IRC logs for #baserock for Monday, 2016-02-29

*** gtristan has quit IRC04:31
*** gtristan has joined #baserock04:57
*** franred has joined #baserock07:21
*** jjardon_ has joined #baserock07:43
*** toscalix has joined #baserock07:49
jjardon_Hi, what is the best approach if I want to compile all the chunks with some specific CFLAGS? Change DEFAULTS? Change the build tool code (I see we do that fore the MAKEFLAGS)??07:50
franredjjardon_, I would say that DEFAULTS is the place to modify that. Said that, the change is not trivial an you should aim to at least manage to build all the packages in any architecture defined that currently built to avoid a massive regression when build the rest of the systems and not the ones you are interested in.08:25
*** ctbruce has joined #baserock08:27
paulsherwoodjjardon_: if you want to make a branch, i can build ci for you on aws08:28
paulsherwoodjjardon_: not sure what you mean by build tool code vs MAKEFLAGS. iirc currently the sandboxing strips any env, and -j is set by default based on cpu_count(), except where it's forced to 1 (max-jobs: 1)08:35
jjardon_yeah, so MAKEFLAGS is set in the code, not in a declarative way (even I think in this specific case makes sense as you have to detect the number of cores available))08:36
jjardon_for CFLAGS/CXXFLAGS I think is a bit different; distros normally default to some optimized ones (-02 -march=native ...) for normal image or debug ones (-g -O0) if you want to generate debug information08:38
paulsherwoodi guess we could support max-jobs: [cpu_count, 1, n]08:39
paulsherwood(defaulting to cpu-count)08:39
jjardon_paulsherwood: that would be great I think08:42
jjardon_about CFLAGS/CXXFLAGS, I think baserock should default to some sane/safe ones, and let users to modify that value if desired08:43
paulsherwoodjjardon_: actually, ybd already does that, defaulting to cpu_count :)08:44
paulsherwoodjjardon_: +1 for CFLAGS/CXXFLAGS08:44
jjardon_paulsherwood: nice. so , if I want to modify the default max-jobs, can I do that changing a parameter in DEFAULTS?08:47
*** rdale has joined #baserock09:03
*** bashrc_ has joined #baserock09:06
paulsherwoodjjardon_: not yet :) - we'd need to fix DEFAULTS09:22
*** jonathanmaw has joined #baserock09:37
*** jjardon_ has quit IRC09:38
pedroalvarez_would be MAKEFLAGS something for DEFAULTS or for morph.conf/ybd.conf?09:41
*** tiagogomes has joined #baserock09:45
pedroalvarez_will different MAKEFLAGS modify the output?09:45
paulsherwoodhopefully not :)09:45
paulsherwoodbut in some cases it does - if the build cannot cope with parallel09:46
pedroalvarez_also, if we use them correctly in definitions09:47
pedroalvarez_same thing with CFLAGS/CXXFLAGS?09:47
pedroalvarez_radiofree, jjardon: stock weston works just fine. There is something wrong with how genivi-weston runs :(09:49
franredpedroalvarez_, it is interesting. so you suggestion is that the build tools define their compile flags, being definitions agnostic about it?09:49
radiofreepedroalvarez_: that's odd!09:50
pedroalvarez_franred: I'm just thinking, that maybe in some cases, you want to build something with max-jobs 1, because you don't have enough hardware.. but then you want to use exactly the same definitions to build with max-jobs:15, because you have a massive build server09:51
jjardonpedroalvarez_: different CFLAGS will produce different binaries, yes09:52
pedroalvarez_and for that... you shouldn't need to change any definitions09:52
paulsherwoodhmmm.... cache-keys need to be affected by CFLAGS09:54
*** ssam2 has joined #baserock09:54
*** ChanServ sets mode: +v ssam209:54
pedroalvarez_radiofree: it is, yes. I'm going to try 4.5-rc5 kernel and latest nouveau oot kernel module09:54
pedroalvarez_paulsherwood: that's what I was thinking09:54
*** gtristan has quit IRC09:57
paulsherwoodpedroalvarez_: hence i think they should be in DEFAULTS09:57
*** edcragg has joined #baserock10:00
*** CTtpollard has joined #baserock10:03
pedroalvarez_- CFLAGS: "-D_GNU_SOURCE -static"10:06
pedroalvarez_- CXXFLAGS: "-foo -B_AR"10:06
ssam2allowing a generic way to set build-environment inside DEFAULTS seems a nice solution10:08
