Skip to end of metadata
Go to start of metadata
paulsherwood== GENIVI Tools Team Meeting Starts ==13:00
paulsherwoodhi all13:00
* paulsherwood notices 46 people in the channel :)13:00
paulsherwoodok, to recap from last week, minutes were published at
paulsherwoodi've put a very draft agenda for today at
CTtpollardhi paulsherwood13:01
paulsherwoodunsure how many people will be active today, since some contribs are on holiday13:01
paulsherwoodso here goes...13:01
paulsherwoodany feedback or thoughts on last week's meeting itself, and/or the minutes?13:01
steve__did get round to commenting but I think I would have just echoed the existing feedback13:02
paulsherwoodok cool :)13:02
steve__did *not* get round to..13:02
* paulsherwood guessed that :)13:02
paulsherwoodok, let's try first agenda item13:03
paulsherwood== Wikis ==13:03
*** mdunford has joined #automotive13:03
paulsherwoodjoelreplogle: any update for us? is toplevel structure set?13:03
*** pespin has joined #automotive13:03
paulsherwood(or have others any feedback from followups with joel)13:03
joelreploglenot much progress... the proposal will be given by 31 aug for wiki structure for full community feedback13:04
pkonopelkodid not see anything since last pmo call13:04
jeremiahI spoke with him in the OSS meetingh13:04
jeremiahI think he's in this meeting13:04
paulsherwoodjoelreplogle: i may have misunderstood, i heard a whisper that you might 'just do it' and see if folks like the result?13:04
jeremiahWe're waiting on feedback from GENIVI Marketing and the Technical product manager that is due in Early Sept.13:05
jeremiahThat is my understanding, Joel can confirm or deny. :-)13:05
paulsherwoodwho is that person?13:05
joelreplogleit's unfortunately more complicated (in terms of players) than just putting something up.13:05
paulsherwoodjoelreplogle: why is it?13:05
paulsherwoodbetter to seek forgiveness than permission? :)13:06
joelreplogleProbably right ;-)13:06
pespinHi, I implemented a simple client app <->service app using common-api-dbus 3.1.3 on the system dbus bus. When I call the method from the client, I can see the service receiving the request and answering it, but it seems just before answering the capi3 layer gets disconnected from the bus (and dispatcher thread is finished), and of course the service no longer works after that.13:06
jeremiahpaulsherwood: The Technical Product Manager is Philippe Gicquel13:06
pespinanybody having similar issues too?13:06
paulsherwoodjeremiah: ok thanks13:06
steve__think you just received some feedback on meetings in irc ;)13:07
paulsherwoodsteve__: yup. we could move it in future to a meetings specific channel :)13:07
paulsherwoodanyways moving on... but folks are welcome to help pespin if they can13:08
paulsherwoodso joelreplogle i guess folks will remain a bit reluctant to use the new wiki til this is resolved13:08
gunnarxpespin, try emailing genivi-ipc list.  We're having a formal meeting here now for the next half hour or so.13:08
pespingunnarx: ok sorry I didn't notice ;)13:08
paulsherwoodany further comments on wiki13:08
gunnarxpespin, no worries13:09
* paulsherwood moves on to Generating Tests with Franca13:09
paulsherwoodKlaus Birken requests feedback/contributions to
*** jeremiah is now known as jeremiah_afk13:09
paulsherwood(klaus is on vacation so hopes to see feedback on his return)13:09
joelreploglere the wiki: I'd like to encourage people to make use of it as much as possible.  Pages may be moved a bit after the final structure is in place, but that's a simple task vs putting the content in.13:09
paulsherwoodjoelreplogle: ok, so you are saying that if we use the new one, for sure our work won't be lost or need to be re-done?13:10
joelreplogleyes, absolutely!13:10
paulsherwoodok, thanks.13:10
paulsherwoodas a show of faith i'll do13:10
paulsherwoodACTION paulsherwood migrate TT wiki content13:11
paulsherwoodothers are welcome to join me :)13:11
joelreplogleperfect - thanks!13:11
steve__happy to move. More an issue of timing for me. Need to concentrate on a deadline this week.13:11
paulsherwoodany comments here on generating tests with franca?13:11
