IRC logs for #baserock for Tuesday, 2015-11-10

*** gtristan has quit IRC03:27
*** gtristan has joined #baserock03:59
gtristanso, here is the upstream of typical truetype chinese fonts:
gtristanno git, no VCS, no tarball06:03
gtristanHow do we lorry that ?06:03
*** ctbruce has joined #baserock08:26
*** tiagogomes has joined #baserock08:42
*** gtristan has quit IRC08:43
persiaWe probably need a "files" type lorry09:04
*** bashrc_ has joined #baserock09:06
*** nowster has joined #baserock09:17
*** gtristan has joined #baserock09:21
*** toscalix has joined #baserock09:23
persiagtristan: For things like that, we probably need a new lorry type.09:23
gtristanyeah :-/09:23
pedroalvarezI don't know what to do with that :/09:25
persiaPatch Lorry?09:25
gtristanwe need a new lorry type for sure, it seems obvious there's going to be "files" as payload sometimes09:25
persiaAnother significant source of "files" distribution is in the Java ecosystem, where source is often in a jar.09:26
persiaThere's also the RFCs, standards documents, etc.09:26
gtristanright, except that normally a git repo, if you have the source at least, will have .java files and rules to build the jar09:26
persiaAnd many organisations UX teams work on a files basis, which assets would want to be consumed for an HMI09:27
persiagtristan: If the Java folk expose their git, yes, but not everyone does.09:27
gtristannod, there are plenty of cases where you want a file, that is something more than config09:27
gtristanI considered putting the fonts in install-files/... but thats just garbage09:28
gtristansome of those are ~10MB unzipped09:28
gtristanthey are not config data, obviously09:28
* gtristan thinks he overheard something similar for binary blobs used in some bsps09:29
gtristan(that those were also somehow problematic)09:29
pedroalvarezwe created a repo for those09:29
pedroalvarezbsp-support.git IIRC09:30
gtristanpedroalvarez, would you think it would be more logical to support those blobs as "file" lorries ?09:31
pedroalvarezhm... in the raspberry pi case, these blobs were already in a git repo, but a 4G git repo..09:31
gtristanwhy not grab the repo then ?09:32
* gtristan thinks webkit was 6GB09:32
gtristanspeaking of which09:33
* gtristan holds breath and dives into another webkit rebuild09:33
CTtpollardwebkit \o/09:34
pedroalvarezI think I only need less than 40M from those 4G09:34
pedroalvarezI thought it was overkill09:34
gtristanpedroalvarez, you mean a checkout is 4GB ? or the entire git ?09:35
pedroalvarezbut I didn't share that with everyone sending it for review, so people still can disagree with my idea09:35
gtristanon my side, I get the optimization of downloading tarball snapshots from the trove09:35
pedroalvarezgtristan: I'd say the checkout09:35
gtristanthat's indeed annoying09:36
gtristanwell, a patched lorry + chinese fonts wont happen and be deployed by the end of today09:37
*** edcragg has quit IRC09:37
gtristanthe chinese fonts are a nice final touch to the branch I'm about to submit to gerrit09:37
gtristanbut of course, one can live without them, if at least you dont only speak chinese09:37
gtristansoooo... I coooould create an un-lorried repo for them while webkit takes her time assembling herself....09:38
pedroalvarezsorry, when I said checkout I meant the entire git09:39
pedroalvarezmy brain is not here today09:39
pedroalvarezbut when you have to build, you have to clone it entirely, right?09:39
pedroalvarez"to build"09:39
pedroalvarezthey are blobs...09:39
gtristanpedroalvarez, thats what I dont think09:39
gtristanfrom ybd I have access to this tar server thing right ?09:39
pedroalvarezsame for morph09:40
gtristanand it downloads something smaller than the whole history does it not ?09:40
pedroalvarezI'm not sure about that09:40
gtristanso if I'm not playing in that module myself and creating deltas, I dont care about the whole history09:40
gtristanoh, you mean it just tars the whole history and is not smarter than that ?09:40
pedroalvarezbut yes, that only works for repos in g.b.o, and I was building from github09:40
* gtristan thought it was smarter, anyway...09:41
pedroalvarezgtristan: no, I mean that I don't know how that works :)09:41
*** jonathanmaw has joined #baserock09:41
gtristanobviously I have time to chat... *cough* webkit...09:41
*** paulw has left #baserock09:42
gtristanwell, I could take a look at the lorry code and try to assemble a patch, it'll help polish off my non-existent python skills :)09:42
persiaI think the tarserver thing is excessive automation applied to the git server: we shouldn't rely upon it.09:43
