IRC logs for #baserock for Tuesday, 2015-03-24

jjardonSotK: Ive just sent pacthes to solve the simple-network problem in the minimal systems, sorry for any inconvenience00:12
*** rdale_ has joined #baserock02:30
*** rdale has quit IRC02:33
*** zoli__ has joined #baserock04:16
*** petefoth_ has joined #baserock04:22
*** petefoth has quit IRC04:24
*** petefoth_ is now known as petefoth04:24
*** zoli__ has quit IRC05:35
*** zoli__ has joined #baserock06:13
*** zoli__ has quit IRC07:21
*** zoli__ has joined #baserock07:28
*** sherm_ has joined #baserock07:33
*** zoli__ has quit IRC07:46
*** fay_ has quit IRC07:49
*** fay has joined #baserock07:49
*** fay is now known as Guest3600507:50
*** zoli__ has joined #baserock07:52
straycatpedroalvarez, gertty gives you a lot more log info if you run it with -d08:14
straycat(debug) >.>08:14
*** Guest36005 is now known as fay_08:16
SotKjjardon: thanks, I'll look at them in a bit08:17
SotKthere was no inconvenience really, I just thought I'd broken the deploy code for a bit :)08:18
*** gfinney_ has joined #baserock08:22
*** gfinney_ has quit IRC08:24
*** gfinney has joined #baserock08:24
*** jonathanmaw has joined #baserock08:33
*** mdizzle has joined #baserock08:50
*** bashrc_ has joined #baserock09:03
straycatWhat is the stance on gerrit, is this a trial run? Will there be some sort of vote?09:09
* Kinnison imagines that ssam2, richard_maw, and pedroalvarez are best placed to discuss that09:13
franredstraycat, I though it was decided we are going to move to gerrit and abandon the ml09:18 soon as possible and for the future patches09:19
*** tiagogomes_ has joined #baserock09:21
straycatfranred, I don't know, it's not clear to me what happening, I assumed there'd at least be some sort of trial run where we get to try it and decide whether we want to use it09:22
*** gary_perkins has joined #baserock09:24
*** fay_ has quit IRC09:33
*** fay has joined #baserock09:33
*** fay is now known as Guest9189009:34
jjardonWhat's the license of baserock definitions?09:37
rjekGood question.09:38
bashrc_maybe there should be a license field09:38
* Kinnison thinks the definitions should be under the ISC licence09:38
Kinnisonit's possible some of the scripts etc might be under something else09:39
Kinnisonbut ISC is the only sane option for making it easy to derive and modify for those who might not wish to release their system definitions09:39
bashrc_better to make it unambiguous09:39
*** Guest91890 is now known as fay_09:41
* jjardon searchs for isc09:44
KinnisonThink "BSD" but without the pesky owned-by-berkeley bit09:45
*** edcragg has joined #baserock09:52
rjek"Don't pretend you made this, but don't blame me either.  Otherwise, do whatever you like."09:54
*** ssam2 has joined #baserock10:04
*** ChanServ sets mode: +v ssam210:04
tlsa^^ that's the output of the reproducibility certification tool10:04
tlsashowing the three types of problems it can catch10:05
tlsathe patches for `morph certify` are on gerrit10:07
* SotK will look at them in a while10:09
tlsathanks :)10:09
*** jmacs has joined #baserock10:10
*** a1exhughe5 has joined #baserock10:20
franred is failing because of  - if someone does have spare time to review it, would be nice to merge it10:23
richard_maweww, I'd forgotten how nasty lsof's build commands were10:26
ssam2franred: looks fine, merged10:26
pedroalvarezdoes that mean that now it can't build with busybox tar?10:27
* pedroalvarez checks10:27
franredssam2, thanks10:27
ssam2pedroalvarez: probably does mean that10:27
ssam2i'm not sure what to do about it10:28
franredother issue which is worth to investigate is if we are shipping licensed code using lsof - fedora stripes part of the tarball, see
ssam2to be honest, if a system based on the minimal system requires GNU 'tar' to build, but uses Busybox 'tar' (or no 'tar' at all) at runtime, it's not a huge issue I think10:28
straycat09:09  straycat$ What is the stance on gerrit, is this a trial run? Will there be some sort of vote?10:28
straycat09:13 ! Kinnison imagines that ssam2, richard_maw, and pedroalvarez are best placed to discuss that10:28
straycat09:18  franred$ straycat, I though it was decided we are going to move to gerrit and abandon the ml10:29
straycat09:19  franred$ soon as possible and for the future patches10:29
straycatssam2, ^10:29
ssam2straycat: at the meetup, we discussed a plan for setting up Gerrit, but didn't discuss any formal process for deciding when it's 'done' or whether we should give up on the effort10:30
ssam2do you want there to be a vote?10:30
ssam2if so, do you have an idea of what the criteria should be for someone to be eligible to vote?10:30
ssam2for the record, i'm fine with putting it to a vote once it's complete (once it sends emails, and perhaps is updated to 2.10, i think it'll be complete)10:36
straycatssam2, My personal stance on this, is that while I was initially very excited by the benefits brought by gerrit the command line tooling I've tried (gertty, gerrymander) seems neither featureful nor mature, I'm also *very* worried by the fact that gerrit has no native concept of a patch series, so a vote seems necessary. I have no idea what the criteria should be for someone to be eligible to vote.10:36
ssam2I think it'd be useful to investigate how other projects that use Gerrit handle the fact that submitting a long series of related changes is awkward to do10:38