paulsherwoodsteve__: understood13:11
paulsherwoodok, let's move onto CommonAPI please13:12
paulsherwoodthere was a lot of discussion on this last week but i think key interested folks are not here today13:12
paulsherwooddoes anyone have anything to discuss on CommonAPI?13:13
CTtpollardI've sent around an email re getting it to work on the GDP13:13
paulsherwoodgunnarx: please go ahead13:14
gunnarxWell I'm not sure if it's a tools thing but we need to reconfirm GENIVI's strategy on shipping generated code or not.13:14
gunnarxbackground is, not all can use or like the java based tools13:14
gunnarxso we have typically committed generated code to git13:14
gunnarxthis is not best practice, since the source for it is there in theory already13:15
gunnarxbut it simplifies building stuff13:15
paulsherwoodseems sensible, though13:15
gunnarxand what I have found in GDP, baseline and elsewhere is a bit messy13:15
gunnarxFor example some recipes include a huge patch to add src-gen using a different commonapi version...13:15
gunnarxneeds cleanup anyway.   I took a look in audiomanager and did not find that it ships src-gen.  Maybe someone can confirm/deny?13:16
KlausUhlOne big problem with this approach is that the generator with which the code is generator usually has to match the runtime library *exactly*.13:16
KlausUhlBut I understand the background.13:16
paulsherwoodso do you have a strong view that we should insist on the from (real) source approach, gunnarx ?13:16
paulsherwood(or just aim to tidy up better)13:17
pkonopelkodoes anybody know, if there is any problem with using command line tools for code generation?13:17
gunnarxI actually have no opinion now because I need more data...  I need to understand in which state each project is, if the component maintainers can commit to maintain src-gen and so on...13:17
pkonopelkoManfred claims that they work in 3.0 and there is no need for ditributing generated code13:17
gunnarxpkonopelko, OK.  We should check that then, but the baseline and GDP needs to catch up with the strategy in that case13:18
jeremiah_afkHas anyone tested the command-line tools for CommonAPI 3?13:18
KlausUhlThe major challenge is to integrate the generators into your build.13:18
*** jeremiah_afk is now known as jeremiah13:18
jeremiahWhat KlausUhl said.13:18
KlausUhlThere is no off-the-shelf integration with anything, be it CMake, autotools, etc.13:18
pkonopelkoKlausUhl: great, so it looks like a problem for tools team13:18
gunnarx+1 :-)13:19
paulsherwoodi think you may be right... provided the tools team has people who can actually do that kind of work?13:19
steve__This is picking up the conversation at the time of Genivi 8.0 about whether the EGs provide generated code or not to the baselines and was postponed as it was a big topic for the baseline maintainers to adopt?13:19
gunnarxsteve__, yes pretty much13:19
paulsherwoodi know there are people here on #automotive who could do it, but they may not consider themselves members of the tools team :)13:20
steve__ok, feedback then from both Baserock and Yocto maintainers was it is a significant piece of work.13:20
paulsherwoodhow many projects use/integrat with commonapi?13:20
jeremiahNot that many13:20
jeremiahIVI radio13:20
gunnarxWhat I need is to have clarity and I'd like to focus on simplicity in building.  So if tools are still difficult my "opinion" will probably revert to committing the generated code still.  Having patches hidden in meta-whatever must be the worst of all solutions.13:20
jeremiahI think there are plugins for AudioManager to use it.13:20
paulsherwood'Having patches hidden in meta-whatever must be the worst of all solutions.' +113:21
gunnarxand that's what we have today, not committed to git, and also not using the tools = patches in meta-layer which is totally nuts.13:21
gunnarxin certain cases I should say13:21
paulsherwoodok, so i'm ok with this being a TT requirement, assuming BIT can not take it13:21
paulsherwoodsteve__: ^^?13:22
gunnarxI'd be happy if TT take it on, figure out if we can make workable build instructions that can be applied to all.  I'd be happy to see it.13:22
