| *** gouthamr has joined #zuul | 00:09 | |
| openstackgerrit | Merged openstack-infra/zuul-jobs master: Switch to http://security.debian.org/ for debian https://review.openstack.org/562349 | 00:32 |
|---|---|---|
| tristanC | corvus: good idea, i'll remember to do the rebase step in an extra ps. | 00:53 |
| *** pwhalen_ has joined #zuul | 01:22 | |
| *** pwhalen has quit IRC | 01:23 | |
| *** JasonCL has quit IRC | 01:31 | |
| *** pwhalen_ is now known as pwhalen | 01:33 | |
| *** pwhalen has quit IRC | 01:34 | |
| *** pwhalen has joined #zuul | 01:34 | |
| *** ssbarnea_ has quit IRC | 01:51 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: Add allowed-triggers and allowed-reporters tenant settings https://review.openstack.org/554082 | 02:10 |
| *** pwhalen has quit IRC | 02:15 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Refactor NodeLauncher to be generic https://review.openstack.org/535555 | 02:23 |
| *** pwhalen has joined #zuul | 02:43 | |
| *** pwhalen has joined #zuul | 02:43 | |
| *** swest has quit IRC | 02:46 | |
| *** swest has joined #zuul | 02:47 | |
| *** harlowja_ has quit IRC | 02:53 | |
| *** toabctl has quit IRC | 03:00 | |
| *** harlowja has joined #zuul | 03:22 | |
| *** harlowja has quit IRC | 04:52 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: Add allowed-triggers and allowed-reporters tenant settings https://review.openstack.org/554082 | 05:33 |
| *** yolanda has joined #zuul | 05:52 | |
| *** hashar has joined #zuul | 05:58 | |
| *** toabctl has joined #zuul | 06:04 | |
| *** dims has quit IRC | 06:54 | |
| *** dims has joined #zuul | 06:56 | |
| *** dims has quit IRC | 07:01 | |
| *** dims has joined #zuul | 07:02 | |
| *** bhavik1 has joined #zuul | 07:05 | |
| *** bhavik1 has quit IRC | 07:07 | |
| *** dmellado has joined #zuul | 07:32 | |
| *** jpena|off is now known as jpena | 07:39 | |
| *** electrofelix has joined #zuul | 07:40 | |
| *** ssbarnea_ has joined #zuul | 08:33 | |
| *** rcarrillocruz has quit IRC | 08:39 | |
| *** hashar is now known as hasharAway | 09:42 | |
| *** pbrobinson has quit IRC | 09:54 | |
| *** pbrobinson has joined #zuul | 09:58 | |
| *** xinliang has quit IRC | 10:39 | |
| *** xinliang has joined #zuul | 10:51 | |
| *** jpena is now known as jpena|lunch | 10:59 | |
| *** mugsie has quit IRC | 11:42 | |
| *** mugsie has joined #zuul | 11:42 | |
| *** mugsie has quit IRC | 11:42 | |
| *** mugsie has joined #zuul | 11:42 | |
| *** jpena|lunch is now known as jpena | 12:02 | |
| *** rlandy has joined #zuul | 12:30 | |
| *** ssbarnea_ has quit IRC | 12:47 | |
| *** ssbarnea_ has joined #zuul | 13:01 | |
| *** ssbarnea_ has quit IRC | 13:03 | |
| *** ssbarnea_ has joined #zuul | 13:09 | |
| *** JasonCL has joined #zuul | 13:09 | |
| *** dkranz has joined #zuul | 13:22 | |
| *** swest has quit IRC | 13:29 | |
| Shrews | tobiash: have you done any ansible 2.5 testing? if not, i'm going to start on that | 14:33 |
| tobiash | Shrews: unfortunately my plan for this week failed and I had to do other things | 14:40 |
| tobiash | So I haven't started an this yet | 14:41 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: WIP: Upgrade to Ansible 2.5 https://review.openstack.org/562668 | 15:06 |
| *** ssbarnea_ has quit IRC | 15:09 | |
| Shrews | I don't guess anyone has written a script to compare differences in Ansible repo modules with the Zuul copies, have they? | 15:12 |
| Shrews | Yeah, didn't think so. :( | 15:12 |
| dmsimard|off | Shrews: hmm... that sort of reminds me of a tool I've seen in the past that scored forks based on amount of changes in a project.. let me find that | 15:13 |
| dmsimard|off | the only thing that comes to mind is debtcollector but that's not what I'm looking for | 15:14 |
| dmsimard|off | I'll need to send an email to find it -- let me get back to you on that | 15:15 |
| *** hasharAway is now known as hashar | 15:15 | |
| Shrews | dmsimard|off: no worries. i'll come up with something | 15:15 |
| Shrews | I guess it's really just the command.py module I should be concerned about | 15:19 |
