*** astrophys has quit IRC | 00:14 | |
*** waltminer has quit IRC | 00:22 | |
*** bjones has quit IRC | 00:23 | |
*** Sisco has quit IRC | 00:35 | |
*** Sisco has joined #automotive | 00:38 | |
*** waltminer has joined #automotive | 01:18 | |
*** mdurnev has joined #automotive | 01:24 | |
*** sanjeev has quit IRC | 01:33 | |
*** waltminer has quit IRC | 01:35 | |
*** kooltux_ has quit IRC | 03:24 | |
*** kooltux_ has joined #automotive | 03:27 | |
*** Sisco has quit IRC | 03:29 | |
*** Sisco has joined #automotive | 03:30 | |
*** rosch has quit IRC | 04:11 | |
*** warter_ has joined #automotive | 04:11 | |
*** warter has quit IRC | 04:12 | |
*** rosch has joined #automotive | 04:12 | |
*** Sisco has quit IRC | 05:12 | |
*** Sisco has joined #automotive | 05:14 | |
*** AlisonChaiken has joined #automotive | 06:06 | |
*** mps_ has joined #automotive | 06:17 | |
*** fredcadete has quit IRC | 07:01 | |
*** mpaolino has joined #automotive | 07:04 | |
* CTtpollard yawns | 07:17 | |
*** leon has joined #automotive | 07:36 | |
*** leon is now known as Guest21599 | 07:36 | |
*** jobol has joined #automotive | 07:38 | |
*** jobol has joined #automotive | 07:40 | |
warter_ | is there a mailing list for the graphics and UI expert group? | 08:10 |
---|---|---|
*** warter_ is now known as jbocklage1 | 08:10 | |
*** picoblw has joined #automotive | 08:11 | |
*** bruce_ has joined #automotive | 08:18 | |
*** picoblw has quit IRC | 08:22 | |
*** jonathanmaw has joined #automotive | 08:27 | |
*** fredcadete has joined #automotive | 08:34 | |
rjek | And another: http://www.bbc.co.uk/news/business-36089558 | 08:40 |
paulsherwood | oh dear | 08:44 |
paulsherwood | ethics matter | 08:45 |
*** steve_l has joined #automotive | 09:01 | |
*** gunnarx has joined #automotive | 09:01 | |
*** gunnarx has joined #automotive | 09:01 | |
gunnarx | Hi | 09:01 |
*** weson has joined #automotive | 09:01 | |
weson | hi all | 09:02 |
gunnarx | No lead for the ToolTeam meeting available today? | 09:02 |
weson | I was testing the latest GDP today morning. Please see the latest failures.WARNING: Failed to fetch URL svn://anonymous@navit.svn.sourceforge.net/svnroot/navit/trunk;module=navit;protocol=http, attempting MIRRORS if available ERROR: Fetcher failure: Fetch command failed with exit code 1, output: svn: E000111: Unable to connect to a repository at URL 'http://navit.svn.sourceforge.net/svnroot/navit/trunk/navit' svn: E000111: Error r | 09:03 |
steve_l | gunnarx: see there were some more email on ML. Various ppl can't join | 09:03 |
gunnarx | toscalix sent an invite but won't be leading it. kbirken here? | 09:03 |
gunnarx | steve_l, ah | 09:03 |
steve_l | gunnarx: klaus can't make it | 09:03 |
CTtpollard | weson: please try to force the package | 09:04 |
gunnarx | weson, CTtpollard. Hmm I thought we had switched navit to the git url... | 09:04 |
gunnarx | force how, CTtpollard? | 09:04 |
weson | CTtpollard: can you share the command? | 09:04 |
weson | wih -C -F? | 09:04 |
CTtpollard | nope navit is still svn | 09:06 |
gunnarx | OK, concluding that we don't have a formal TT meeting today. Could have a brief informal discussion if someone has something to report. | 09:06 |
CTtpollard | bitbake -f -c compile navit | 09:06 |
paulsherwood | well, i could report that the above is a clear example of the mirroring problems i tried to solve last year | 09:07 |
steve_l | I remember that navit had moved its hosting so the URL redirects, but I guess the recipe was updated for that... If not, possible they moved again and the redirect is broken. Its a while since I downloaded again | 09:07 |
gunnarx | that's not a report :) | 09:08 |