ssam2just allowing CFLAGS and CXXFLAGS would fine, but then one day someone would need to set LDFLAGS, etc.10:08
jjardonSetting CFLAGS, CXXFLAGS and LDFLAGS should be enough I think10:11
tiagogomesHow will that play with CFLAGS already defined in the chunk morphs? Will the values of CFLAGS in DEFAULTS and CFLAGS in the chunk morph be joined together?10:15
tiagogomesThe environment spec should also be a dictionary, and not a list of dictionaries10:16
ssam2tiagogomes: it might break things10:19
franredtiagogomes, I would suggest to overwrite the flags with the ones written in the chunk(definition)10:19
ssam2tiagogomes: but most likely it wouldn't10:19
ssam2tiagogomes: chunks could easily do "CFLAGS=$CFLAGS -O0" or whatever if they want10:19
tiagogomesssam2 that sounds good10:20
pedroalvarez_should CFLAGS change the cache key of everything? or only of the chunks that are actually using them>10:24
jjardonpedroalvarez_: everything10:24
pedroalvarez_well, I guess something at a really low level uses them anyway :)10:24
jjardonthe idea is that I will build the whole system with -O2 -march=native , for example10:24
pedroalvarez_tiagogomes: I almost put another '-' to environment :)10:29
ssam2if cache keys are implement sensibly, this shouldn't require any change to how they are calculated10:31
ssam2the entire build environment should be included in the cache key10:31
*** gtristan has joined #baserock10:37
*** locallycompact has joined #baserock11:09
*** franred has quit IRC11:30
*** franred has joined #baserock11:43
*** rjek_ is now known as rjek12:19
*** gtristan has quit IRC13:50
*** bashrc_ has quit IRC14:04
*** bashrc_ has joined #baserock14:05
*** gtristan has joined #baserock15:03
*** ctbruce has quit IRC15:06
locallycompactHi peeps. Shot in the dark - pedroalvarez_ was kind enough to add me to the gerrit Non-Interactive Users group so I could test this stream-events thing on gerrit, but it seems to timeout after exactly 50 seconds every time:16:00
locallycompact'Connection to closed by remote host.'16:00
rjekHow rude.16:01
paulsherwoodwe seem to short of reviews, still....16:01
* richard_maw doesn't have time to trawl through things to review of his own volition but can spare a few minutes here and there if people ping him links16:02
paulsherwoodlocallycompact: could you help with some reviews, while you wait for help on that? :)16:02
* locallycompact doesn't mind trawling if it's a priority16:02
paulsherwoodwell, i think there's updates to things that jjardon is working on16:03
*** pedroalvarez_ is now known as pedroalvarez16:08
pedroalvarezlocallycompact: I wonder if we could get some help in #gerrit16:09
ssam2locallycompact: do you get any events? it doesn't look like it from that log..16:10
locallycompactno events16:10
locallycompactbut possible there were none in that timespan16:10
ssam2i thought that stream-events was available for everyone, because doesn't gertty require that?16:10
pedroalvarezevents arrive if any16:10
ssam2is anyone using gertty?16:10
* SotK thought it was too16:10
pedroalvarezssam2: nope, gertty uses download-commands plugin16:10
locallycompactI had authorisation problems without being in that user group16:10
ssam2ah, right16:11
locallycompactif you do just ssh -p 29418 gerrit stream-events16:11
pedroalvarezI've tested, ant events arrive if any, but times out after 50 seconds of inactivity16:11
*** faybrocklebank has quit IRC16:11
ssam2is your gerrit username the same as your local username?16:12
locallycompactah, it is not16:12
locallycompactok now it's different16:12
ssam2i'm getting events16:13
locallycompactI have events16:13
paulsherwoodw00t :)16:13
locallycompactsee if timeout16:13
pedroalvarezwho is making noise in gerrit!16:13
*** ctbruce has joined #baserock16:13
pedroalvarezbut it still times out, right?16:14
locallycompactit still times out yeah16:14
pedroalvarezlocallycompact: maybe now that pygerrit script you showed me works?16:14
locallycompactdoesn't seem to be printing anything eventy16:14
locallycompactif it's supposed to16:15
pedroalvarezlocallycompact: because nothing is happening in gerrit I guess16:16
paulsherwoodreviews! :)16:16
locallycompactI'm trying with python -g -u locallycompact16:17