gtristanpersia, well... in todays day in age it's much less expensive than only a decade ago, but I would think; anything to reduce the cost of transport is a bonus09:45
*** ssam2 has joined #baserock09:45
*** ChanServ sets mode: +v ssam209:45
gtristanof course, if you are in such a position where you happen to have the trove in the same office, you dont care09:45
gtristanand in korea, I happen to have unlimited high speed internet09:46
persiagtristan: My problem is that it ties tools in ways that I have yet to find a user that wishes.09:46
gtristanI might be hurting if I was in brazil, though :)09:46
persiaI don't have a problem with a generic tarserver: I have a problem with any build tool asserting that it can only have full functionality with a specific git server, when that git server doesn't offer large chunks of functionality that everyone wants, so all the users have to run internal mirrors, which, because Lorry cannot subscribe to event feeds, means everything has built-in multi-hour delays.09:47
gtristanpersia, that looks more like baserock implementation details bleeding out of the box09:48
gtristanI perhaps shouldnt even really have to know that each module is a git, except that 'a git is the input of the trove'09:48
* persia dislikes the very concept of "trove", and has failed to find many people who want such a thing.09:52
persiaI *like* knowing stuff is git.  I think that developers mostly know and use git, so hiding git only causes confusion, suspicion, and unhappiness.09:52
*** edcragg has joined #baserock10:00
gtristanlook at that... lorry supports an undocumented "zip" type !10:11
gtristannow, that importer does not support gzip for some reason, but looks like it could...10:11
pedroalvarezundocumented /o\10:12
gtristanwell, it's not in the README file afaics10:16
persiaFirst step: add it to README.10:16
persiaNext step, look at gzip10:16
gtristannod, lemme try to get it to do gzip at the same time :)10:16
persia(each commit stands on it's own)10:16
gtristanof course10:16
pedroalvarezoh, so you only need to get one of the files of that ftp folder?10:17
pedroalvarezI somehow thought you needed the entire folder10:17
persiaThat would be another useful lorry type :)10:19
gtristanpedroalvarez, that might be cool, but I dont see why lorry would have to be that complex10:19
gtristandocbook-xml.lorry is a "zip" :)10:25
gtristanok first docs10:25
gtristanI'll do a separate importer for gzip, it doesnt fit into the same semantics really10:25
gtristanas zip files are structured and can contain multiple files10:25
persiaAnd tgz is already handled by "tar"?10:30
gtristantar supports all the compressed tarball types yeah10:32
gtristanits documented that it used to have a "compression" option which is now deprecated10:32
gtristanDocumentation patch:
gtristanOk to just push that one I guess ??10:33
*** locallycompact has joined #baserock10:36
*** ssam2 has quit IRC10:37
franredgtristan, importing zip files do not create any branch?10:38
gtristanfranred, nope10:39
franredgtristan, well, docbook-xml.git creates master branch10:39
pedroalvarezI was going to same the same about the tag10:39
gtristanjust ran it on docbook-xml...10:40
gtristanfor context:
gtristanmaster is not something that is created, though; it exists by default in any git10:40
franredgtristan, AFAIK, it is the default branch, but you can have a git repository without master, I think10:41
CTtpollardyou can have a git repo without master10:42
pedroalvarezgtristan: so, I can see a tag here:
pedroalvarezand I see "tagged byLorry Zip Importer <lorry-zip-importer@lorry>"10:42
pedroalvarezgtristan: regarding "Ok to just push that one I guess ??", is that to gerrit?10:43
gtristangit --init && touch pony && git add pony && git commit && git branch.... *master10:43
persiagtristan: Yeah, just push commits to gerrit for doc patches10:43
gtristanpedroalvarez, I dont know where that branch came from, though, however it is indeed there10:44
gtristanpedroalvarez, it wont be there if you run lorry on your machine :)10:44
gtristanpedroalvarez, nor will there be a tag10:44
gtristanoh !10:45
gtristanthere is a tag10:45
pedroalvarezright :)10:45
gtristanpersia, gerrit ? does gerrit grok the lorry repo ?10:50
gtristanoh apparently yeah10:50
persiaIf there are any tools for which Baserock is upstream that are not managed in gerrit, that is considered a bug.10:50
gtristanthere are those extra redundant commit ids in the history10:50