CTtpollard | although your error weson could be due to being behind a firewall | 09:08 |
paulsherwood | gunnarx: no, it's not, and i'm no longer in the tools team. that doesn't make me wrong | 09:08 |
paulsherwood | :) | 09:08 |
gunnarx | no you were never "wrong", just the question was intertwined with many other things | 09:09 |
CTtpollard | I can't generate the navit fetch error personally, so it is more than likely a network hiccup, or a firewall issue | 09:09 |
gunnarx | point is paulsherwood that we could keep a mirror to avoid the above but it is not a solution for "car companies that want repeatable builds" as you stated | 09:09 |
gunnarx | because car companies have their own mirrors. | 09:09 |
gunnarx | they have no reason to trust GENIVI more than any other outside entity.... | 09:10 |
gunnarx | the recent failure of GENIVI git server shows that | 09:10 |
paulsherwood | gunnarx: i'm not going to go around this again :) | 09:10 |
gunnarx | ok | 09:10 |
gunnarx | we're in agreement though. I have no problem with GENIVI keeping mirrors. First step would be to move navit URL to git though, since it exists | 09:11 |
gunnarx | Main problem seems to be that everyone keeps fetching the same source code over and over, it's some weird artifact of Yocto build culture | 09:11 |
* CTtpollard tries not to do that | 09:12 | |
gunnarx | CTtpollard is a good Yocto builder | 09:13 |
CTtpollard | fsvo good :) | 09:13 |
gunnarx | Very high on my list now is the wish to standardize local mirror on all Go agents. I know CT does this but it's not everywhere | 09:14 |
gunnarx | However, I have first to finalize migration of Go server. Very close now. | 09:14 |
steve_l | +1 | 09:14 |
CTtpollard | +1 | 09:15 |
pedroalvarez | gunnarx: "standardize local mirror" ? what's that :/ | 09:17 |
gunnarx | :) | 09:17 |
gunnarx | Stating that /srv/dl is the path at which downloads should be stored on all agents, and subsequently fetched by the pipeline definitions | 09:18 |
gunnarx | that's all basically | 09:18 |
gunnarx | maybe some other name than /srv/dl can be agreed - anything that is very unlikely to clash with anything existing on any typical system... | 09:18 |
rjek | /tmp/genivi-dl/ | 09:19 |
rjek | Or something | 09:19 |
rjek | 'dl' is a bit vague | 09:19 |
steve_l | you will need to pad that 1 page of content with another 789 pages to call it a standard | 09:19 |
gunnarx | Yes but /tmp is tricky, sometimes it's a limited space etc. | 09:19 |
rjek | /var/spool/genivi/downloads | 09:19 |
gunnarx | and it might be wiped. I think the point is it should be persistent | 09:20 |
rjek | gunnarx: Which path has unlimited space? :D | 09:20 |
gunnarx | My /volumes/ does :) | 09:20 |
rjek | Have you backed up the internet to it yet? :) | 09:20 |
steve_l | Just the cat videos | 09:20 |
gunnarx | if we agree on the path then the recipes themselves could define a mirror variable in the bb recipes. If it's there it's used, if not it's not | 09:20 |
gunnarx | Simple. I put one of those chinese USB sticks there. When it gets to the end, it loops around. | 09:21 |
rjek | heh | 09:21 |
gunnarx | (No offence intended to chinese USB manufacturers that actually work) | 09:22 |
CTtpollard | gunnarx: we can pre mirror in local.conf for each branch/build | 09:22 |
rjek | Well, /var/spool or /var/cache strike me as idiomatic | 09:22 |
rjek | (As defaults) | 09:22 |
*** mpaolino has quit IRC | 09:23 | |
*** mpaolino has joined #automotive | 09:23 | |
gunnarx | /var/cache/gdp_downloads ? | 09:23 |
*** jeremiah_ has joined #automotive | 09:24 | |
jeremiah_ | ohai! | 09:24 |
fredcadete | CTtpollard: why ore mirror over DL_DIR? (http://www.yoctoproject.org/docs/1.8/mega-manual/mega-manual.html#var-DL_DIR) | 09:24 |
