and giving feedback. This is very promising! Jun 24 14:20:21 emrius: https://wiki.yoctoproject.org/wiki/Documentation_Sphinx Jun 24 14:22:23 Hi, I reboot continously a lot of time for testing but systemd services sometimes occur failed and boot hanging. Now write "A start job is running for Permit User Sessions" (18min 40s / no limit). But not always hanging. How can I fix this stiuation? Jun 24 14:22:25 (y) Thanks Jun 24 14:25:03 beratiks: use sysv Jun 24 14:26:01 qschulz but Can I use sysv with qt5? I think there is dependency? Jun 24 14:26:37 beratiks: we use sysv with qt5.6 don't know about newer releases but that;d be surprising Jun 24 14:28:40 qschulz thanks for support. I am starting to convert my yocto project systemd to sysv. I tried before but I cant. I might handle , if can't I will ask you :) Jun 24 14:29:52 * Letothe2nd facepalms and slaps qschulz with a large trout Jun 24 14:34:57 beratiks: Mmmm... For large and complex systems, I'd advocate against. Systemd brings many improvments Jun 24 14:35:25 :) Jun 24 14:35:49 honestly if size and boot time aren't extremely critical, go with systemd. It's well maintained, widely used and you'll get help easily Jun 24 14:35:50 yes but I think systemd is not stable :) Jun 24 14:36:04 beratiks: tell that to all desktop distros :) Jun 24 14:36:18 * paulbarker "stable" is always relative Jun 24 14:40:44 paulg: if i can sit on it with out it breaking, it has to be considered "stable" Jun 24 14:43:02 Letothe2nd: struggling with autocompletion I see Jun 24 14:43:25 s/struggling with/battling/g Jun 24 14:44:38 hey, leave me out of any system disease discussions! Jun 24 14:45:16 hehe Jun 24 14:45:18 what, you aren't ghonnorea'ing to help ? Jun 24 14:45:56 Who ordered the puns? Jun 24 14:46:06 s/puns/puss/ Jun 24 14:46:08 do they have to be ordered here? Jun 24 14:46:15 ever seen "Error contacting Hash Equivalence Server" starting to pop up during a build? Jun 24 14:46:29 Letothe2nd: At least they're not fish puns Jun 24 14:46:44 fish buns? Jun 24 14:47:08 Letothe2nd: This isn't really the plaice for fish puns Jun 24 14:47:41 paulbarker: why not? i like fish. Jun 24 14:49:04 and any IRC place is good for that very specific fish pun. (for the youngters among us: https://en.wiktionary.org/wiki/trout_slap) Jun 24 14:49:46 yann: Yes, I think it has trouble procssing requests when running locally with high parallelism Jun 24 14:50:24 yann: It wasn't a problem on my 8 core machine... it shows up a lot more on the 36 core machine Jun 24 14:50:47 I make marine electronics; my whole career has been a fish pun Jun 24 14:51:00 JPEW: \o/ where can i sign up? Jun 24 14:51:30 JPEW: Ah that's interesting, I got into embedded by working on underwater noise monitoring equipment Jun 24 14:51:42 paulbarker: yay for more fish puns! Jun 24 14:51:50 Letothe2nd: for the career or the puns ;) Jun 24 14:51:57 JPEW: both? :P Jun 24 14:52:17 paulbarker: Ya, this is all consumer grade stuff... expensive yacht toys. Very interesting market Jun 24 14:52:39 JPEW: yeah there are 24 cores here - but then, don't we launch a hashequiv server once for the whloe build ? that looks like it terminated ot sbd removed the socket ? Jun 24 14:53:08 what's the impact for the remaining tasks of the build ? Jun 24 14:54:06 yann: It's not any worse than not having the server, it just means you might not find some equivalence you would have otherwise Jun 24 14:54:28 Letothe2nd: http://www.garmin.com/careers Jun 24 14:55:12 JPEW: no fish puns mentioned there... no deal :( Jun 24 14:55:34 Letothe2nd: Well, for good reason they don't let me work on the website Jun 24 14:59:10 yann: The server is single threaded, so I think there is some connection limit slowing it down.... I wonder if listen() needs a longer backlog.... Jun 24 15:01:31 hm, would it crash when its connection queue is full ? Jun 24 15:01:41 does it leave logs anywhere ? Jun 24 15:05:17 yann: No, it wouldn't crash when the log is full Jun 24 15:05:30 yann: Sorry, when the *connection* queue is full Jun 24 15:06:17 can we get it to run in debug mode, or anything that would help understand ? Jun 24 15:06:51 yann: Probably the easiest is to run one manually on the side Jun 24 15:07:30 yann: `bitbake-hashserve --log DEBUG` Jun 24 15:29:45 hmm, what's the status of hashequiv nowadays? i stopped using it for a while since it was in flux at the time Jun 24 15:32:06 kergoth: it's activated by default in dunfell, it seems Jun 24 15:50:10 kergoth: We seem to have most of the big kinks worked out Jun 24 15:53:48 Hi ! I'm getting back to my faulty recipe.. I got this recipe: https://paste.ubuntu.com/p/Cv5DNfpv6n/ and experiencing an issue with the path of delivered .service file.. if I use sbindir, package is delivered but