**** BEGIN LOGGING AT Sat May 25 02:59:58 2019 **** BEGIN LOGGING AT Thu May 30 09:29:05 2019 May 30 10:18:03 New news from stackoverflow: Confused about install -m and install -d using Yocto May 30 10:20:20 Hi channel. I've seen that support for Eclipse Yocto plug-in has been removed https://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#migration-2.7-eclipse-support-dropped . I've tried to search for the reasons that brought to this decision and to see if there are suggested alternative options, but couldn't find any specific reference. Can anybody elaborate on the topic? **** BEGIN LOGGING AT Thu May 30 11:39:20 2019 May 30 14:35:34 There are a couple of errors building thud branch for powerpc. Do I have submit a bug report? May 30 14:36:20 One could fixes with a single cherry-pick. May 30 14:53:19 is there some way to stop systemd from crapping into the serial console in warrior ? May 30 14:53:23 I don't need to see the kernel log twice May 30 14:53:47 worse, it garbles it when there's some meaningful kernel output by babbling into it May 30 14:54:03 I cant seem to find a way to just stop this braindead behavior May 30 14:54:27 it didnt happen in thud or earlier May 30 15:05:48 armpit: Do you want me to assign https://bugzilla.yoctoproject.org/show_bug.cgi?id=13359 to you for backporting? May 30 15:05:49 Bug 13359: normal, Medium+, 2.8 M1, JPEWhacker, RESOLVED FIXED, Loading the parsing cache when multiconfig is enabled is slow May 30 15:06:10 JPEW_, sure May 30 15:13:44 armpit: Done. Thanks May 30 15:18:20 in this bash script: https://paste.fedoraproject.org/paste/6-ybTmQRE6CLPg7JafGqeg May 30 15:19:21 the device is left in an inconsistent state. "ls /dev/sdd?" shows only two partitions when there should be three. May 30 15:19:52 however if you run fdisk /dev/sdd, and type "p" to print the partitions, then it shows 3 partitions May 30 15:20:30 and when you exist fdisk and do another "ls /dev/sddd?" you then see three partitions. May 30 15:20:38 any clues why? May 30 15:24:02 which partprobe May 30 15:38:21 /sbin/partprobe May 30 15:51:41 If anyone is new(ish) to yocto and has time to help, please take a bug from: https://wiki.yoctoproject.org/wiki/Bug_Triage#Newcomer_Bugs May 30 15:58:08 diego_r: re the Eclipse plugin, my understanding is that Intel had been doing the maintenance of it, and have stopped doing so May 30 16:06:40 smurray: :-( I haven't seen any specific message in the eclipse-yocto mailing list about this but I trust you. Any confirmation from the Intel guys in here? Any suggestion about alternative IDE setups for non-Qt projects? May 30 16:09:57 diego_r: there's not many Intel folks left working on YP, sadly. Re IDEs, no good ideas here, I switch between vi and emacs and build from the command line May 30 16:13:47 smurray: did I miss some important news about Intel relationship with Yocto? May 30 16:19:32 diego_r: nothing has been explicitly stated, but there were layoffs in their open source center, and a bunch of people have left. The only public declaration I can recall was that their manual testing support was ending with thud, think that was mentioned in status reports on the mailing list a couple of times a while back. Hence the significant effort to automate as much testing as possible... May 30 16:24:02 also diego_r if you know people with time an declipse knowledge, you could look at figurin gout what needs doing with th eeclipse stuff May 30 16:30:31 smurray: thanks. Very insightful May 30 16:32:04 diego_r: np May 30 16:32:20 Crofton: time and resources are often a problem when it comes to maintaining such a piece of software. May 30 16:37:15 diego_r: Vscode is very popular here recently. May 30 16:46:47 diego_r, exactly the problem the YP has May 30 16:47:14 and most embedded people do not understand why anyone would care about Eclipse, so really hard to get help May 30 16:53:21 sigh, this mixed output on serial is really annoying May 30 16:56:32 Marex: Had you already take a look of the kernel command line? May 30 16:57:00 alessioigor: for what ? **** ENDING LOGGING AT Thu May 30 17:08:18 2019 **** BEGIN LOGGING AT Thu May 30 19:17:05 2019 May 30 19:26:40 yeah, I always figured that would be populat, even if it wasn't for me :) May 30 19:28:17 course when teams and companies are small, the two roles are often filled by the system developer anyway, so a certain amount of bias in that direction makes sense May 30 19:34:48 Hmm, the bash-ptest Makefile includes ${UNINATIVE_LOADER} (from BUILD_LDFLAGS).... whats the best way to make that reproducible? May 30 19:53:26 JPEW_: patch it out? :/ May 