ssam2you should have just seen some events16:19
locallycompact2016-02-29 16:17:05,703 INFO No event16:19
locallycompact2016-02-29 16:17:06,705 INFO No event16:19
locallycompact2016-02-29 16:17:07,707 INFO No event16:19
locallycompact2016-02-29 16:17:08,708 INFO No event16:19
locallycompactuntil timeout16:19
pedroalvarez2016-02-29 16:19:37,028 INFO Event: <UnhandledEvent>16:19
ssam2this is what I saw:
ssam2and I definitely have been connected more than 50 seconds16:20
pedroalvarezIME after 50 seconds from last event, timeout!16:20
ssam2especially weird since you and me should be in all the same groups16:21
ssam2no timeouts for me16:21
paulsherwoodlocallycompact: what do you mean by 'opopopop' ?16:25
*** franred has quit IRC16:26
* paulsherwood notes that posting frivolous review messages may or may not endear one to others in the upstream community16:26
locallycompactit's roughly equivalent to "move it along"16:27
paulsherwoodeek.... leaking my creds all over the interwebs,, thanks :)16:27
pedroalvarezcreds ?!?!16:28
paulsherwood(and yours too, locallycompact )16:28
locallycompactWhat creds?16:28
paulsherwoodwell, email address, name, nick16:28
pedroalvarezthe email is public in gerrit already16:28
paulsherwoodtrue. i'll get my coat :)16:28
pedroalvarezheh :) no need16:28
pedroalvarez**you are being watched**16:29
ssam2oh, this might be haproxy that closes the connection16:29
ssam2    timeout client 50000ms16:29
ssam2    timeout server 50000ms16:29
ssam2that is suspiciously like 50 seconds...16:29
pedroalvarezgood catch16:30
ssam2locallycompact: i've changed the timeout for ssh to gerrit to 1 hr, has that improved matters?16:35
* locallycompact relaunches16:36
locallycompactIt has not16:38
locallycompact50 secs still16:38
ssam2this is log from haproxy for all the ssh connections to gerrit in that time:]16:41
ssam2no idea what it says though :-)16:41
locallycompactThat didn't seem to link anywhere16:42
ssam2the 'cD' seems the significant part16:43
ssam2"c : the client-side timeout expired while waiting for the client to16:43
ssam2            send or receive data."16:43
ssam2so, I guess it is haproxy messing things up16:43
locallycompactfascinating log16:44
ssam2except that I put "    timeout client 1h"16:44
ssam2for that backend16:44
ssam2oh, 'timeout client' goes in a different section16:45
ssam2ok, try again now16:45
locallycompactso far so good16:48
locallycompactssam2, looks good16:50
locallycompactssam2, OOI, is it possible to make that infinite?16:50
pedroalvarezI was actually thinking about t hat. 1h timeout is not long enough16:50
pedroalvarezbut the code should handle that too I think16:51
ssam2hmm.. no, it can't be infinite16:51
ssam2i could make it 1 day or something16:51
ssam2but yeah, your code can't assume network connections will last forever anyway16:51
*** fay has joined #baserock17:20
*** fay is now known as Guest6602917:20
*** ctbruce has quit IRC17:21
* paulsherwood is building devel-system-armv8l64 on jetson TX117:26
paulsherwood16-02-29 00:02:35 [0/372/372] [stage1-binutils] Elapsed time for build of stage1-binutils.7cff4748c65f817a269a40ea3a7b40fa5fd559b87820e523cca042426ef0efce 00:02:0517:28
pedroalvarezI wonder if that would still build hehe17:28
*** Guest66029 is now known as faybrocklebank17:29
* richard_maw would be interested to know whether the TX1 proves to be a useful build machine17:29
paulsherwoodcpu_count = 417:29
CTtpollardit's quite beastly17:31
CTtpollardand 4GB of ddr417:31
CTtpollardnut sure if any of the cuda/visual stuff is of any use though17:32
locallycompactuse for what?17:32
locallycompactI like cuda17:32
paulsherwoodi'm sure it's of use, but not for building sw :)17:32
CTtpollardlocallycompact: building (not in general)17:32
* CTtpollard would prefer OpenCL17:35
*** toscalix has quit IRC17:36
CTtpollardanyway, that's a different conversation17:36
paulsherwoodrjek: any ideas about that? or edcragg ?17:40
pedroalvarezthis is were all the fun starts17:42
paulsherwoodit's been fun here for years already :)17:43
edcraggpaulsherwood: maybe no armv8 support in baserock? but i thought that had already been patched for the moonshot17:43