with systemd_system_unitdir, nothing happens.. any clue ?? Jun 24 16:01:36 PaowZ: "delivered"? what do you expect and what do you have? Jun 24 16:02:12 dreyna: joining in a minute.. Jun 24 16:02:15 I expected a rpm package to be shipped.. but I only get -dev and -dbg packages Jun 24 16:02:22 HAPPY HOUR is on now ! Jun 24 16:02:28 ndec ok Jun 24 16:02:33 https://zoom.us/j/99151508871 Jun 24 16:04:53 qschulz: .. I expected rpm to be generated, and now, I'm getting this error (Empty tag: Group).. https://pastebin.com/DmZLUnLY Jun 24 16:05:36 ok.. for the error.. I had SECTION blanked.. I fixed it.. Jun 24 18:03:08 yann: Ah, the default buffer size is too small to transfer some of the larger packets. Probably needs a chunking protocol of some kind on the server side Jun 24 20:01:26 yann: Ok, I implemented a chunking protocol for hash equivalence and it seems fix the issue. I'm going to try it out and send the patches Jun 24 20:03:18 qschulz, kroon, rburton, PaowZ_, thank you. Jun 24 20:04:12 I'm curious, that is the difference between dunfell and dunfell-next ? Is dunfell-next ready to be used for production? Jun 24 20:06:33 dv|2: dunfell-next are the patches being tested. That branch head will get rebased (e.g. move in a non linear fashion), where as dunfell will only move forward Jun 24 20:21:21 I do have a variable ABC which gets assigned a random value (similar to DATETIME). ABC gets assigned to XYZ. I added now XYZ to vardepsexclude of a task which uses that variable to fix basehash changed issues. Question: Which one is the right one to exclude? ABC or XYZ? Why/what are the tradeoffs? Jun 24 20:21:59 I see lots of tasks excluding DATETIME, altough they only use it through indirection. Personally it seems that the actual variable in use in this task seems to be more sensible (XYZ)... Jun 24 20:30:41 JPEW: cool, will shake that :) Jun 24 20:38:48 JPEW: what list should I check for the patches ? Jun 24 20:39:39 yann: bitbake-devel, but I haven't sent them yet Jun 24 20:40:24 ok my first guess was correct after all :) Jun 24 20:43:35 hm brainfart Jun 24 20:43:51 if a recipe does export FOO='bar', bitbake will track that and rebuild the recipe if FOO changes to 'baz', right? Jun 24 20:44:06 Or does it ignore environment unless told otherwise Jun 24 20:44:08 can never remember Jun 24 20:45:18 Another issue I get with my dunfell builds: I have spurious rebuilds. A build just completed, but the gitlab.com hard timeout of 3h on runner killed it while uploading the images (sigh). So I'm just relaunching the build ... and it's gone out rebuilding things, starting with opkk-utils and libtool-cross - wow I fear that build will hit the same timeout... Jun 24 20:45:27 rburton: yes, every exported env var afffects all task checksums Jun 24 20:45:41 which is why i was working on a patch series to move to minimizing global exports and using task level exports Jun 24 20:45:45 never did finish that.. Jun 24 20:46:26 right, thought so Jun 24 20:46:31 how to exempt all tasks from that? Jun 24 20:46:49 specific use case: license key needed to run a native binary is in envrionment variable Jun 24 20:47:08 export FOO=bar in the local.conf works but i fear a license key change will cause a world rebuild Jun 24 20:47:10 probably need to add it to the hashbase whitelist, or unexport it and explicitly export it where needed Jun 24 20:47:19 we do the former for our license at mentor Jun 24 20:47:33 needed to add it both to BB_ENV_EXTRAWHITE and the hashbase whitelist Jun 24 20:48:37 oh, there it is .. https://github.com/kergoth/meta-kergoth-wip/blob/master/meta-minimal-exports/conf/task-exports.conf .. hmm, i should really finish that.. it's tough to know all the potential fallout of a change like that though Jun 24 20:49:00 needed per-recipe additions Jun 24 20:49:45 suppose i should split that into two pieces, first use task level across the usual suspects for all currently exported variables, then minimize that set later Jun 24 20:49:49 * kergoth yawns Jun 24 20:50:28 BB_HASHCONFIG_WHITELIST? Jun 24 20:50:37 ah no BB_HASHBASE_WHITELIST Jun 24 20:51:37 thanks kergoth Jun 24 20:52:12 rburton: You could also do something like FOO[export] = "1" in the specific recipe that needs it and not export in local.conf Jun 24 20:52:28 ueaj tjat Jun 24 20:52:33 erk Jun 24 20:52:56 yeah, that's an idea unless its needed for something particularly invasive like a toolchain component or something Jun 24 20:53:04 really dislike global exports in general Jun 24 20:53:14 i mean who'd have a toolchain with a license key Jun 24 20:53:18 * rburton slowly walks away Jun 24 20:53:20 there's no way to track what's used and what isn't, which is irritating Jun 24 20:53:23 haha **** ENDING LOGGING AT Thu Jun 25 02:59:57 2020