ssam2a vote is probably a good idea, my view is certainly biased because i spent the last month working to set up Gerrit10:38
straycatheh :)10:39
bashrc_straycat: +1 on the command line tooling10:45
jjardonFor a vote you need an alternative (even i like mailing list, its not an option because one of the whole points of use gerrit is that it will check the patches build and give us a +1 or +2 automatically)10:45
bashrc_but there is an API, so the commandline tooling issues can be solved10:45
tlsawhen it finally pushed, I think my patch series was handled OK10:45
Kinnisontlsa: is there a URL you can give me so I can review your series?10:45
* Kinnison would like to get a feeling of how reviewing a multi-patch series goes10:46
tlsaI can vive you a link to the first patch, and you can click to the next10:46
Kinnisonif that's the way it works, then sure10:46
* Kinnison already knows how review feels in mutt, so wants to get the proper gerrit feel10:46
tlsasee "related changes" on the right10:47
tlsatop is the last patch of the series10:47
tlsathe gerrit web UI is a bit poor, but maybe it improves with a later version10:47
* Kinnison is confused10:50
Kinnisonlast time I signed into gerrit, it used my openid10:50
Kinnisonnow it's sending me to openid.baserock.org10:50
straycatI like that I can easily get a list of what needs to be reviewd, but there are other systems such as patchwork that can do but still allow to use an ml. I find the sets of changes in a topic difficult to follow10:50
KinnisonHow do I sign into my existing account?10:51
straycat*can do this10:51
pedroalvarezKinnison: when was this last time?10:51
pedroalvarezI believe that now only our openid is allowed10:51
Kinnisonpedroalvarez: Gosh, quite a while ago when sam was first testing gerrit10:51
pedroalvarezmaybe when I was testing gerrit? (
KinnisonPerhaps it may even have been that long ago10:52
pedroalvarezin that case your account may not exist in this gerrit instance10:52
* Kinnison creates yet another identity then10:53
Kinnisongerrit doesn't like my ssh key :(10:55
* Kinnison skips for now10:55
Kinnisontlsa: I've added three comments to the first patch, can you see them? (They say 'draft' on my screen and I don't know why)10:59
tlsaI can't10:59
* Kinnison wonders how to make it so you can10:59
pedroalvarezKinnison: i think you have to add a vote to send the comments11:00
* straycat suggests that the set of people who have ever submitted a patch to baserock-dev OR submitted a review to baserock-dev ought to be eligible to vote11:00
* pedroalvarez suggests that only the subset that have tried gerrit11:01
straycatThat's unfair, not everyone has time to try gerrit11:01
ssam2how can you vote on something that you know nothing about?11:02
KinnisonI think I'm with pedro here -- if you've not even tried gerrit how can you have an opinion on it?11:02
rjekIt's possible to have opinions on vices without having partaken.11:02
KinnisonOpinions, yes.  an explicit vote on whether to use it or not, IMO not so much11:03
Kinnisontlsa: Right, I had to hit 'reply' (which wasn't obvious in the least) to publish my comments11:03
Kinnisontlsa: can you see them now?11:03
pedroalvarezstraycat: some people that have only used the mail list may be willing to move to something else and without trying it before11:04
pedroalvarezand viceversa11:04
straycatThat's a fair point11:04
pedroalvarezI think is fair that they try it before they can choose one11:04
straycatIt's difficult, it took me several days to decide how I feel about gerrit, several contributors don't have that much time to spare11:06
tlsahow do you send a v2 with gerrit?  Just push it all again?11:06
ssam2push it again with the same Change-Id11:07
ssam2that's the main point of the Change Id11:07
tlsaso if I sqaush into it, it should retain the same change ID?11:08
pedroalvarezthe change ID is part of the commit message, so if you can keep it when sqashing, then yes11:10
* richard_maw thought he saw something about allowing gerrit patches to have dependencies11:12
straycatyou can send changes with the same topic11:12
richard_mawTopics don't imply any dependency information11:12
* Kinnison finishes reviewing tlsa's series11:12
straycatno, you can have a change depend on another, i believe11:13
KinnisonThat took roughly 20 minutes11:13
Kinnisonsubtract maybe 5 minutes for being confused about workflow, and call it 15m to review that 4 patch series11:13
KinnisonThat's not awful but also not great11:13
pedroalvarezrichard_maw: `git review -d 1234` being 1234 the number of the change that you want as dependecy11:14
pedroalvareztaken from here:
Kinnisonssam2: Is it possible to extend gerrit's review mechanism?11:15
Kinnisonssam2: I've given reviews, but because my responses were always -0 or +0, gerrit doesn't count those as reviews11:15
* richard_maw wonders if git-review or one of its friends supports sending a patch series to give it both a common topic, and fill in all the dependencies to make it a chain11:16
ssam2i don't really understand what you mean11:16
ssam2(to Kinnison)11:16
ssam2for I don't see why you didn't just give a -111:17
ssam2in Gerrit, I think -1 doesn't block merging11:17
ssam2-2 blocks merging11:17
KinnisonI don't feel strongly enough to -1 it11:18
Kinnisonor in the +ve case, to +1 it11:18