*** ssam2 has joined #baserock10:51
*** ChanServ sets mode: +v ssam210:51
franredgtristan, it looks fine by me10:51
* gtristan submits a gerrit:
franredgtristan, and regarding the discussion about if master is by default created you could replace "branch_name = 'master'" by "branch_name = 'foo'" in and master shouldn't exists10:53
gtristanfranred, eh... thats a hidden feature, dont tell anyone10:55
gtristanfranred, the last thing we need is people to create gits without any master :)10:55
* persia has lots of gits without "master", and likes it that way10:57
franredgtristan, master is just your reference/default branch - in may projects development/production/"any other name someone likes" become "master" ;-)10:57
*** wschaller has joined #baserock10:59
*** franred has quit IRC11:05
persia"trunk" is a common one, for example11:05
tiagogomesfor svn surely, not git11:05
persiaFor git: lots of git repos were once svn repos, and the teams migrating didn't change names.11:06
persia(but extra points for immediately understanding why "trunk" was a popular development focus branch name)11:06
rjekWe only removed the CVSisms from NetSurf's relatively recently11:06
rjek(CVS -> SVN -> git)11:06
richard_mawpersia: well, presumably everything else branches off it, so the metaphor holds11:10
persiarichard_maw: Indeed.11:11
* persia has actually seen new native-git projects use "trunk" as a result of being created by folk comfortable with subversion.11:11
*** locallycompact has quit IRC11:16
*** locallycompact has joined #baserock11:16
tiagogomesanyone knows what is the sparse-gunzip script about in morph?11:29
edcraggi know busybox tar doesn't handle compressing sparse files very well11:32
ssam2last commit is from 2012 and the message refers to a script called 'loopback-rootfs' that doesn't exist any more11:36
ssam2so I imagine it is not for anything11:36
ssam2seems to have been used in which also doesn't exist any more11:37
tiagogomesyes, git told me that :)11:39
persiaYay: orphan code that can be removed!11:39
*** jmacs has quit IRC11:42
*** jmacs has joined #baserock11:46
*** toscalix has quit IRC12:14
*** toscalix_ has joined #baserock12:14
*** mariaderidder has joined #baserock12:20
*** gary_perkins has quit IRC12:40
*** gtristan has quit IRC12:47
*** gtristan has joined #baserock12:47
pedroalvarezew.. I just tried to change the password to a freshly built and deployed baserock system12:52
pedroalvarezand I could only press one key before passwd asked me to retype the password12:52
*** gary_perkins has joined #baserock12:54
* gtristan has a gzip lorry importer that works :)13:14
tiagogomesdid we change pam recently?13:21
* gtristan hopes not13:22
gtristantiagogomes, last I touched pam config data I believe it was completely gnome specific - systemd installs its own thing (which was upgraded recently), and other than that my pam meddling is limited to install-files/gnome and strata/gnome/gdm.morph13:23
gtristaninput method goodness finally built and ready for review:
gtristan :)13:32
* gtristan about to hit the road...13:40
gtristanbut... I have a gzip importer !13:40
gtristanbeware my awful python skillz !13:40
*** gtristan has quit IRC13:42
* paulsherwood wonders what thrift is, and why it suddenly refused to build
ssam2"virtual memory exhausted: Cannot allocate memory"14:00
ssam2not seen that before!14:00
pedroalvarezdoes that mean "not enough ram?"14:02's vbox on my lpatop14:02
*** gtristan has joined #baserock14:02
paulsherwoodrerun worked14:05
radiofreepaulsherwood: could add swap14:05
paulsherwoodtrue :)14:05
pedroalvarezgtristan: boo, it shares a lot of code with the
gtristanpedroalvarez, sure, it's cheap ass copy paste mostly14:09
pedroalvarezI don't feel like I can review it sadly14:10
gtristanbut it turns out not *that* much, it's less than zip-importer14:10
gtristanyou can test it !14:10
pedroalvarezI can yes14:10
gtristanI do not vouch for the excellence of quality of python, but I can vouch for it actually working14:10
gtristanits weak point is, that it assumes the target filename to be <gzipfile> - .extension14:11
gtristanand that it does not extract the encoded mtime, as python ignores the mtime in it's GzipFile14:12
gtristanI'm not entirely sure, but I have a feeling that a gunzipped file.gz is simply named 'file' without the extension too14:13
gtristanas gzip is not a dictionary, I doubt it encodes the file name14:13
* gtristan checks that up14:13
* pedroalvarez realises that gzip files only have one file inside14:14
* pedroalvarez wonders how useful is going to be it, and how many times are going to use it14:15