fredcadete | I've been sharing DL_DIR between builds on the same machine and it works fine; I have not tried pre mirror | 09:24 |
rjek | gunnarx: /var/cache/gdp/downloads perhaps; you never know what else might want caching in future :) | 09:25 |
gunnarx | better | 09:25 |
gunnarx | yocto/downloads... | 09:25 |
CTtpollard | fredcadete: dl_dir is also a sane posibilty (I also share dl_dir locally) | 09:25 |
gunnarx | That's the question. DL_DIR is required to exist right? | 09:26 |
* jeremiah_ lurks | 09:26 | |
gunnarx | I was thinking MIRROR is used if it is there and if it's not it still works | 09:26 |
gunnarx | ? | 09:26 |
fredcadete | I think so. if you pre-mirror it will fetch from your mirror but still put things in a download dir | 09:26 |
*** jeremiah_ has quit IRC | 09:27 | |
fredcadete | I *think* | 09:27 |
gunnarx | PREMIRRORS is the variable | 09:27 |
gunnarx | fredcadete, right. It uses space still but I imagine it's acceptable. | 09:27 |
CTtpollard | dl_dir defaults to $builddir/downloads | 09:28 |
CTtpollard | http://www.yoctoproject.org/docs/1.8/ref-manual/ref-manual.html#var-PREMIRRORS | 09:30 |
gunnarx | OK, so option 1: Setting PREMIRRORS to /var/cache/gdp/yocto/downloads in recipes. Encouraging agents to provide persistent storage at that location. That means files are copied from the shared space to downloads for each build but we avoid the network traffic. | 09:30 |
CTtpollard | personally I'd probably vote for mirrors (so it tries to pull from upstream first, less load on the genivi mirror) | 09:31 |
gunnarx | option 2. Setting DL_DIR to /var/cache/gdp/yocto/downloads It means it must exist and Go Agents that don't provide it will fail. Downloads will be shared across all GDP builds run on that agent. | 09:31 |
fredcadete | gunnarx: right, but that space grows. My downloads dir is now at 14G. If you multiply this by all variants, it adds up. You also add the copying from your premirror to downloads. | 09:31 |
gunnarx | in option 2 the DL dir is shared among all variants so will grow to your 14G something | 09:33 |
gunnarx | in option 1 you manually need to fill the content of the PREMIRROR location though | 09:33 |
gunnarx | + as you say you will get duplicates for each pipeline. However this is already the case today and the source big but not the biggest part - all the build artifacts are IIRC | 09:34 |
fredcadete | to save fetching and disk space on the same agent -> DL_DIR. to have a single GDP mirror that will maybe be more reliable than upstream fetches -> PRE_MIRROR | 09:34 |
fredcadete | to save fetching and disk space on the same agent -> DL_DIR. to have a single GDP mirror that will maybe be more reliable than upstream fetches -> PRE_MIRROR | 09:35 |
CTtpollard | are you wanting to have the pre_mirror openly available and set by default? | 09:35 |
gunnarx | fredcadete, yes. I'd say option two also also saves fetching (over network) though, it copies locally since we're talking about a local mirror. | 09:36 |
gunnarx | fredcadete, maybe not clear I'm not talking about the reliability problem primarily actually | 09:36 |
fredcadete | yes | 09:36 |
gunnarx | My first point was to reduce network traffic and speed up builds | 09:37 |
fredcadete | oh, ok | 09:37 |
gunnarx | but reliability might be a good offshoot | 09:37 |
fredcadete | for that use case DL_DIR has been working fine for me | 09:37 |
gunnarx | But are you running a Go Agent? (just putting it in context) | 09:37 |
gunnarx | As we said above both are possible for Go Agent | 09:38 |
fredcadete | no, sorry :) | 09:38 |
fredcadete | no, sorry :) | 09:39 |