tlsaKinnison: where did you find the 'reply' button?11:18
Kinnison(to me -1 means 'I veto this')11:18
Kinnisontlsa: near the top of the change page11:18
ssam2Kinnison: ok. so I don't see why Gerrit should pay attention to taht11:18
ssam2to me, it's not a very useful review11:19
ssam2i mean, the comments are useful, but if I see a patch with '-0', it tells me nothing11:19
tlsaKinnison: replied to a comment11:19
Kinnisonssam2: to me, it means someone is uncomfortable with the patch, but doesn't believe that discomfort warrants veto11:19
ssam2in Gerrit, that's what -1 means11:20
ssam2someone in the Mergers group can still +2 and merge a patch which has 10000 -1s11:20
ssam2but they cannot if it has a -2, so -2 is 'veto'11:21
pedroalvarezssam2: but then we should consider if that person belongs to the mergers group :P11:21
ssam2pedroalvarez: Gerrit only allows you to give -2 or +2 if you are in the Mergers group11:21
*** Krin has joined #baserock11:21
* SotK reads Gerrit reviews as -2 == old -1, -1 == old -0, 0 == old +011:22
KinnisonOkay, so as a non-merger I cannot raise a veto11:22
* Kinnison can appreciate that11:22
* Kinnison adjusts his mental model as per sotk's comment11:22
Kinnisontlsa: Since you don't get email notification yet, I have responded to your comment11:23
Kinnisonssam2: will 2.10 fix the lack of email notifications from gerrit?11:23
tlsaKinnison: yeah, thanks11:23
* straycat thought lack of email notification was due to lack of exim or similar in baserock11:23
Kinnisonstraycat: oh right11:24
Kinnisonexim is a pain to compile :(11:24
tlsaI think gerrit is not ideal, but not impossible to work with.  (assuming yesterday's push trouble was a one-off freak.)11:24
KinnisonSo far, it has required an adjustment in how I understand voting to work (not an issue) and has forced use of a browser (not ideal, but I understand 2.11 will fix that somewhat)11:24
KinnisonI've not tried submitting to it11:25
* SotK had push trouble the first time he tried, but since then its been smooth11:25
KinnisonBut I know I'll dislike that11:25
pedroalvarezKinnison: I don't remember how and where, but it was announced that people in the baserock-writers group will be welcome in the Mergers group. But we couldn't do that for the people that didn't have an account at the time.11:25
pedroalvarezKinnison: if you want to be part of that group, you are welcome to11:25
* SotK also had to adjust his mental model of voting to the above11:25
Kinnisonpedroalvarez: I understand (I also don't think I know enough about current definitions or the current morph/lorry/etc codebases to want to be in the group right now :)11:25
pedroalvarezbeing in the group doesn't mean  that you have to make that kind of decissions though11:26
* straycat isn't convinced that everyone in baserock-writers should be in mergers11:27
straycatbaserock-writers corresponds to being able to push to a repo in the baserock prefix on gbo, that's doesn't map exactly to mergers, even though it happens to grant merge powers on gbo11:28
pedroalvarezi thought that people in the baserock-writers group could do merges in master as well11:28
pedroalvarez(as you said, sorry)11:29
pedroalvarezyou are right, not everyone should be in that group11:29
pedroalvarezsome people were only in baserock-writers group to be able to push branches to g.b.o. And that with gerrit is not needed anymore11:30
*** petefoth_ has joined #baserock11:30
SotKgah, ./check --full fails in master again11:31
*** petefoth has quit IRC11:32
*** petefoth_ is now known as petefoth11:32
SotKthe copyright line in COPYING causes "ValueError: invalid literal for int() with base 10: ''" somehow11:34
tlsa"the comment on code --> draft --> return to main patch page --> reply --> post" saga is a bit laborious11:36
SotKhm, maybe its not that line11:36
SotKaha, there's a template line in there11:38
SotKI guess that the copyright checking script should ignore COPYING11:39
*** radiofree has quit IRC11:50
*** radiofree has joined #baserock11:50
*** radiofree has quit IRC11:51
*** radiofree has joined #baserock11:51
franred<franred> other issue which is worth to investigate is if we are shipping licensed code using lsof - fedora stripes part of the tarball, see
franredis something we have to worry about? ^^11:53
*** petefoth has quit IRC11:55
ssam2i've not got a clue11:56
ssam2I think lawyers would generally us a polite cease and desist notice before filing an actual court case against us for violating whatever license it is11:57
tlsaI've rebased, squashing a change into the first commit of my series.  Now the commit has two Change-Ids.12:01
tlsaIs that correct?12:01
* SotK has no idea12:02
straycat ! [remote rejected] HEAD -> refs/for/master (no new changes)12:02
straycaterror: failed to push some refs to 'ssh://'12:02
ssam2tlsa: nope, pick one change-id12:02
straycatAm I seeing this because the object already exists in another branch?12:02
tlsastraycat: I had that yesterday12:02
tlsado you have the git hook for adding change ids?12:02
straycatNo, didn't know I needed one12:03
tlsa`scp -p -P 29418 <Username> .git/hooks/`12:04
straycatwhy do i need that the gerrit guide i was reading said nothing about this?12:05
pedroalvarezI think that the change Id requirement can be disabled12:06
straycatthat sounds good, it's not exactly convenient to have to copy some git hook in everytime i want to submit for review12:07