30 19:53:42 JPEW_: we have sed expressions for a lot of stuff in ptest :( May 30 19:53:58 Ya, I was think I could maybe sed replace it with /bin/false? May 30 19:54:20 JPEW_: depending on context, yes May 30 19:54:34 I don't *think* any of the "native" build stuff done by bash is needed for ptest May 30 19:54:43 shouldn't be May 30 19:55:45 I'll try it. May 30 19:57:16 I think that will get me down to 18 perl packages and 2 python packages... is it a requirement to get them fixed before we add the QA test? May 30 19:57:46 If we do log bugs for them it seems like an easy way to reproduce them (e.g. the QA test) would be nice for whomever tries to fix it May 30 19:59:12 JPEW_: I can't have red targets on the autobuilder. We may be able to add it disabled May 30 19:59:32 (red targets quickly develop into other bugs) May 30 20:01:12 Right May 30 20:34:18 JPEW_: do we really need coreutils-native? :( May 30 20:34:27 that is a horribly heavy dependency May 30 20:35:15 Oh... um... Probably not? We just need a way to convert SOURCE_DATE_EPOCH to a string. Or alternativly, #define PERL_BUILD_DATE "" would work too I suppose May 30 20:35:27 JPEW_: inline python? May 30 20:35:48 Isn't ${SOURCE_DATE_EPOCH} a string? May 30 20:36:04 Yes, but it is an integer May 30 20:36:52 * RP doesn't follow as an integer is a string May 30 20:37:07 We could pass it as is, but then perl would say: "Build on 123456" May 30 20:37:38 ah, right, human readable May 30 20:37:44 correct May 30 20:37:44 I still suggest inline python May 30 20:37:58 Ok, I'll do that May 30 20:38:42 So it would need to DEPEND on python-native then? May 30 20:38:57 JPEW_: bitbake is written in? May 30 20:40:17 Ah, Ok. So.... there are actually 2 SOURCE_DATE_EPOCHs :) May 30 20:40:29 There is the environment variable that reproducible.bbclass shoves into the environment before a task runs which is calcualted based on the source code May 30 20:40:55 well, python3 is available in the shell if you want to use the environment May 30 20:40:58 And there is the bitbake variable ${SOURCE_DATE_EPOCH}.... they are not the same and AFAICT, aren't related May 30 20:41:07 how confusing :( May 30 20:41:11 indeed May 30 20:41:46 I think the bitbake variable is used for making images and stuff, but if a given recipes source doesn't have any reliable SDE, it just gets 0 May 30 20:42:53 It sort of makes sense; it means that the SDE for a recipe will probably change to something intelligent when the recipe is updated (as opposed to be fixed to the bitbake variable) May 30 20:43:05 that sounds vaguely familiar May 30 20:43:26 Sounds like something we could better document May 30 20:43:31 Indeed May 30 20:45:51 jonmason: around? May 30 20:55:42 * RP is shocked when arm ptest with kvm works May 30 20:55:50 https://autobuilder.yocto.io/pub/non-release/20190530-15/testresults/qemuarm64-ptest-fast/ :) May 30 20:56:37 cool, does that close the bug then? May 30 20:58:15 armpit: well, I need to sort the patch out and submit it and integrate it to the full/quick targets but the fact its actually run is a nice start May 30 20:58:58 next someone needs to sort through the results as something isn't right there compared to x86 in some cases May 30 20:59:10 nice! May 30 20:59:24 interestingly the gcc9 bluez5 issue reproduces on arm too :) May 30 20:59:33 are the results Endian related ?; ) May 30 21:00:24 someone will need to fix the ptest reporting to work per arch... May 30 21:00:34 this is going to confuse the reporting :/ May 30 21:01:05 just add it to the long list May 30 21:02:00 armpit: well, the reports are about to get trashed May 30 21:03:11 awesome May 30 21:03:13 RP: Sorry, busy day. What can I do for you? May 30 21:03:31 * armpit jonmason working in thumb mode ? May 30 21:03:46 jonmason: will take this to privmsg May 30 21:11:09 RP: I think I have another large chunk of perl ptest dependencies sorted May 30 21:11:39 RP: libxml-parser-perl ptests are passing in core-image-full-cmdline May 30 21:12:51 RP: well, almost... May 30 21:15:00 moto-timo: that sounds promising. core-image-minimal is my gold standard! :) May 30 21:20:36 RP: I'll do that next, now that I don't need to debug the sun and the moon May 30 21:21:01 moto-timo: right, I'll settle for any improvements! :) May 30 21:21:16 moto-timo: if you have a test environment, quilt-ptest was the other key one with perl module issues May 30 21:21:17 RP: me too May 30 21:21:28 RP: good to know! May 31 02:51:02 New news from stackoverflow: What is the difference between Linux Modules, Linux Drivers, and packages **** ENDING LOGGING AT Fri May 31 02:59:58 2019