gunnarxMost of all, a recommendation to SAT would be nice so we know which way to go and can apply that consistently13:23
steve__sorry afk13:23
steve__yes happy for TT to push on good solution13:23
paulsherwoodok any more comments on CommonAPI?13:23
steve__was just grepping the source but didn't get clear pic back on what was using13:24
paulsherwoodnext up, CIAT13:24
paulsherwooda topic dear to my heart :)13:24
gunnarxand that has gone to your head :)13:24
paulsherwoodindeed :)13:24
steve__On gfx drivers..13:25
paulsherwoodgunnarx: do i take i you think this is a bad topic?13:25
*** jeremiah is now known as jeremiah_afk13:25
gunnarxno I'm jesting13:25
steve__Happy with suggestion we discussed.13:25
* paulsherwood strongly believes we need a solution 13:25
paulsherwoodsteve__: which suggestion precisely?13:25
steve__If I was on a phone I would have told you by now ;)13:26
paulsherwoodbut there would be no record :)13:26
steve__Remove gfx drv from binary distro..13:26
steve__Have script to pull needed parts from click through download and "make good" again.13:26
paulsherwoodah, yes. i think that's a separate topic, though, for GDP?13:26
steve__Script would be maintained by Genivi.13:26
paulsherwoodyup, that makes sense13:27
steve__Not part of CIAT output?13:27
paulsherwoodi think it would make GDP possible in CIAT, yes13:27
paulsherwood(and also reduce new-user-friction for GDP in general)13:27
CTtpollardI'm not sure how you could script for the clickthrough download however13:28
paulsherwoodCTtpollard: you couldn't13:29
steve__End user downloads, then runs the script13:29
paulsherwoodgunnarx: one of the key discussions last week was about not forking recipes, iirc13:29
steve__on the gfx download13:29
CTtpollardwe have scripts for if the end user manually downloads the binaries13:29
gunnarxpaulsherwood: yes?  not sure i follow13:30
CTtpollardthey are not maintained by genivi though13:30
paulsherwoodbut since then i think i've discovered that adding config/tweaks is default behaviour anyway13:30
paulsherwoodgunnarx: this was the idea of a single git mirror13:30
paulsherwoodinstead of pulling direct from upstreams13:30
steve__Paul kindly signed the GDP maintainer up for it :)13:30
gunnarxpaulsherwood, is this related to steve__ 's topic, or a different one?13:30
gunnarxjust trying to follow13:30
paulsherwoodsorry, different one.13:31
paulsherwoodi think steve__'s topic is in hand :)13:31
paulsherwoodso i'm mainly interested in a) confirming if the picgture i've drawn makes sense to others)13:32
paulsherwoodb) identifying which parts we can implement off the shelf with minimum pain13:32
paulsherwoodc) identifying how we would approach things that don't fall into b)13:32
gunnarxI think we've already given the feedback that it appeared a bit too complicated, but that we would start in one end with the most important?13:32
paulsherwoodok. do folks have any clear recommendation on what's most important?13:33
gunnarxI would start with b), let c) follow naturally later on13:33
*** Egy has quit IRC13:33
paulsherwoodfair enough13:33
gunnarxBuild automation + source mirror is most important for me.  Then follows ability to upload code for "trial build", which is typically integrated with a code-review ability also (Gerrit style)13:34
*** jeremiah_afk is now known as jeremiah13:34
paulsherwoodso, at the rightmost end, i think the only realistic FOSS candidate for test soln is LAVA today13:34
paulsherwood(ie test farm/cloud... frameworks are further down)13:35
gunnarxthat fits earlier consensus afaict13:35
paulsherwoodi think gerrit + ml was the previous consensus on review13:35
steve__perhaps for real h/w, but presumably more flexibility if QEMU13:36
paulsherwoodi think LAVA can work with QEMU too13:36
steve__Any issue for LAVA on x86?13:36
paulsherwoodsteve__: it's python, i'd be astonished if there were13:36
steve__Was thinking of out of box solution13:37
paulsherwoodand the default server-side soln is x86 anyway13:37
CTtpollardlava can deploy qemu / kvm images13:37
paulsherwoodCTtpollard: so doesn't care about architecture?13:37