gunnarx | Checked a current GDP build on real silicon - 4GB DL DIR, QEMU is around 3. | 09:39 |
gunnarx | So I imageine 14G is if you let it accumulate over many versions and variants... | 09:39 |
fredcadete | just many builds for troubleshooting: vanilla poky, yocto build as recommended by the silicon provider, a GDP build, variants of our internal distro | 09:39 |
gunnarx | So we should expect 4*(number of variants) or 14. Not 14*(number of variants) necessarily... | 09:40 |
fredcadete | yes it is accumulated | 09:40 |
gunnarx | fredcadete, good point and I realized it may be good to avoid specializing the cache path too much | 09:40 |
gunnarx | many may want to build different systems that are similar | 09:40 |
fredcadete | also over a year or so, so it has caught two poky upgrades and for many packages there are a couple of versions | 09:40 |
fredcadete | you may want to consider this too :/ | 09:40 |
gunnarx | at the same time I'd like to be able to roll it out directly in the configs without clashing with something | 09:41 |
fredcadete | you mean clashing with things people may have in their local.conf? | 09:42 |
gunnarx | no just choosing a path that isn't used already | 09:43 |
*** Sisco has quit IRC | 09:43 | |
gunnarx | those that have the need to edit their conf manually will sort it out. but for GDP I'm eager it works out of the box with no manual fiddling | 09:43 |
gunnarx | and the third aspect is that it should build successfully on any Go Agent. | 09:43 |
pedroalvarez | I asked in #yocto a while back about this, and some people with CI setups were using DL and SSTATE dirs to share all the things that were possible | 09:43 |
pedroalvarez | and every weekend cleaning everything from them | 09:44 |
gunnarx | I still don't trust sstate sharing | 09:44 |
fredcadete | pedroalvarez: yes, SSTATE also speeds up a lot, I discovered that recently though so I don't have it set up | 09:44 |
gunnarx | sorry, in theory I guess it's supposed to be OK but I've seen too much weirdness | 09:44 |
gunnarx | At least I can confirm that an interrupted yocto build can end up being unbuildable until you wipe state and other things. | 09:45 |
gunnarx | That said, for basic CI I might be OK with it. | 09:45 |
gunnarx | If it works it works. If it fails, we'll know. CI is not about product quality (yet) :) | 09:46 |
pedroalvarez | gunnarx: that was one of the reasons they had to clean up sstate sometimes | 09:46 |
gunnarx | ok | 09:46 |
CTtpollard | we've had to kill the sstate on the codethink agents before due to sharing weirdness | 09:46 |
pedroalvarez | hm.. I wonder if it's possible to run a job in all the agents to syncronise the cleanup every weekend | 09:46 |
jeremiah | Shave that yak! | 09:47 |
pedroalvarez | yeah.. better use baserock :P | 09:47 |
gunnarx | :) | 09:47 |
jeremiah | Can we run baserock on a Go.CD instance? | 09:48 |
gunnarx | pedroalvarez, there's no separation between pipelines AFAIK. We can have a manually triggered pipeline that deletes stuff on the agent. | 09:48 |
gunnarx | Set each agent up with its own resource and there can be one job per agent. Not very pretty but it should be quite OK. | 09:49 |
gunnarx | s/job/pipeline/g | 09:49 |
gunnarx | s@s/job/pipeline/g@s/job/pipeline/@ :) | 09:49 |
pedroalvarez | it could be less ugly if we can make it run automatically | 09:50 |
pedroalvarez | jeremiah: it should be possible yes | 09:50 |
gunnarx | yes but that also means installing that solution on all agents, which makes them specialized. I think simplicity there is quite important. | 09:50 |
pedroalvarez | gunnarx: nope, I was just suggesting that gocd could run the job for us, instead of having to click a button of the manually triggered pipeline | 09:52 |