| pabelanger | where did we land on 2.4 support for zuul? Given that it is likely disruptive to jobs, we'd likely need to have some hands on deck for openstack. But should we start communicating we are looking to bump the ansible version soon? | 15:21 |
| dmsimard|off | They released Ansible 2.5.1 today. We're late :) | 15:23 |
| dmsimard|off | 2.3 is actually EOL | 15:23 |
| Shrews | pabelanger: I suggested a while back we just make the jump to 2.5 since that is released. | 15:23 |
| Shrews | pabelanger: but we need to make sure that works, which is what i'm doing now | 15:24 |
| Shrews | that would give us more time to come up with a multi-version solution | 15:24 |
| pabelanger | dmsimard|off: where did you see 2.3 was EOL? | 15:25 |
| dmsimard|off | pabelanger: they only support the two latest releases https://github.com/ansible/ansible/blob/devel/docs/docsite/rst/reference_appendices/release_and_maintenance.rst | 15:26 |
| dmsimard|off | (I linked to github because for some reason http://docs.ansible.com/ansible/latest/reference_appendices/release_and_maintenance.html is not up to date) | 15:26 |
| corvus | anyone else want to review the mqtt driver addition? | 15:27 |
| pabelanger | dmsimard|off: thanks | 15:27 |
| corvus | (it has 2x+2 -- i just didn't want it to slip by anyone interested) | 15:27 |
| pabelanger | Shrews: my concern about 2.5, is it might be too new. I'd be curious to see if anybody in openstack community is actually using it yet (as a gauge) | 15:28 |
| pabelanger | but agree, we need to do something | 15:28 |
| dmsimard|off | pabelanger: there is a 2.5.1 so I would like to think the issues from 2.5.0 (if there was any) have been ironed out | 15:29 |
| Shrews | pabelanger: newness doesn't scare me. backward compatibility does | 15:29 |
| pabelanger | +1 | 15:29 |
| pabelanger | I haven't had a good upgrade storey with ansible yet | 15:29 |
| pabelanger | story* | 15:30 |
| Shrews | but you know if we were to announce to openstack we intend to upgrade 2.4, we're inevitably going to be asked "why not 2.5?" | 15:30 |
| pabelanger | sure, I think we'd also see why 2.5, why not 2.4? | 15:31 |
| Shrews | because 2.5 goes to 11 | 15:32 |
| openstackgerrit | Merged openstack-infra/zuul-website-media master: Run zuul-website jobs https://review.openstack.org/561750 | 15:32 |
| *** ssbarnea_ has joined #zuul | 15:35 | |
| pabelanger | OSA teams is currently testing ansible 2.5, few issues but seems to be around their connection plugins | 15:42 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: late bind pipelines https://review.openstack.org/553618 | 15:45 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: late bind pipelines https://review.openstack.org/553618 | 15:54 |
| *** electrofelix has quit IRC | 16:08 | |
| *** hashar is now known as hasharAway | 16:17 | |
| dmsimard|off | Shrews: regarding our earlier discussion: https://softwarefactory-project.io/r/gitweb?p=debtor.git;a=blob;f=README.rst;h=cd43850a5f1721d86731ee968ad4130de82dc6c6;hb=HEAD | 16:28 |
| dmsimard|off | (it's way overkill for what we need but I said I would follow up) | 16:29 |
| dmsimard|off | It's also made to track /project/ forks (not "module" forks) | 16:29 |
| clarkb | re ansible EOL what does that actually mean? was ansible properly supporting old branches? I don't recall any of my bugfixes ever getting backported for example | 16:38 |
| clarkb | I guess its more a matter of keeping up with what third party modules will support? | 16:38 |
| *** ssbarnea_ has quit IRC | 16:53 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Late bind projects https://review.openstack.org/553618 | 17:04 |
| *** gouthamr is now known as gouthamr|afk | 17:06 | |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: WIP: Upgrade to Ansible 2.5 https://review.openstack.org/562668 | 17:08 |
| *** jpena is now known as jpena|off | 17:10 | |
| corvus | the late-bind projects change (formerly known as late-bind pipelines) passes all of the non-postgres tests locally (i haven't set up pg yet), so i think it's ready for review | 17:12 |