SotKits very annoying, I agree12:08
pedroalvarezcan be annoying but before to reply resend a patch you had to get the message-id of the previous patch12:09
* SotK sends a patch to fix ./check12:16
ratmice__ssam2: the lsof problem seems to be the dialects/ directory contains code copyrighted by sco, (and apple, and others), though it likely isn't even compiled into the resulting executable12:18
ratmice__ssam2: so seems like they are covering the case of source distribution12:19
ssam2ratmice__: ah, thanks for the explanation12:19
ssam2yes, fedora produce SRPMS so need to worry about that12:19
* bashrc_ thought that sco were defunct12:19
tlsaI've experimentally pushed a v2 of my patch series to gerrit, and it worked fin12:21
straycatpedroalvarez, basically means i'll have some script that git clones then puts the hook in place for me i guess12:21
straycattlsa, i still get the same error12:26
tlsahave you rebased the patch series to add change-ids to the commit msg?12:26
straycatyeah, i needed to amend though >.>12:28
SotKstill the same error?12:28
straycatno, now it's bitching that the emails don't match12:29
* straycat sighs12:29
*** fay_ has quit IRC12:29
straycat huzzah12:31
* straycat does not think the benefits of this warrant the obviously needless complexity12:32
Zarahow do the methods compare if they're given to a new user? do we have a guinea pig new to both ML patching and gerrit? I found using git to send patches to the ML extremely fiddly at first (I'm thinking that people here might be so used to the ML that it seems simpler because it's familiar, rather than because it *is* simpler-- though that's speculation)12:42
ratmice__ssam2: fwiw gleaned from $ tar Oxf lsof/lsof_4.83_src.tar | grep -i Copyright | egrep -vi '(Purdue|static)' on baserocks lsof.git12:45
*** fay_ has joined #baserock12:45
ratmice__with the troublesome files all in the dialects/ subdir, they just nuke all the directories there except linux/12:46
jjardonZara: you have a good point there. People in general is biased for what they are used to use. That's why new comers feedback is important12:49
persiaThe main reason I've been a vocal gerrit advocate was the simplicity of submitting my first patches to a project using gerrit, as opposed to every other project to which I've contributed.12:51
Kinnisonpersia: It's certainly very good for that12:52
* Kinnison is finding that it's also hard to determine if review comments have been addressed12:52
KinnisonThough there's supposed to be ways to improve that12:53
* Kinnison wonders how one gives an opinion on a patch series as a whole rather than on each individual patch in the series12:53
KinnisonAgain, I assume it'll have to be "process" (put it on the tip, or put it on the base fr.ex)12:53
Kinnisontlsa: Your v2 patch series doesn't address at least one of my comments12:55
Kinnisontlsa: Could you go back and decide on responses?12:56
* SotK put it on the tip when he reviewed Javier's patch earlier12:59
jjardonSotK: thanks for the quick review btw :)13:01
KinnisonMy concern there is that everything-but-the-tip could be accidentally merged13:01
* Kinnison is wondering if he should -1 every patch even if individual patches are "okay"13:02
straycatZara, try gerrit and compare :)13:02
* Kinnison is quite happy to admit that he finds the ML more comfortable because it's the way he's worked for F/LOSS contribution for nearly a decade and a half13:02
KinnisonBut Gerrit is "usable"13:02
straycatpersia, How exactly is it simpler?13:02
straycatLet's ignore the fact that I've lost the freedom of choice mail gave me over tooling, and the fact that this forces all review to be centralised on a single service13:03
straycatAnd the fact that tooling gerrit *does* have to support offline work is seemingly incomplete13:04
* Kinnison wonders if someone wrote a tool to join gerrit to a ML13:04
* Kinnison has a quick google13:04
jjardonas I said before, I think the whole point of using gerrit is that a machine will tell us if a patch is suitable to consider to merge or not. AFAIK you cannot do that with ML13:05
straycatSo run the tests? Why is that hard?13:05
Kinnisonjjardon: there's no reason that wouldn't be integratable with a ML using something like patchwork, but yes, tooling already exist for gerrit13:05
Zarastraycat: I don't have any patches to send, though I could try with a dummy patch full of nonsense13:06
KinnisonI've not found a showstopper yet in my interactions with gerrit13:06
straycatZara, okay with me :)13:06
pedroalvarezZara: I've done the same when trying gerrit :)13:06
jjardonKinnison: Id suggest 100% patchwork, but doesnt meet the requisite I said before13:07
SotKjjardon: no problem :)13:07
Kinnisonjjardon: aye13:07
jjardonstraycat: I do not really fancy spending hours building stuff (imageine a patch to upgrade glibc)13:08
straycatjjardon, for baserock/baserock/definitions I'm willing to concede that it's useful13:08
straycatI'm just saddened at the level of complexity this introduces to the review process13:09
jmacsIt's perfectly possible to do automatic builds and merges based on a mailing list; it just needs some software to extract patches from it13:09
jmacsI think Kinnison was looking for something to do this yesterday13:09
* Kinnison has looked and decided that over-all there's nothing nice13:10