gunnarx | pedroalvarez, forget what I said. You could set them up as pipelines, not cron-jobs on agents. You can time-trigger pipelines. | 09:52 |
gunnarx | You got it | 09:52 |
pedroalvarez | ppeeeeeeeerfect | 09:52 |
steve_l | off to work. but my 2c on above. maybe avoid sstate sharing across pipelines as first step - as others note it can be cause of issues when mixing machines. | 10:10 |
steve_l | premirror would be useful to many people outside of go.cd. shared downloads folder on go is obvious first step. | 10:12 |
CTtpollard | after the genivi git woes, having a central premirror scares me, I think a normal mirror (checked after upstream) is safer and probably faster | 10:13 |
paulsherwood | what is a 'normal mirror' ? | 10:13 |
CTtpollard | paulsherwood: bitbake checks dl_dir - pre mirror - upstream - mirror | 10:14 |
CTtpollard | so if we move to pre_mirror default everything, it needs to be robust and stable | 10:15 |
steve_l | that's one nice thing about yocto premirror entirely optional. If I drop it from my local.conf it just goes to the upstream. If the premirror isnt working it will also go to upstream. but yes your right. should look at the needs and decide what is best. | 10:16 |
steve_l | low hanging fruit is shared DL folder :) | 10:16 |
CTtpollard | and controlled by genivi, so we don't have to go through layers of loops to get things fixed :) | 10:17 |
*** steve_l has quit IRC | 10:18 | |
weson | CTtpollard: sorry, I missed the updates. What is the reason behind the navit failure on latest GDp for rasp? | 10:22 |
CTtpollard | weson: I believe it was either a network hiccup (why I asked you to try and force the recipe) or your firewall does not like it | 10:24 |
CTtpollard | as gunnarx suggested, we should move the recipe to a git source, if you wish to add that to a jira ticket :) | 10:24 |
CTtpollard | or send a patch ofc | 10:25 |
gunnarx | Yes, https://github.com/navit-gps/navit.git as the source location | 10:28 |
gunnarx | ^^ is the official source location according to home page, I meant to say | 10:28 |
CTtpollard | great | 10:29 |
weson | ok, so what is the temporary workaround to pass the build? | 10:30 |
CTtpollard | weson: have you tried pulling from the svn link manually on your system? | 10:31 |
weson | CTtpollard: let me check it.. Please hold | 10:32 |
weson | unable to connect, connection refused. | 10:43 |
weson | svn co <svn://link to navit> | 10:43 |
CTtpollard | sounds like it's your firewall then | 10:43 |
CTtpollard | or the svn server not liking your IP | 10:43 |
weson | ok, then I will try from home :) | 10:45 |
CTtpollard | weson: for now you could use an append, or manually edit the recipe to pull from the git repo | 10:46 |
weson | CTtpollard: git address please | 10:47 |
CTtpollard | https://github.com/navit-gps/navit.git | 10:47 |
weson | ok, where I need to change? | 10:47 |
CTtpollard | navit_svn.bb, you'll probably have to find the corresponding commit, or w/e term svn call it | 10:49 |
weson | I will check this file in the GDP folder and will change the link to take it from git instead of svn | 10:50 |
weson | Thank you. | 10:50 |
weson | brb | 10:50 |
*** weson has quit IRC | 11:58 | |
*** gunnarx has quit IRC | 12:05 | |
*** CTtpollard has quit IRC | 12:16 | |
*** CTtpollard has joined #automotive | 12:19 | |
*** waltminer has joined #automotive | 13:10 | |
*** hammer_gaidin has quit IRC | 13:19 | |
*** mvick has quit IRC | 13:19 | |
*** mdurnev has quit IRC | 13:50 | |
*** Sisco has joined #automotive | 13:52 | |
*** CTtpollard has quit IRC | 14:02 | |
*** CTtpollard has joined #automotive | 14:16 | |
andrunko | hi, I've created SPEC-172 on jira to add support for some genivi components to agl - if someone could please take a look and let me know what I should do next to push the changes I'd appreciate, tnx | 14:31 |