CTtpollardCTtpollard: I've had it deploy a debian x86 image and an arm image they provide, both worked13:38
steve__On QEMU just meant requirements are not so hard, no board to reboot etc.13:38
paulsherwoodCTtpollard: talking to yourself? :)13:38
CTtpollardwhere the host was a x86 laptop13:38
steve__ok good13:38
paulsherwoodsteve__: yup, but in practice i think if we only test on QEMU lots of hw-level issues maybe missed and land on users13:39
jeremiahYeah, part of the benefit is testing on real hardware13:39
CTtpollardin their lab, they have working jetson test beds13:39
paulsherwoodwho is they?13:40
CTtpollardlinaro that is13:40
paulsherwoodaha :)13:40
* gunnarx is soon returning to vacation mode13:40
paulsherwoodok... so wrt the middle bit, i think that's where things are less clear13:40
paulsherwoodgunnarx: ack. any strong recommendations for build and artifact publication?13:41
paulsherwood(before you disappear)13:41
gunnarxno, whatever works.  I've used Jenkins before but I'm not surprised if others are just as capable or more today13:42
paulsherwoodok, anyone else strong feelings on this?13:42
jeremiahI plan to use a Test Anywhere Protocol test for the SoW13:43
jeremiahTAP is an IETF standard, or is becoming one13:43
jeremiahThat should fit into Jenkins (or hopefully mustard or any framework.)13:43
paulsherwoodi think our soln needs to support multiple test approaches, though? we can't expect to dictate13:43
jeremiahpaulsherwood: You may well be right.13:43
paulsherwoodok, any more on CIAT? i'm concerned we're losing gunnarx :)13:44
gunnarxdon't mind me...13:44
jeremiahIn which case we have Jenkins --> * --> LAVA13:44
steve__personally I think if it does not require full time admin and a specialistic degree in its admin you are more likely to get others accepting it.13:44
paulsherwoodsteve__: is that an implied comment on Jenkins?13:44
gunnarxexactly.  There's little room for significant resource needs to keep this running13:45
paulsherwoodok, good point13:45
paulsherwoodhowever, in effect genivi is a subset of a 'distro'... and with GDP it becomes a whole distro...13:46
paulsherwoodand AGL is a distro too...13:46
jeremiahAnd then we get on to boiling the ocean13:46
paulsherwoodand distros require this infrastructure, and the resources to keepe them running13:46
steve__Depends on your goals. But assuming you are not trying to definitively provide a system for a 100+ production team, feels to me like a good straight forward setup that can easily be used by an in-house R&D team would have benefits for contribution13:46
paulsherwoodsteve__: we could set that as a criterion13:46
paulsherwoodjeremiah: fair enough :)13:47
steve__Jenkins makes it so easy to connect in tests you would need a pretty convincing argument not to provide tests for a contribution13:47
paulsherwoodbut it's java :)13:48
waltminerIs there an arguement against Jenkins somewhere in here?13:48
jeremiahIt's Java?13:48
* jeremiah goes silent.13:48
steve__Not that I made Walt :)13:49
CTtpollardAs is Gerrit :)13:49
jeremiahCTtpollard: Is it? I knew I didn't like it.13:49
paulsherwoodwaltminer: no, i don't think there is an argument against it, really13:49
waltmineryes it is Java.13:49
paulsherwoodexcept that other things are more attractive, maybe easier to setup13:49
jeremiahAnd, BTW, Oracle claims all use of the Java API is copyright Oracle. :-)13:50
* paulsherwood decides to move on from CIAT13:50
paulsherwood== Trello Clearup ==13:50
waltminerI guess I was hoping to see some kind of decision matrix for CIAT solutions13:51
paulsherwoodjeremiah: can you remember what triggered
paulsherwoodwaltminer: good idea13:51
paulsherwoodi'll take that as  a next action13:51
* paulsherwood notices that some things have been in review for a long time13:52
paulsherwoodare we intending to do anything with them?13:53
jeremiahpaulsherwood: I was thinking about that but I can't13:53
jeremiahpaulsherwood: I have created a GENIVI-esque Docker image based on Debian though13:53
jeremiahI suppose I can slap that URL in there and we can close that.13:53
paulsherwoodjeremiah: sounds like a plan!13:53