| corvus | it'll probably be at least an hour before the tests results come back though because of openstack backlog | 17:12 |
| clarkb | that reminds me I should rereview the secrets change | 17:12 |
| corvus | i'm hoping (i need to regroup and see where we are now) that puts us pretty close to severing all of the links between config objects. once that's done, we can start individually caching them and re-use them on dynamic reconfiguration. then hopefully use much less ram. | 17:14 |
| corvus | fbo_: ^ that will conflict with the load-broken-config change | 17:15 |
| pabelanger | Shrews: clarkb: shocking testing ansible 2.5.1 with existing playbooks seems to have worked | 17:18 |
| clarkb | oh its an entire stack I'll start at bottom of stack | 17:22 |
| Shrews | corvus: pabelanger: mordred: oh, 2.5 brings us this lovely thing: "Added a configuration file that a site administrator can use to specify modules to exclude from being used." | 17:25 |
| pabelanger | yah, I heard about that. But haven't looked at it yet | 17:26 |
| Shrews | maybe we don't have to have zuul carry around those disabled shim modules | 17:26 |
| clarkb | ya dmsimard|off pointed it out. | 17:28 |
| pabelanger | Shrews: +1 | 17:29 |
| Shrews | i am pestering bcoca to point me to docs on it | 17:32 |
| clarkb | it won't help where we conditionally disable modules based on host or source etc | 17:33 |
| clarkb | but for straight up disabled modules may be useful | 17:33 |
| Shrews | we always disable the networking modules | 17:33 |
| Shrews | so at least useful there | 17:33 |
| Shrews | jimi|ansible: gundalow: is that new feature documented? | 17:33 |
| Shrews | cannot seem to find it in the current docs | 17:35 |
| *** patriciadomin has quit IRC | 17:35 | |
| dmsimard|off | Shrews: I have it somewhere in my #ansible-devel logs when I asked about it -- I'll pick it up when I'm not on my phone :p | 17:42 |
| jimi|ansible | i honestly thought the whitelist/blacklist stuff would have been based on what you all were doing, i'll have to check | 17:45 |
| jimi|ansible | i'd be surprised if we didn't have docs for that | 17:45 |
| Shrews | jimi|ansible: yeah, not jumping out at me anywhere | 17:50 |
| *** ssbarnea_ has joined #zuul | 17:50 | |
| openstackgerrit | Merged openstack-infra/zuul master: Test base job secrets https://review.openstack.org/561030 | 17:53 |
| gundalow | Hi | 17:55 |
| * gundalow reads | 17:55 | |
| gundalow | Shrews: Docs for what? | 17:56 |
| pabelanger | Shrews: well, all my tests on playbooks / roles passed using ansible 2.5.1. I'm going to start bumping things, but really good that nothing broke or needed to update playbooks | 17:56 |
| gundalow | pabelanger: \o/ | 17:56 |
| pabelanger | gundalow: indeed! | 17:56 |
| Shrews | gundalow: for the new 2.5 config file that "a site administrator can use to specify modules to exclude from being used" | 17:56 |
| *** gouthamr|afk is now known as gouthamr | 17:57 | |
| Shrews | pabelanger: what playbooks are you testing? | 17:58 |
| pabelanger | Shrews: roles / playbooks for windmill | 17:58 |
| Shrews | oh | 17:58 |
| pabelanger | but still using 2.3 to drive the 2.5.1 ansible-playbook | 17:58 |
| pabelanger | and talking to OSA, they don't seem to have many changes that are needed either | 17:59 |
| pabelanger | hoping that means when we do release 2.5.1 for zuul, minimal changes needed to our jobs | 17:59 |
| Shrews | pabelanger: the porting guides for 2.4 and 2.5 might be useful (http://docs.ansible.com/ansible/latest/porting_guides/porting_guides.html) | 18:00 |
| Shrews | i didn't see anything when i skimmed 2.5 porting guide. haven't looked at 2.4 one yet | 18:00 |
| Shrews | well, i guess those docs do not exist and I'll have to go skimming code. i can do that after my walk. biab | 18:02 |
| openstackgerrit | Merged openstack-infra/zuul master: Make gearman calls async in ZuulWeb https://review.openstack.org/560026 | 18:13 |
| *** pabelanger has quit IRC | 18:17 | |
| *** pabelanger has joined #zuul | 18:17 | |
| jlk | folks, we picked bubblewrap for executer, partly BECAUSE it can be used within a container, right? | 18:29 |