* richard_maw hopes that error isn't that the host C compiler can't compile aarch64 code17:44
pedroalvarezI hope it's not related to the `case "$MORPH_ARCH" in` workaround in stage1-gcc17:45
pedroalvarezalthough that one would be easy to fix17:46
pedroalvarezYou could try to build the definitions from when aarch64 was merged17:46
edcraggi don't see an armv8 case in the error for stage1-gcc17:47
*** jonathanmaw has quit IRC17:50
edcraggyeah looks like gcc flags aren't set for armv8 in master17:51
edcragghave we not been building on armv8?17:52
paulsherwoodaarch64, iirc17:52
edcraggsame thing, surely?17:54
ssam2i don't think so17:55
edcraggwell armv8 can run in 32 bit mode...17:56
paulsherwoodedcragg: yup, but where's the fun in that? :)17:56
paulsherwoodedcragg: i believe you have a similar board :) patches (to definitions) gratefully accepted17:56
edcraggi did patch that previously as it happens but never got a chance to test it before the moonshot was taken away17:57
paulsherwoodmoonshot should be back online soonish17:57
paulsherwoodedcragg: i could test your patches?17:57
edcraggpaulsherwood: just finding them now17:58
pedroalvarez2016-02-29 16:24:43 Using git clone.17:59
pedroalvarezERROR: Cannot find remote git repository:
pedroalvarezjjardon: ^ :P17:59
*** bashrc_ has quit IRC18:00
pedroalvarezdefinitions v8 will solve this problem I guess18:01
paulsherwoodpedroalvarez: what are you building to trigger that?18:01
pedroalvarezThis is Mason trying to build lastest Gnome system18:02
paulsherwoodpedroalvarez: in master?18:02
pedroalvarezMason is still alive, yes :)18:02
pedroalvarezpaulsherwood: yes, master18:03
pedroalvarezafter "strata/xdg-app-common: Upgrade xdg-app to 0.4.13"18:03
paulsherwoodack. i guess i need to slap myself... did i review that?18:03
pedroalvarezplease, don't slap anyone. Reviewing that the repo being used doesn't have gitmodules is too much18:04
paulsherwoodpedroalvarez: shall i test it with v8 applied?18:04
pedroalvareztbh I don't know if that would help :/18:05
edcraggpaulsherwood: what the patch?18:06
edcraggpedroalvarez even18:06
pedroalvareznope, that wasn't for you18:06
edcraggok :)18:06
paulsherwoodedcragg: does not apply18:07
*** faybrocklebank has quit IRC18:08
paulsherwooderror: patch failed: strata/build-essential/stage1-gcc.morph:7118:08
paulsherwooderror: strata/build-essential/stage1-gcc.morph: patch does not apply18:08
edcraggpaulsherwood: i just rebased it on master, one sec18:11
edcraggwould gerrit be any more or less useful to you?18:12
paulsherwoodedcragg: thanks! trying now18:19
ssam2 exists now18:20
paulsherwoodcan we make it green, though?18:20
paulsherwoodred is scary18:21
paulsherwoodwhat's the magic for that, btw?18:21
ssam2mkdocs --
ssam2and branch sam/website of spec.git18:21
ssam2i've not looked at how to tweak the theme, I can put it on my list18:22
ssam2although perhaps we should change the wiki at the same time -- seems to me that they should look similar18:22
pedroalvarezthanks! I almost send it first :)18:23
edcraggpaulsherwood: i think that could be gcc architecture problems - assembler complaints at the end18:25
edcraggif i'm not mistaken, i think this could be a 32 bit ubuntu rootfs, trying "readelf -a -W `which gcc` | grep Tag_CPU" for example18:25
*** locallycompact has quit IRC18:26
paulsherwood  Tag_CPU_name: "7-A"18:26
paulsherwood  Tag_CPU_arch: v718:26
paulsherwood  Tag_CPU_arch_profile: Application18:26
paulsherwood  Tag_CPU_unaligned_access: v618:26
paulsherwoodoh well18:27
edcraggat least, v7 definitely isn't 64-bit18:27
edcraggit's probably cross-bootstrap time18:27
* pedroalvarez smiles18:28
paulsherwoodhow do i do git fetch into a bare repo?18:28
richard_mawthe same way as a non-bare repo usually18:28
richard_mawwhich bare repo is this?18:28
paulsherwoodnever mind, i figured it out... i'm in my local gits/*linux18:29
*** ssam2 has quit IRC18:36
paulsherwoodarmv7 build also borks ^^18:53
* paulsherwood goes home18:55
*** rdale has quit IRC19:09
*** edcragg has quit IRC19:40
*** locallycompact has joined #baserock20:36
*** locallycompact has quit IRC22:01
*** gtristan has quit IRC22:23
*** edcragg has joined #baserock22:32

Generated by 2.15.3 by Marius Gedminas - find it at!