paulsherwoodjeremiah: but i think you should publish it on the wiki too?13:54
jeremiahI'm slowly, but steadily, building up a Debian based image. I'll wait till its finished for feedback13:54
*** gunnarx has left #automotive13:54
paulsherwood(and tell the ML?)13:54
jeremiahpaulsherwood: Okay, I'll do that.13:54
*** gunnarx has joined #automotive13:54
jeremiahI don't know how useful it is at the moment, hopefully it becomes more robust shortly. :-)13:54
jeremiahThen I'll send it to the list, meanwhile there's some stuff on the wiki.13:55
gunnarxso it's in "doing" I suppose.  Which card are we on?13:55
* paulsherwood thinks GDP is no longer a thing for tools team, so can delete that card?13:55
gunnarxyep, kill it13:55
paulsherwoodgunnarx: all of them. i asked jeremiah about the debian one, he's advancing that13:55
paulsherwoodi know fall amm planning is on me13:56
paulsherwoodlongterm strategy for franca and commonapi i'll leave where it is13:56
paulsherwoodnormal build integration is the new one13:56
paulsherwoodanyone want to add any new cards or suggest advances for existing ones, please fire at will :)13:57
* paulsherwood will move to last topic...13:57
paulsherwood== AOB ==13:57
gunnarxCommonAPI works in franca automation13:58
gunnarxI think.  Needs some more testers I'd say.13:58
paulsherwoodhow would we solicit those?13:58
gunnarxDon't know.  How are TT members using Franca / CommonAPI tooling?13:59
* paulsherwood is noticing that it's hard to armtwist volunteers at times13:59
paulsherwoodi don't know that unfortunately13:59
gunnarxYes, it's me and Klaus B testing this stuff.  There is limited resource.13:59
KlausUhlWe are integrating it from source ourselves.14:00
gunnarxAnyhow, it's there if someone wants to try it.14:00
paulsherwoodnoted... would you be interested in that, KlausUhl ?14:00
gunnarxintegrating it from source... you mean the source of the tools?14:00
gunnarxOK, so building your own Eclipse basically14:00
KlausUhlNot a complete Eclipse. Only the Franca and CommonAPI parts.14:01
gunnarxOK, downloading binary eclipse, building franca and common API from source, and then installing in eclipse?14:01
KlausUhlWe package them as an Eclipse update site from their sources ...14:01
gunnarxok, got it.14:01
KlausUhlessentiall, yes.14:01
gunnarxso the final step is a manual install in eclipse, but you can point that to your own update site?14:02
KlausUhlWe had to patch both sources from time to time.14:02
gunnarxbugs filed?14:02
KlausUhlBut we hope that we will be able to use the "stock" versions in the near future.14:02
KlausUhlYes, bugs are filed. I will also prepare our patches for upstreaming this week.14:02
gunnarxthanks for the info.  Anyway we'll keep pressing on with the automation - ultimately we need this to develop into a GDP SDK anyhow.14:03
paulsherwoodok, any other aob?14:03
* paulsherwood gently wonders if folks think we are discussing the right things in tools team, or if there are things we should change/drop/add to our agenda14:03
gunnarxI think the discussions are on things that need progress.  So if we can achieve progress then... :)14:04
jeremiahpaulsherwood: I think that the CIAT topic might benefit if it were broken out of TT and into its own (somewhat massive) project?14:04
pkonopelkopaulsherwood: anything particular? most things are in scope, imho14:04
paulsherwood(and this is a question to folks more widely than just the active members of the discussion... we would be happy to get feedback, suggestions)14:04
gunnarx+1. lurkers jump in.14:05
paulsherwoodpkonopelko: i'm just asking, really, to check i'm not accidenrtally driving in the wrong direction(s) :)14:05
jeremiahsteve__: Are you gonna hang around? I have some GLES patching fun. :-)14:05
* paulsherwood has heard in other fora that his diagram didn't come out right... so
rjek10 points to paulsherwood for using correct plural of forum.14:06
paulsherwoodok, well, on that basis i propose to end the formal meeting... folks are welcome to continue discussio n of course14:06
paulsherwood== GENIVI Tools Team Meeting Ends ==
  • No labels