gtristanpedroalvarez, it will be used more than zip !14:19
gtristanwhich is only used once :)14:19
gtristanarguably, we should simply have a raw file lorry as well14:20
gtristanso that a chunk morph can do whatever it wants with a blob payload14:20
gtristanin anycase, without one, or the other; selecting a chinese keyboard in gnome control center will give you garbage in terminal and entries and editors14:22
gtristanand its so ridiculously simple to fix that, the fonts just need to go into /usr/share/fonts/freetype/chinese during the build14:22
*** toscalix_ has quit IRC14:23
persiaI'm not a fan of cargo-cult code, even if the copy source is the same system.  Better to refactor a bit to abstract the common elements into a single place.14:23
persiaIf you're just learning a new language, maybe ask the usptream developers for help with language features (e.g. for potential patches to Lorry, ask for python help here)14:24
gtristanNot sure I compute that, cargo-cult or such, oh you mean that it's duplicating a couple of lines from the zip ?14:25
gtristanbah, it's really hardly doing that at all14:25
gtristanthere is a convenience function which could be shared14:25
*** toscalix_ has joined #baserock14:25
gtristanwhich is 3 lines long14:26
persia3 lines is fine :)14:34
*** straycat has joined #baserock14:39
*** jjardon has quit IRC14:42
gtristan... anyway ... it's not the prettiest code, I will be happy to make any changes to it for any review comments14:44
gtristanof course not tonight (almost midnight)14:44
*** jjardon has joined #baserock14:46
pedroalvarezI've tested it and it works :)14:48
gtristanit does, honestly the zip converter wasnt particularly pretty either - but I can make the gzip one more readable with some more comments and better formatting etc if that's desirable14:50
*** wschaller has quit IRC15:06
pedroalvarezgit-review is only sending one of the 5 patches..15:12
persiarebase and try again?15:19
pedroalvarezrebase on top of master? I did that15:20
pedroalvarezI'm pondering rewriting every change-id15:20
pedroalvarezbut that shouldn't work15:20
pedroalvarezbut it worked ...15:22
*** paulw has joined #baserock15:22
pedroalvarezRight! This is to add support to install SSL certificates in troves
pedroalvarezthere is another patch for trove.configure, coming soon15:26
*** paulw has quit IRC15:28
*** franred_ has joined #baserock15:31
*** franred_ is now known as franred15:32
pedroalvarezhere trove.configure patch16:11
pedroalvarezforgot to ^V16:12
richard_mawaaa, there's faces everywhere!16:15
richard_maw(this is the first time I've used the new UI)16:16
richard_mawI miss the old one, which provided dependency information16:16
franredI like the new interface :)16:18
*** toscalix_ has quit IRC16:19
radiofreewhat is new about the ui?16:20
persiaradiofree: : specifically hackergotchi support, in-line editing, etc.16:21
pedroalvarezfranred: answered16:21
radiofreein-line editing as in make patch changes using the ui?16:22
radiofreepedroalvarez: ^16:22
SotKI think so, yes16:22
pedroalvarezI think you can only edit existing patches16:22
pedroalvarezall of it, not only the commit msg16:22
pedroalvarezah, not me :)16:22
richard_mawwtf was wrong with the term "Avatar"?16:22
persiarichard_maw: Different cultural sources, mostly.16:23
radiofreeghostery blocks gravatar so i don't seem them anyway16:23
persiapedroalvarez: You can answer better than I anyway.16:24
pedroalvarezfranred: I'm happy to change the patch if there is a better way16:25
franredpedroalvarez, what about
*** ctbruce has quit IRC16:26
pedroalvarezfranred: I don't find that easier to understand tbh16:28
franredwell, it does the same, Im going to change my -1, it is not something that should block this merge16:30
* richard_maw can vouch for it being logically equivalent, but despite the extra indentation, I think the separate check for whether the variable is defined is clearer16:31
*** toscalix_ has joined #baserock16:31
pedroalvarezthat is what I thought16:33
pedroalvarezI think I want to add an extra check, to check that the certificates are valid16:33
pedroalvarez(otherwise lightppd fails to start)16:33
pedroalvarezonly if I have time though..16:36
*** ssam2 has quit IRC16:42
tiagogomesmm, ssam2 just got away. His patch to modify the heuristic for max-jobs broke the tests16:53
persiarevert it!16:54
persiaOh, but we can't, because that would break history, or at least make it ugly.16:54
persiaCan we do pre-merge testing?16:54
persiaHow far away from completion is Mason v3?16:54
