IRC logs for #baserock for Monday, 2016-01-25

*** gtristan has quit IRC06:52
*** gtristan has joined #baserock07:26
*** toscalix has joined #baserock08:00
*** franred has joined #baserock08:08
*** CTtpollard has joined #baserock08:11
*** ctbruce has joined #baserock08:31
*** gtristan has quit IRC08:56
*** tiagogomes_ has joined #baserock09:03
*** bashrc_ has joined #baserock09:07
*** edcragg has joined #baserock09:09
*** jonathanmaw has joined #baserock09:31
*** rdale has joined #baserock09:40
pedroalvarezthis should fix gnome-session build: https://gerrit.baserock.org/#/c/1762/1/strata/gnome/gnome-session.morph09:59
radiofreeaye10:00
pedroalvarezBroken since 4c18622560ff115db00c6deb592371cbb3eccaad10:00
pedroalvarez(my guess)10:00
radiofree[17:17:53] <radiofree> --disable-docbook-docs might fix it, though it's odd that this has ever worked if that's the case10:00
radiofree[17:18:15] <radiofree> aaah10:00
radiofree[17:18:17] <radiofree> http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/definitions.git/commit/?id=fe64c00f17a4922c7499f9bc48f672cca293fc6510:00
radiofree[17:18:43] <radiofree> previously it probably didn't build the docs, because that dependency wasn't there10:00
pedroalvarezheh looked for a patch but there wasn't any10:01
radiofreemight be worth a quick check to make sure it's not just gnome-session doing it now10:01
pedroalvareztrue10:01
pedroalvarezbut with webkit there, is not a quck check :/10:02
pedroalvarez2 webkits, btw10:02
radiofreelol10:02
radiofreetwice the fun!10:02
pedroalvarezbut you are right, I'll leave this VM building10:02
pedroalvarezit can take several hours ... :/10:03
radiofreeyeah i noticed that10:04
radiofreeis it building MAX_JOBS: 1?10:04
pedroalvareznope, but my VM is not ready for this kind of builds10:04
radiofreemight be possible to split webkit out?10:06
radiofreethough looking at it there's a number of things common to a lot of components that would probably have to be moved out as well10:06
pedroalvarezyeah, that's why all of gnome is on the same stratum10:07
radiofreei'm just thinking that it's probably quicker to rebuild the gnome-core if there's a needed change to webkit, rather than rebuilding webkit because of a change to gnome-core10:07
pedroalvarezyeah, but when you are working in gnome, and something lower in the stack changes.. and you have to wait for webkit....10:08
pedroalvarezI guess that's the view from the people maintaining it10:09
radiofreepedroalvarez: yeah, so the idea would be to make a gtkwebkit strata that didn't depend on the gnome-core stuff10:09
radiofreeby core i mean session, shell, mutter, clutter etc..10:10
radiofreethey could depend on gtkwebkit, however i don't think gtk webkit depends on them?10:10
*** ssam2 has joined #baserock10:10
*** ChanServ sets mode: +v ssam210:10
pedroalvarezthe idea from gtristan was to http://listmaster.pepperfish.net/pipermail/baserock-dev-baserock.org/2015-November/013337.html10:12
pedroalvarezso, stop thinking about how to organise strata, and just worry about the dependencies for each component10:16
CTtpollardre the email to baserock dev that I got mentioned in, I don't know about the status of genivi systems in baserock10:21
ssam2i'm asking more about the state of them in GENIVI10:24
*** paulw has joined #baserock10:48
rdaleif you build a stratum such as build-essential with morph, does the artifact that gets cached consist of a filesystem containing all the chunks that make up stratum?11:28
rdalestratum^the stratum11:29
SotKIIRC morph just creates a file containing the names of the chunks that make up the stratum11:29
ssam2yes11:30
ssam2(to sotk's answer)11:30
rdaleah, and so you expect ybd to do the same?11:30
rdalei was assuming that the artifact would actually consist of all the files, not just a list of them11:30
ssam2there's no reason ybd has to do the same11:31
ssam2but it seems like morph's approach is the most efficient11:31
richard_mawstratum artifacts are redundant information ideally, since all that information is parsed from the definitions anyway11:31
rdalei assumed it would be better to untar a single stratum into a build area, rather than untarring the individual chunks11:32