| clarkb | jlk: iirc it was picked because it aims to do proper isolation whereas other runtimes do not? But it can be run in a container too (tobiash does this iirc) | 18:30 |
| *** ssbarnea_ has quit IRC | 18:30 | |
| pabelanger | it also could be run as non-root users | 18:32 |
| pabelanger | which right kernel | 18:32 |
| pabelanger | with* | 18:32 |
| clarkb | ya my local install is no setuid and I can run it as my normal user | 18:36 |
| pabelanger | I actually wonder how hard it would be to wrap DIB with bwap, since people often complain about the root permissions it needs | 18:38 |
| clarkb | I don't think bwrap would address those (mount in particular) | 18:39 |
| pabelanger | yah, i think more of the extra isolation if DIB breaks out of a chroot, eg: recent iptables issue on nodepool-builder | 18:40 |
| clarkb | same issue | 18:40 |
| pabelanger | yah? | 18:40 |
| clarkb | pretty sure if running brwap as non root user you don't get access to things that would require root like allowing mounts or getting new network namespace | 18:41 |
| pabelanger | oh, maybe. I thought there was a way to allow that via cgroups | 18:41 |
| clarkb | but ya isolating networking would be nice to prevent hosing the builder | 18:47 |
| pabelanger | +1 | 18:48 |
| pabelanger | I mean, I view them as disposable personally, aside from the time it takes to rebuilt / update dns | 18:49 |
| clarkb | mount is likely to be the trickiest bit | 18:49 |
| clarkb | unless you coordinated the mounting outside of bwrap and bind mounted the results into where the elements are running | 18:50 |
| clarkb | but ya I use disposable vm for dib locally | 18:50 |
| clarkb | pabelanger: from bwrap readme "Bubblewrap could be viewed as setuid implementation of a subset of user namespaces. Emphasis on subset - specifically relevant to the above CVE, bubblewrap does not allow control over iptables." | 18:55 |
| *** hasharAway is now known as hashar | 18:57 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Increase unit testing of host / group vars https://review.openstack.org/559405 | 19:00 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Inventory groups should be under children key https://review.openstack.org/559406 | 19:00 |
| pabelanger | clarkb: so, I parse that as it might of helped with recent iptables issue, if we figured out mounts | 19:02 |
| clarkb | I guess that would be fine in the dib case since the build doesn't want ot update iptables at runtime but when image actually boots | 19:02 |
| clarkb | pabelanger: ya it would've failed the build I think | 19:02 |
| clarkb | (and prevented it from derping the host iptables) | 19:02 |
| pabelanger | yah | 19:02 |
| pabelanger | a friday afternoon experiment | 19:02 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: WIP: Update to Ansible 2.5 https://review.openstack.org/562668 | 19:06 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: WIP: Update to Ansible 2.5 https://review.openstack.org/562668 | 19:09 |
| tobiash | clarkb, pabelanger: yocto (builds custom distros from source) can generate diskimages without root using fakeroot | 19:11 |
| tobiash | maybe that also would be possible for diskimage-builder | 19:11 |
| clarkb | ya there are tools for making docker images now without root too | 19:11 |
| clarkb | I'm sure its possible its just a matter of figuring out what the constraints are | 19:11 |
| tobiash | but I guess that would be more than a friday afternoon experiment | 19:11 |
| pabelanger | yah, I'm not sure why we don't use fakeroot for DIB. I'm assuming there is a reason | 19:29 |
| clarkb | probably because you still need root to mount | 19:34 |
| clarkb | so it was seen as unneeded overhead? spamaps may know the history there | 19:34 |
| clarkb | I think if dib itself needs root for mount but the elements don't that would be straightfoward enough | 19:35 |
| clarkb | then you can isolate build host from the building of iamges and then shuffle bits as necessary | 19:35 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Late bind projects https://review.openstack.org/553618 | 19:38 |