KinnisonBut then again, I'm a very particular person :)13:11
* SotK wonders how difficult it would be to write something useable as a trigger for Zuul which watches a mailing list13:12
straycatSotK, or gets the output from patchwork?13:12
ssam2I found submitting patches to gerrit a pain at first, but it's undeniably much easier to merge patches in Gerrit than it is to merge patches from the mailing list13:17
ssam2and now we have git-review in baserock, submitting patches to morph.git or definitions.git or infrastructure.git for me is literally a case of typing `git-review`13:17
* straycat doesn't find the current process for merging series difficult in the slightest13:17
* richard_maw finds keeping up with patches to review difficult though13:17
straycatrichard_maw, patchwork would solve that13:18
ssam2I don't think anything solves that, except actually having people spending time regularly clearing the review backlog13:18
ratmice__jjardon: not only the spending hours, but may not have access to all the types of systems baserock is intended to run on13:18
ssam2only richard_maw seems to really spend time doing that right now13:18
richard_mawstraycat: if it does, that's grand, I'm just saying that the current approach is difficult for me to keep up with, rather than saying Gerrit will fix it13:18
straycatSotK, pathwork reads the ml and constructs a list of stuff that needs to be reviewed from it13:19
Zarahm, following instructions, hit difficulty at 'developing your change'-- first bit said to use public key from host machine, but then it says that you develop change in devel vm (do you always have to?), sshing into gerrit. I'm guessing I also need to add the vm's public key to gerrit.13:19
straycaterr, ssam2 rather ^13:19
KinnisonZara: You use *your* public key and use ssh agent forwarding typically13:19
KinnisonZara: There's no need for your VMs to have their own keys13:20
* SotK doesn't have a key in his VM, but does `ssh -A root@sotk-devel`13:20
KinnisonZara: ^^^ exactly that13:20
* Kinnison is currently suffering because gerrit says my ssh key is invalid :(13:20
richard_mawSotK: I do `while ! vmip="$(vm-ip br-dev qemu:///system)" || ! homely-ssh root@"$vmip" -A; do sleep 1; done`13:20
KinnisonAah apparently this is fixed in 2.1113:21
ZaraKinnison (and SotK): Okay, I don't know how I'd translate the instructions currently on the wiki to that.13:21
richard_mawSotK: but that's because I prefer not to have to retry logging in to my VM myself, and I CBA remembering the VM's IP address, and I don't run a system that has systemd-machined or libnss_mymachines13:21
* SotK looks at the current instructions13:22
KinnisonHmm 2.11-rc1 was out a week or so ago13:22
Zaraah, nvm13:23
Zarafigured it out, it was the way I was sshing into the vm to begin with13:23
ZaraI was dooing 'ssh whatever' rather than 'ssh -A whatever'13:23
Zaraall is well in the world.13:24
Kinnison implies that work is still going on apace on the 2.11 release13:25
ZaraI vaguely remember people talking about the 'BAD COPYING' thing, should I do anything to make it go away?13:29
Kinnisonmerge master of morph?13:30
KinnisonI think SotK's fix got merged13:30
SotKyep, just now13:30
Zaraah, okay, this is just for a dummy test so I'll leave it for now13:30
SotKthanks for the quick reviews Kinnison and ssam213:30
*** petefoth has joined #baserock13:32
ZaraI got 'missing Change-ID in commit message footer'13:35
SotKyou need a change ID in your commits, there is a git hook which adds them for you13:36
SotK`scp -p -P 29418 <Username> .git/hooks/`13:37
Zaraah, I thought I did that, but the wiki instructions are a little different, so maybe it's that13:39
Zara(wiki says /src/morph/.git/hooks)13:39
franredif you go to Projects -> Geneal -> click your project -> clone with commit hook - all of us would avoid confusion (for newbies)13:39
franredfor example: git clone && scp -p -P 29418 morph/.git/hooks/13:39
SotKZara: the command I posted is intended to be run in your git checkout13:39
SotKif you made the commit before you got the hook you'll need to rebase and/or `git commit --amend`13:40
SotKfranred: +1000000000013:41
Zarahm, the wiki seems to have things in the order 'clone, get hooks, create branch and checkout, then commit'13:43
*** a1exhughe5 has quit IRC13:43
SotKthat seems right to me13:43
ZaraI followed it in order. Weird. I'll try again in case it's me.13:44
Kinnisongetting the wiki to be *clear* and *obviously correct* will help here13:44
ZaraI think we should add fran's suggestion to the wiki13:45
Zara(that is, get the command from projects -> General)13:46
tlsaKinnison: yep, my v2 was more an exercise in seeing what would happen when I pushed it.  I only did anything with the first 2 patches of the series13:47
Zaraahh, the wiki instructions mustn't point to master of morph, since this time it's acting differently (copyright didn't hold it up)13:51
Zaraoh, unless the fix *just* got merged13:51
* Zara toddles off13:51
Kinnisontlsa: I'll let you submit a v3 before I consdier again then13:53
Kinnisonssam2: Did you decide if you could grant the stream events permission to everyone?13:55
Zarahm, now I'm getting this at the 'push' step: fatal: Authentication failed for ''13:55
*** a1exhughe5 has joined #baserock13:56
SotKah, franred's command was an anonymous http clone, you'll need to set your remote to be the ssh url for the repo to be able to push13:56