*** ssam2 has joined #baserock16:55
*** ChanServ sets mode: +v ssam216:55
tiagogomesone version increase :)16:56
persiatiagogomes: No: there was lots of work done on v3 earlier this year: I think it never got deployed, or tested with more than one worker, but I think it was finished.17:02
SotKit never had meaningful review, we thought of more work to do, and then didn't have time to do it17:04
SotKwell, the latest version didn't have meaningful review afaicr17:05
SotKthere was useful review on previous versions17:05
*** toscalix_ has quit IRC17:05
SotK!/story/8 was what ssam2 and I came up with for it in May17:06
persiaSo, it needs some rebasing, and a bit more development, and then all the deployment work?17:08
SotKpretty much17:08
* persia hopes someone gets excited at the prospect and does that17:09
*** franred has quit IRC17:36
*** toscalix_ has joined #baserock17:49
*** ssam2 has quit IRC17:51
*** jonathanmaw has quit IRC17:58
*** mariaderidder has quit IRC17:58
*** nowster has quit IRC17:59
*** bashrc_ has quit IRC18:02
gtristanjjardon, oops :)18:18
gtristanactually it's lorried18:19
gtristanjust that commit is fucked18:19
toscalix_tiagogomes: what does it needs to happen for that code to be reviewed and deployed ?18:22
*** toscalix_ is now known as toscalix18:22
toscalixtiagogomes: SotK how can I help?18:22
* gtristan is just checking libgnomekbd still builds against the already existing libxklavier18:23
gtristanjjardon, did you catch the link btw: ?18:24
jjardongtristan: cool!18:24
gtristanjust missing a way/lorry to install the chinese fonts, but input methods in general, everything works pretty smooth18:25
* gtristan only wishes he had a distro that actually got it right for once18:26
gtristananyway branch updated18:27
jjardongtristan: maybe we should add and as well?18:28
*** edcragg has quit IRC18:28
SotKtoscalix: with what? mason v3?18:29
gtristanjjardon, I think you +2'd the patches right before I pushed the fix, which caused gerrit some confusion :-/18:29
gtristanjjardon, for ibus-chewing, no... covers bopomofo...18:30
jjardongtristan: ah, ok then18:32
gtristanand I think the other is covered by ibus-anthy18:32
jjardongtristan: any idea why -enable-private-png=yes is required in ?18:33
gtristanjjardon, no idea why, I just followed jhbuild modulesets to add that flag18:33
*** locallycompact has quit IRC18:33
gtristanthe python3 thing is a requirement though18:33
gtristannot sure why there is an option exposed18:34
gtristanas I first ran it with python2, which is available, but then it just craps out trying to import gi18:34
jjardonok, lets merge this then we can investigate the enable-private-png thing18:36
gtristannext on my list; find some way to allow my vm to access the internet18:37
jjardongtristan: thanks for the patches! :)18:37
* gtristan assumes it will be some kind of hell which involves routing tables and providing a dns18:37
* gtristan shudders at the thought18:37
* gtristan is alergic to routing tables :)18:38
gtristananyway, its a necessary evil, in order to ensure geoclue and online accounts actually work18:39
jjardongtristan: or you can dd the image in a pen drive and boot it in real hardware18:39
gtristanI can do that, its an exchange of one pain for another :-/18:40
gtristanas I dont happen to have more than one machine lying around right now ;-)18:40
*** toscalix has quit IRC18:56
persiagtristan: How are you hosting your VM?19:28
persiaThere's a lot of glue in most environments these days that can make that nearly seamless, if you like.19:28
persialibvirt likely being the most accessible to provide transparent pre-configured DNS/NAT/etc. to a VM on a laptop.19:31
SotKvirtualbox makes it easy too19:32
persiaYes it does.19:37
persiaAlthough somewhere in my head is the idea that fewer distros have working virtualbox included.  I don't know why.19:38
persiaThat said, someone should probably pick up the vagrant basebox work at some point, so one can `vagrant up` a baserock VM.19:38
SotKwas that working at some point or not?19:39
persiaAs I understand it, it reached the point where it was possible to define a basebox, but no default basebox was ever generated or hosted.19:40
persiaThe problem being, of course, that one probably wants to have a *specific* basebox to do anything, and at the time, one already needed to have a dedicated baserock VM in order to build Baserock, so would already have working infrastructure meaning that running `vagrant up` was uninteresting.19:41
*** Walkerdine has joined #baserock23:58

Generated by 2.14.0 by Marius Gedminas - find it at!