| *** harlowja has joined #zuul | 20:03 | |
| pabelanger | ah, new build dependency for zuul, if not using wheels: re2/re2.h: No such file or directory | 20:06 |
| corvus | pabelanger: oh we forgot to add that to bindep | 20:08 |
| corvus | pabelanger: libre2-dev | 20:08 |
| pabelanger | I believe bindep is right, roles haven't been updated to use bindep.txt yet | 20:09 |
| clarkb | tobiash: because i totally got nerd sniped, I think the way yocoto does it is they don't support partitioning. They write a filesytem to a file and then produce an image with that (so its single fs) | 20:09 |
| clarkb | http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/image_types.bbclass magic seems to happen there | 20:09 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: Make config objects freezable https://review.openstack.org/562816 | 20:11 |
| corvus | clarkb, tobiash: ^ can you take a quick look at that and tell me if you like that approach? | 20:11 |
| clarkb | corvus: sure, I should get back to code review now that i have thoroughly been nerd sniped on the how to avoid root/mount problem for image building | 20:12 |
| corvus | clarkb, tobiash: i spotted an error in my last change; that's a response to that. see my comment on https://review.openstack.org/553618 for more context | 20:12 |
| clarkb | I only see a -1 without a comment on 553618 | 20:13 |
| corvus | (that change should actually fail tests because of the error in 562816 -- i mean, it's not an *error* yet, it would only be an error in the future once we start re-using config objects across layouts) | 20:13 |
| clarkb | ? | 20:13 |
| corvus | drat | 20:13 |
| corvus | clarkb: fixed. was on prev patchset. | 20:13 |
| clarkb | corvus: off the top of my head iirc subclassing Dict and List has issues but i'd hvae to go reread on what those issues are | 20:16 |
| corvus | clarkb: yeah, my understanding is that this particular usage should be okay, but i can change it if that's wrong | 20:18 |
| corvus | (because, basically, we're just adding an extra check to the underlying methods) | 20:19 |
| clarkb | it has to do with __setitem__ and __getitem__ being implemented in C separate from the python implementation | 20:19 |
| clarkb | or at least getitem is | 20:19 |
| corvus | clarkb: right, so in this case, we've got a python __setitem__ method which performs an extra check, then calls the C __setitem__. and we're not changing setitem. | 20:20 |
| corvus | er, not changing getitem | 20:20 |
| clarkb | corvus: I think you may need to handle dict.update() too | 20:25 |
| corvus | clarkb: good call | 20:25 |
| clarkb | but reading on the issues doing this assuming __setitem__ is used for all updates would be the issue so explicit handling of udpate too should cover the bases? | 20:25 |
| clarkb | __getitem__ and get() being similar this way | 20:26 |
| pabelanger | is anybody else seeing a failure in test_job_list_in_project_template_not_dict_error ? | 20:26 |
| clarkb | corvus: https://docs.python.org/3/library/collections.abc.html#collections.abc.Mapping appears to be the recommened way to do this (you implement the methods for that abc) | 20:27 |
| pabelanger | http://logs.openstack.org/05/559405/2/check/tox-py35/728f6a9/testr_results.html.gz | 20:28 |
| corvus | clarkb: i'd like to keep the performance of list/dict as much as possible; if i do that, i'm adding in another layer, right? | 20:29 |
| clarkb | corvus: ya you'd likely have a dict as a class attribute rather than subclessing dict | 20:29 |
| corvus | it's not like the python docs say "dont't do this": https://docs.python.org/3.5/library/collections.html?highlight=userdict#userdict-objects "The need for this class has been partially supplanted by the ability to subclass directly from dict" | 20:30 |
| corvus | if we use the abc, we have to implement the getters too, but if we subclass, we get the straight C versions of those | 20:30 |
| clarkb | ya | 20:31 |
| clarkb | corvus: https://docs.python.org/3.5/library/types.html#types.MappingProxyType there is also that | 20:32 |