richard_mawrdale: we used to do that, we moved to the current model since it is redundantly storing the artifact contents in the cache, since all the chunk artifacts have the files11:33
ssam2the main overhead is the actual untarring, it doesn't matter whether it's one tar file or 50011:33
ssam2i assume11:33
rdaleah i see11:33
ssam2morph (and ybd?) actually cache the extracted chunks and hardlink the files in place in the staging area, anyway11:33
ssam2that's pretty much essential if you don't want to spend 5 minutes at the start of each build unpacking 1GB of data11:34
richard_mawssam2: even for non-trivial builds, since you then have to tar up strata too11:34
rdaleyes, except when it assembles the final system11:34
richard_mawrdale: aye, at which point you can either extract them from the tarballs, or do file copies,11:34
richard_mawrdale: potentially there's room to optimise the file copies with CoW on filesystems that support it11:35
richard_mawrdale: or you can try an overlayfs on top of the hardlinks for less filesystem coupling, but requiring root11:36
rdalei want to get artifact splitting working in ybd, before i even think about optimization11:36
richard_mawack, easiest is to re-use the existing extraction logic presumably11:37
rdalei build a list of files for a split artifact, but currently each chunk is expected to be in the stratum, so i need to change the file list to point to the individual chunks '.unpacked' directory11:39
rdaledirectories rather11:39
rdaleanother question: is 'this' a special thing in python? is it possible that it would suddenly change while a function is being executed11:39
richard_mawrdale: AIUI 'this' is just Paul's preferred variable name11:40
richard_mawit has no semantic meaning in python11:40
rdaleit's very confusing11:40
rdaleah ok11:40
*** bashrc_ has quit IRC12:12
*** bashrc_ has joined #baserock12:14
radiofreedoes anyone have a build log for xserver from a gnome system?12:25
radiofreeoutput of ./configure is what i'm after12:25
ssam2try `morph show-build-log`12:27
ssam2it might work ... :)12:27
pedroalvarezthat worked: http://paste.baserock.org/ocodabotis12:30
pedroalvarezradiofree: ^12:30
radiofreeTA!12:32
* radiofree hmms12:33
radiofreebtw i manually compiled x on a gnome system for a jetson, gnome-shell works, and it's fast!12:33
radiofree(have to manually start things though)12:34
radiofreenone of the apps appear to work thogh12:35
radiofreehttp://git.baserock.org/cgi-bin/cgit.cgi/delta/xserver.git/log/?h=baserock/1.18%2bjetson12:37
radiofreerdale: interestingly in qt 5.6 doc building fails12:47
radiofreedon't think qtbase is providing qtdoc anymore (or might need to be turned on in the configuration)12:47
rdalei would have thought it should be built by default12:58
radiofreeit used to be, but 5.6-beta is failing everywhere on making the documentation13:02
* radiofree skips it for now since he just wants to test wayland13:02
radiofreealso need to lorry https://code.qt.io/cgit/qt/qtxmlpatterns-testsuites.git/13:03
rdaleit looks like qtdoc has been moved to the qttools repo: https://bugreports.qt.io/browse/QTBUG-4923313:05
radiofreeso you need to compile qtbase to compile qttools to compile the qtbase docs?13:06
rdalehmm13:07
* radiofree adds https://code.qt.io/cgit/qt/qtdeclarative-testsuites.git/ to the list of lorries13:07
radiofreegetting as bad as boost this13:07
rdalewe would probably need to have some kind of bootstrap scheme and build qtbase and qttools twice13:11
*** toscalix has quit IRC13:12
radiofreeqtbase-docs?13:12
radiofreeyou can probably just call "make html_docs" in the qtbase rebo13:12
radiofrees/rebo/repo13:12
*** Lachlan1975 has joined #baserock13:51
*** toscalix has joined #baserock14:14
*** tiagogomes_ has quit IRC16:04
*** edcragg has quit IRC16:06
*** edcragg has joined #baserock16:08
*** tiagogomes_ has joined #baserock16:19
paulsherwoodrdale: i used 'this' to mean the definition we are currently working on (assembling/building etc..).... what would you recommend as an alternative name16:27
paulsherwood(i borrowed this approach from hobo, fwiw)16:28
rdalei would normally take it to mean the current instance as in c++ inside a method (or self in other languages), but if it isn't in a method then 'definition' maybe if it is always a definition16:30
rdalei don't think that would particualarly help actually16:31