SotKafaik anyway13:56
franredSotK, yes, it was13:57
persiastraycat: As a first benefit, it doesn't require me to sort my MTA.  While choice in tooling is nice, getting mail to work from most of the places I go is signficantly less so.13:57
franredbut you can clone using ssh if you press the option13:57
ssam2Kinnison: didn't investigate yet13:57
KinnisonFair enough13:57
Zaraokay, third attempt! :D13:58
straycatpersia, Most people who use the internet seem to have worked out how to use email13:58
persiaYes.  Most of them use some form of webmail.  This is not amenable to patches.13:58
jmacsWhy not?13:59
persiaMost of the environments in which I have network access block port 25, and many block 587.13:59
persiajmacs: Because patches in HTML make git cry13:59
persiaBut even if one disables HTML, the problem is usually line wrapping, etc.14:00
straycatpersia, So use whatever imap service your webmail provides14:00
straycator, failing that, rent a vps and deploy your own mta14:00
persiastraycat: IMAP doesn't support sending mail.14:00
straycatpersia, you were talking about reading patches14:00
persiaAnd yes, it is possible to sort MTA: my point is that it is annoying.14:00
persiaI said that my first experience with gerrit for *sending* patches was lovely.14:01
persiaFor reviewing patches, I find it a bit annoying to fetch the special ref, etc.14:01
Zarapatch sent14:01
Zaraif you merge it, terrible things will happen14:01
straycatpersia, That's nice for you, everyone else seems to manage just fine with mail. My first patches sending to the ml were just lovely because I can use email. I find it slightly annoying that instead of reviewing a series I have to review a dependant set of individual changes and am coerced into using some web ui, that said, now that I've been coerced I'm working out how to make this a less crappy experience than it already is.14:04
persiaExcellent for you.  But neither of us alone makes a decision.  My renewed statement was in support of Zara's comment about the experience for the first-time submitter.  I agree that for many of the workflows used by many of the developers of this channel, gerrit lacks tooling.14:05
SotKZara: how would you compare the experiences?14:06
Zaraso far, I've found gerrit easier, though that could mostly due to it being documented14:07
Zaramostly *be due14:07
Zarawhereas for the ML I have 'git format patch' and 'git send email' commands, with flags, saved on my computer, that I adjust, because otherwise I have no way of guessing the flags.14:08
ssam2one idea for how to handle features that are lots of individual commits is to merge them all to a staging feature branch14:12
ssam2I've been dreading reviewing the OStree patches, because it's such a big series14:13
ssam2but it doesn't really make sense to merge it piece by piece to 'master'14:13
bashrc_gerrit is maybe ok and I think I can live without the web UI. I understand the limitations of high traffic mailing lists, having seen what happens for the kernel14:13
ssam2maybe a good middle ground would be to submit it piece by piece to gerrit, but merge it to a staging/ostree branch instead14:13
ssam2then when it's all working, staging/ostree can be merged to master14:14
*** mdunford has joined #baserock14:15
straycatAfter speaking with ssam2 and Kinnison there doesn't seem to be a better alternative14:15
ZaraTo sum up: That took me an hour, compared to at least 2 afternoons for the ML. *But* gerrit had comprehensive documentation on the wiki, which wasn't the case for the ML process. I like the fact that gerrit has a UI, where I can find the git commands.14:19
Zara*GUI, even. I also want to know if gerrit is compatible with editing baserock files on the host machine.14:21
straycatAt least not if we want pre-merge testing for morph and definitions, personally I can live without pre-merge testing for morph, but maybe waiting a few minutes for check is too long14:21
*** zoli__ has quit IRC14:24
SotKgiven that I've sent patches to fix check in master twice in the last week or so, I'm looking forward to pre-merge testing for morph14:24
SotKZara: what do you mean by "editing baserock files on the host machine"?14:25
Zara(I don't merge things, so I can only comment as someone submitting patches. So I'd vote for gerrit, given my focus is on UI; I think this makes baserock easier to use, and encourages wider contribution)14:25
ZaraSotK: I cloning a baserock repo to the host, changing files on there and then pushing them. So not using the devel vm during this process (eg: I want to edit a README and for some reason I am irrationally attached to gedit or some other tool with a GUI)14:26
Zarahaha, ignore that 'I' :P14:26
ZaraI can't see why there would be a problem with that but there could be something I've missed.14:28
SotKyou can do that just fine with both the ML and with Gerrit14:28
bashrc_gedit is ok and has a lot of addons14:28
* SotK used to be attached to gedit before he started working on Baserock and had to use VMs all the time too14:29
Zaracool. :) wanted to check in case I discovered some weird incompatibility later. (I used 'irrationally' because I can't explain why I prefer it, but I prefer it a lot. Might just be from a generation used to GUIs.)14:29