*** jobol has quit IRC | 14:42 | |
*** Sisco has quit IRC | 14:44 | |
*** tjamison has joined #automotive | 14:45 | |
*** astrophys has joined #automotive | 14:49 | |
*** Sisco has joined #automotive | 14:54 | |
CTtpollard | Weekly GDP call is about to start | 14:59 |
*** gunnarx has joined #automotive | 15:01 | |
*** gunnarx has joined #automotive | 15:01 | |
*** Sisco has quit IRC | 15:02 | |
CTtpollard | Can anybody here us in the call? | 15:05 |
CTtpollard | gunnarx ^ | 15:05 |
gunnarx | Yes I hear you | 15:05 |
CTtpollard | hmm, will try a different device | 15:05 |
CTtpollard | brb | 15:05 |
CTtpollard | I heard you join | 15:06 |
CTtpollard | odd | 15:06 |
*** smiller6 has joined #automotive | 15:06 | |
gunnarx | np, we will wait :) | 15:06 |
*** amcgee7 has joined #automotive | 15:07 | |
*** mvick has joined #automotive | 15:08 | |
gunnarx | We're speaking | 15:12 |
gunnarx | You're not hearing us CTtpollard | 15:12 |
waltminer | my wife says that to me a lot | 15:13 |
gunnarx | weird... you have audio only in one direction gunnarx | 15:13 |
gunnarx | waltminer, do you use Webex to communicate with your better half? | 15:13 |
gunnarx | I hope not :) | 15:13 |
gunnarx | CTtpollard, we can hear you fine but you don't hear us | 15:14 |
CTtpollard | it looks like we're having problem then | 15:14 |
gunnarx | yes | 15:14 |
CTtpollard | will try and reconnect one more time | 15:14 |
gunnarx | It's like we're in two separate calls with a one-way bridge :) | 15:14 |
gunnarx | Can you hear us?? | 15:14 |
gunnarx | oh no :( | 15:15 |
gunnarx | Try the volume control | 15:15 |
*** jbocklage1 has quit IRC | 15:24 | |
*** warter has joined #automotive | 15:25 | |
CTtpollard | thanks everyone | 15:48 |
CTtpollard | and for baring with the technical difficulties | 15:48 |
*** bruce_ has quit IRC | 15:51 | |
CTtpollard | leon-anavi-cloud: are you going to the AMM? | 15:54 |
*** Sisco has joined #automotive | 15:59 | |
*** gan has joined #automotive | 16:00 | |
*** gan has joined #automotive | 16:00 | |
*** gunnarx has quit IRC | 16:01 | |
*** Guest21599 is now known as leon-anavi | 16:02 | |
*** AlisonChaiken has quit IRC | 16:08 | |
leon-anavi | CTtpollard, yes, I will arrive on Monday (25 april) and I will go back home on 28 April. | 16:08 |
leon-anavi | CTtpollard, so I will be there for the meetings & sessions on 26-27 April. | 16:09 |
*** waltminer has quit IRC | 16:12 | |
*** leon-anavi has quit IRC | 16:16 | |
*** gan has quit IRC | 16:18 | |
*** fredcadete has quit IRC | 16:20 | |
*** jonathanmaw has quit IRC | 16:33 | |
*** mpaolino has quit IRC | 16:35 | |
*** bjones has joined #automotive | 16:55 | |
*** tjamison has quit IRC | 17:06 | |
*** tjamison has joined #automotive | 17:06 | |
*** waltminer has joined #automotive | 17:17 | |
*** fury- has joined #automotive | 17:44 | |
*** fury has quit IRC | 17:44 | |
*** fury- is now known as fury | 17:44 | |
*** praneeth_ has quit IRC | 17:59 | |
*** praneeth_ has joined #automotive | 17:59 | |
*** gan has joined #automotive | 18:32 | |
*** gan has quit IRC | 18:45 | |
*** bjones has quit IRC | 18:49 | |
*** gan has joined #automotive | 19:08 | |
*** smiller6 has quit IRC | 20:12 | |
*** gan has quit IRC | 20:25 | |
*** mps_ has quit IRC | 20:56 | |
*** waltminer has quit IRC | 21:26 | |
*** waltminer has joined #automotive | 22:16 | |
*** Sisco has quit IRC | 23:21 | |
*** amcgee7 has quit IRC | 23:27 | |
*** waltminer has quit IRC | 23:37 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!