rdalefor sandbox.install() possibly 'component' and 'dependency' would be clearer than 'this' and 'component'16:33
paulsherwoodyup16:33
rdalethe other part of the discussion where i found out that build-essential should contain any chunks, only meta files describing the contents of chunks is quite important - is that how it is supposed to work?16:35
rdale^should contain^should not contain16:35
paulsherwoodrdale: as of master, strata only contain metafiles. i would prefer to keep it that way, but my wip for the #150 problem has the files being tarred up...16:37
paulsherwoodwhich as others have said is slower and a waste of space16:37
rdaleright - i don't have any opinion - it's just that i was assuming that it would contain the chunks. i've change my artifact splitting code to copy the chunk files from the individual chunk artifacts16:38
paulsherwoodi was originally trying to treat components (chunks) and compound components  (strata, systems) generically but that is proving difficult :)16:39
rdaleindeed!16:39
paulsherwoodrdale: that would be preferable :)16:39
paulsherwoodrdale: any improvements you can make to the clarity of the code would be welcome :)16:40
rdalewell currently i'm confused, and usually coding confused ideas tends to not be very clear16:40
paulsherwoodack16:40
rdalei was getting strange behviour in sandbox.install() and your comment about exceptions getting eaten in your mail about the system artifact dependencies might explain that16:42
rdalei would like to turn off any multiple instance code, and also turn off the calls to randomize so that everything is deterministic and behaves the same way in every test run16:43
paulsherwoodrdale: multiple instance only happens if instance: set to > 1, so without that exceptions should show up16:45
paulsherwoodi'd be happy with a randomize: True flag, or similar16:46
rdalehere is my current sandbox.install() function: http://paste.baserock.org/ibuhucugoc16:52
rdaleif i don't have the two booleans 'is_system' and 'is_component_stratum', the code behaves differently - why would that be16:53
*** edcragg has quit IRC16:56
*** edcragg has joined #baserock16:56
ssam2i'm going to do a few tweaks to cgit on git.baserock.org, it may be down briefly over the next 30 mins17:02
pedroalvarezthat's ok17:07
pedroalvarezdon't worry if takes ages to load, that won't be your fault17:10
*** ctbruce has quit IRC17:10
paulsherwoodrdale: i don't understand your question17:18
paulsherwoodrdale: /win 4617:19
paulsherwood:(17:19
rdaleif i don't use the 'if is_system:' boolean and just use if this.get('kind', 'chunk') == 'system': as per the original code the test is never true, even though it was true a couple of lines earlier in the function17:20
paulsherwoodwow.. that is weird17:22
rdalei've tried to use pdb, but it seems broken inside a baserock vm as i assume we use a dummy readline() and it doesn't echo what you are typing17:23
paulsherwoodrdale: maybe try this on your host? ybd doesn't require a baserock vm17:24
rdalei don't want to run anything as root where i'm debugging copying files around17:25
paulsherwoodrdale: use vagrant up, debug on the vm?17:25
paulsherwood(if you have virtualbox)17:25
rdaleyes, i have but a baserock vm with kvm works fine apart for this pdb problem17:26
paulsherwoodok17:26
paulsherwoodsorry.. boarding now...17:26
rdaleok17:27
*** jonathanmaw has quit IRC17:40
*** toscalix has quit IRC17:49
*** toscalix has joined #baserock17:49
*** bashrc_ has quit IRC18:02
ssam2finally the cgi-bin/cgit.cgi is gone from git.baserock.org URLs!18:04
paulsherwoodw00t!18:05
ssam2old URLs should still work of course18:05
ssam2well.. maybe18:08
ssam2pedro points out this will break all other troves18:08
ssam2ok, seems it's much better to have them at http://git.baserock.org/cgit/18:13
ssam2we can still redirect git.baserock.org/baserock/ to git.baserock.org/cgit/baserock etc. as well18:13
ssam2but redirecting git.baserock.org/* would break stuff like lc-status.html18:13
pedroalvarezit's nice as it is right now18:14
*** ssam2 has quit IRC18:31
*** edcragg has quit IRC18:34
*** rdale has quit IRC18:38
*** toscalix has quit IRC18:52
*** Lachlan1975 has quit IRC19:11

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!