* SotK doesn't think liking GUIs is irrational :)14:32
Zarahahaha, I just saw sam's review of my dummy patch, I'm heartbroken ;_;14:36
pedroalvarezZara: do you want to test more things with that patch?14:37
pedroalvarezmaybe send a second version?14:37
straycatMy favourite editor is a gui, some guis are good, I'd argue the gerrit web ui is not an example of such a ui and that in coercing contributors to use this tool the project risks alienating them14:37
straycatAnyhow, I'm going to crawl under my rock, defeated14:38
Zaraaw. :( I think the gerrit web ui needs work but I prefer it to none.14:38
persiaI also aggressively dislike the gerrit web ui, and wish there was any alternative that provided the other features.14:38
Zarapedroalvarez: I'm okay for now, but if I'm struck by more artistic inspiration I'll be sure to patch appropriately.14:39
persiagit-review seems OK for submission.  I've seen people happy with gertty for reviewing, but we need to sort permissions, a certificate (would self-signed do?), etc.14:39
KinnisonI believe we're organising a certificate14:40
*** wschaller has joined #baserock14:45
persiaI also believe we're organising permissions :)14:53
ssam2what do you mean by 'permissions'?15:07
ssam2is this the streamEvents thing?15:07
*** mdunford has quit IRC15:09
persiassam2: Yes, which I understood you to be doing as soon as you had time (but that didn't look that soon by wall-clock).15:11
*** wschaller_ has joined #baserock15:17
ratmice__persia: for webmail (well gmail at least), the thing is that the user needs to set the appropriate mime type for patches in their browser15:18
ratmice__not implying all webmail things may be able to send patches appropriately, but some can be configured to do so15:19
*** wschaller has quit IRC15:20
*** mdunford has joined #baserock15:23
persiaratmice__: Agreed.15:24
ratmice__anyhow that is something that could be included in code submission guidelines.15:25
bashrc_can settings be addd to morph at runtime?15:28
pedroalvarezbashrc_: what do you mean?15:28
persiabashrc_: There's command-line options to override many of the morph.conf values, if that is what you mean.15:29
bashrc_ok, but can those options be added dynamically or do they need to be specified within ?15:29
persiaThey are command-line options.  So `morph foo bar baz quux ...`15:30
bashrc_if I want a new commandline option do I need to edit add_settings in ?15:31
SotKbashrc_: you mean can a morph plugin add a new option?15:31
persiaIf you need to override something not exposed to the command line nor to morph.conf, then yes, you have to modify code.  If you encounter one of these, please expose it to a confugrable interface.15:31
*** ssam2 has quit IRC15:35
*** mdunford has quit IRC15:49
*** bashrc_ has quit IRC15:49
*** franred has quit IRC15:49
*** sherm_ has quit IRC15:51
*** ssam2 has joined #baserock15:55
*** ChanServ sets mode: +v ssam215:55
*** mdunford has joined #baserock15:55
*** bashrc_ has joined #baserock16:01
*** sherm_ has joined #baserock16:02
*** ssam2 has quit IRC16:05
*** franred has joined #baserock16:05
*** sherm_ has quit IRC16:08
*** mike has joined #baserock16:08
*** mike is now known as Guest8717916:09
*** franred has quit IRC16:10
SotKKinnison: I replied to your comments16:11
SotKit seems kinda weird to me that I can vote +1 on my own patches in Gerrit16:12
*** franred has joined #baserock16:12
paulsherwoodSotK: you could do that on the ML too... but please don't :)16:13
SotKI guess16:13
*** bashrc_ has quit IRC16:17
*** franred has quit IRC16:17
*** franred has joined #baserock16:18
*** bashrc_ has joined #baserock16:19
*** ssam2 has joined #baserock16:20
*** ChanServ sets mode: +v ssam216:20
KinnisonSotK: I responded again16:24
SotKKinnison: thanks, I will send a reworked series16:25
SotKKinnison: do you have time to take a look at the partial deploy stuff, since I imagine you'll have the same complaint about --partial there too (and `morph deploy CLUSTER CHUNK CHUNK STRATUM...` didn't make much sense to me)?16:26
SotKThat is on the mailing list since it depends on my OSTree work (but comments on the approach would be useful)16:27
Kinnisonpartial deploy?!16:28
KinnisonThat sounds odd16:28
* Kinnison is almost 1200 mails behind on the ML sadly16:28
KinnisonI'm unlikely to reach it today16:28
SotKIt is odd16:28
doffmIs there an aws deployment extension for baserock?16:29
Kinnisondoffm: I don't think I've seen one16:29
Kinnisondoffm: though if there is, that'd help me trial an idea I had16:29
KinnisonSotK: what happened re: mason via zuul and turbo-hipster?16:30
KinnisonSotK: almost an entire screenful of baserock-dev is dedicated to a patch series related to that which isn't merged or rejected16:31
SotKKinnison: tlsa addressed the review comments and sent new patches for it IIRC16:32
richard_mawdoffm: no, what's missing is a) enable support for being a XEN guest in the kernel, b) import the disk image build tools and add a write extension to generate images in the correct format and c) have that write extension generate a /boot/grub.cfg (or whatever it's called) since the AWS bootl-loader smells like GRUB rather than extlinux16:32
KinnisonSotK: So he did, pity he didn't say that on the old thread :)16:32
* Kinnison sighs in relief as 36 is removed from the total of things he's tracking16:33
* SotK wonders if anyone reviewed that yet16:33
* richard_maw failed to do so last week16:34
richard_mawbut since more patches are going to gerrit instead, I may stand a chance of getting through the mailing list backlog16:34
bashrc_ah, I may be getting somewhere. Adding gerrit settings to morph/ does indeed get past some of the errors I've been seeing16:40
*** zoli__ has joined #baserock16:43
pedroalvarezI wonder what are you trying to achieve16:50
pedroalvarezmixing gerrit settings with morph doesn't sound "right"16:50
*** a1exhughe5 has quit IRC16:51
*** wschaller___ has joined #baserock17:00
* SotK looks up what the magic numbers mean in the OSTree stuff17:02
*** wschaller_ has quit IRC17:03
rjekThat all sounds quite familiar17:04
*** wdutch has joined #baserock17:06
tlsaKinnison: sent v317:11
tlsaand replied to most of your orig comments17:12
SotKrjek: it does indeed17:12
*** zoli__ has quit IRC17:13
ssam2the input format looks a bit more complex than Baserock definitions17:15
perrylyeah, the user manual seems rather...convoluted17:15
rjekThey're not as clear, certianly17:16
radiofreethe coloured dmesg output in the new util-linux is beautiful, thanks for upgrading it17:17
ssam2the concepts of the actual engine must be similar though. I remember someone who works at Google once describing an internal tool that sounded like it worked in a similar way to Morph, when I told them about Baserock17:18
ssam2but this was a couple of years ago, and it seemed the tool was private at that time17:18
ratmice__not sure, it says that all builds are incremental17:20
mwilliams_ctssam2: if I understand the hacker news post, the internal tool is called Blaze and this is a release of some of Blaze17:21
jjardonradiofree: yw17:21
ssam2it may be a lot smarter than Morph if it can do incremental builds while also being certain they are reproducible17:21
ssam2that would be awesome17:21
ratmice__there was a talk that cary coutant did a while back, on incremental linking in gold, where he briefly discussed some of the google build stuff iirc17:22
rjekrummage in its sources for ideas and inspiration17:22
ratmice__IIRC rather than storing like entire sets of source trees they more rely on checking out one source file + include files rather than having an entire dedicated staging area17:24
ratmice__so, the compiles I think are a lot more parallel17:24
jjardonsorry, I always get confused about this: if I want a binary in /bin in the final system, should I use --bindir=/bin or --bindir="$DESTDIR"/bin in configure time?17:24
bashrc_it looks like setting values within the firehose plugin isn't early enough to prevent configuration settings from not being recognised17:24
pedroalvarezjjardon: /bin17:25
jjardonpedroalvarez: thanks17:26
* jjardon fixing patch17:26
*** Krin has quit IRC17:26
* jjardon just discovered how amazing "git review" is17:26
CTtpollardgit-review is nice17:27
ratmice__so like, compiles happen in parallel on various machines, and the object files get sent to the linker machine which links in object files as they come in17:28
ratmice__anyhow I believe the systems must be quite different17:28
ssam2that does sound very different17:31
ssam2and a lot more complex17:31
ratmice__that was the talk iirc (been some time since i've watched it, so hopefully recalling correctly)17:35
tlsamore like distcc then?17:36
bashrc_is there some process needed to enable a plugin, other than setting MORPH_PLUGIN_PATH ?17:37
richard_mawyou can either do that, or put the plugin in morphlib17:37
bashrc_the values from --config seem to take effect before the plugin is enabled, causing errors17:39
*** mdizzle has quit IRC17:39
ssam2if that's so, it must be a bug in cliapp. But I think we'd have already spotted it if that were the case17:45
ratmice__tlsa: not sure, one thing is for certain they don't need to take legacy build systems into account :)17:45
bashrc_I guess I'll need to read the morph code to confirm the order in which things are defined17:48
richard_mawbashrc_: what exactly do you want to do?17:48
bashrc_read some config settings with --config, where the settings are for a plugin17:53
bashrc_exec ${MORPH} firehose "$@" --config=/etc/firehose.conf17:54
bashrc_if the settings are defined in the plugin enable, the morph command crashes saying it doesn't recognise the settings fields17:55
*** pdar has quit IRC17:55
*** tiagogomes_ has quit IRC17:56
richard_mawThat's odd, the distbuild plugin adds settings, and we add config files like that for various distbuild bits like you are trying to do, and it adds settings in the `def enable(self)` function17:56
bashrc_I'll look at how disbuild does it17:56
richard_mawfrom what you describe, it sounds like it does exactly the same thing as you17:57
*** pdar has joined #baserock17:58
*** bashrc_ has quit IRC18:01
*** gfinney has quit IRC18:04
*** Guest87179 has quit IRC18:08
*** ssam2 has quit IRC18:16
*** zoli__ has joined #baserock18:27
Kinnisontlsa: I'll try and look at it tomorrow18:30
*** wschaller___ has quit IRC19:08
*** edcragg has quit IRC19:10
*** gary_perkins has quit IRC19:35
*** rdale_ has quit IRC19:59
*** zoli__ has quit IRC20:54
*** bjdooks has quit IRC21:48
*** bjdooks has joined #baserock21:49
*** robtaylor has quit IRC23:18
*** robtaylor has joined #baserock23:18
*** zoli__ has joined #baserock23:56

Generated by 2.14.0 by Marius Gedminas - find it at!