| clarkb | which i think may be implemented as the thing you don't want | 20:32 |
| clarkb | (but not sure of that) | 20:32 |
| clarkb | I don't see a corresponding list proxy | 20:32 |
| corvus | clarkb: interesting; yeah, that seems like it would incur a performance penalty on reads | 20:33 |
| corvus | it's pretty much the interface i want though :) | 20:33 |
| corvus | i'll look at the implementation real quick | 20:33 |
| clarkb | oh the docs direct link to the source now thats nice | 20:34 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: Make config objects freezable https://review.openstack.org/562816 | 20:35 |
| corvus | biab | 20:35 |
| corvus | back. i don't understand mappingproxytype. | 20:38 |
| clarkb | https://github.com/python/cpython/blob/d019bc8319ea35e93bf4baa38098ff1b57cd3ee5/Objects/descrobject.c#L808 it appears to also be implemented in c | 20:42 |
| corvus | clarkb: yeah, i just found that | 20:42 |
| corvus | clarkb: well, the list version of this would be to just convert it to a tuple | 20:43 |
| corvus | so i could have my freeze method replace lists with tuples, and dicts with mappingproxies before freezing the object itself | 20:43 |
| clarkb | thinking out loud that may have memory benefits because tuples don't need to be able to grow like lists do? | 20:44 |
| clarkb | the implementation of the mapping proxy doesn't seem to hae that benefit | 20:45 |
| corvus | maybe? i'm not sure about that. | 20:45 |
| clarkb | https://nedbatchelder.com/blog/201608/lists_vs_tuples.html says tuple are a little more space efficient. Probably not worth worrying too much about | 20:48 |
| clarkb | (which confirmed my suspicion that because lists overallocate and reallocate if you append enough there is a small amount of extra pointer memory floating around) | 20:48 |
| clarkb | I'm going to go and give this cold some rest, I'll try to be back later to catch up on ^^ if you learn anything else | 20:49 |
| clarkb | also that type(type.__dict__) thing is magic | 20:49 |
| corvus | clarkb: ok. i'm inclined to go with mappingproxy and tuple, since they're likely easier and more efficient. | 20:50 |
| corvus | i'll work that up in just a bit | 20:50 |
| corvus | clarkb: thanks, and feel better | 20:50 |
| *** gundalow has quit IRC | 20:56 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: Make config objects freezable https://review.openstack.org/562816 | 21:03 |
| corvus | clarkb: that's a version with mappingproxy/tuple | 21:04 |
| clarkb | corvus: left a comment on ^ but its a minor thing | 21:35 |
| clarkb | I like that version its requires a lot less magic | 21:36 |
| corvus | clarkb: yeah; i could drop configurationobject, but i was thinking that it might be nice to indicate *why* it's freezable... i think we might end up with some freezable things which aren't configurationobjects... but i dunno... we'll see how it shakes out. | 21:44 |
| clarkb | there is a question about override-checkout and handling different branches in different repos on the zuul mailing list. Can the job specify an override-checkout for both repos then when that job runs on branch foo on repo A it does the right thing and when run on branch bar on repo B it does the right thing if both repos are set to override to foo and bar respectively? | 21:47 |
| corvus | clarkb: if i'm following correctly, i *think* https://zuul-ci.org/docs/zuul/user/config.html#attr-pragma.implied-branches may help. what do you think? | 21:56 |
| corvus | clarkb: (but, also, i think the solution in the email would work too. that's the explicit version, pragma.implied-branches is the implicit version) | 21:57 |
| corvus | i'll draft a reply and paste it here before i send it out | 21:58 |
| clarkb | ah yup, the docs even seem to point out it is this specific case | 21:58 |
| clarkb | just not "stable" but a mapping between the two none the less | 21:59 |
| corvus | yeah, we wrote this for, i think, openstack <-> ceph branches or something like that. | 21:59 |
| corvus | like, queens uses emerald or somesuch | 22:00 |
| *** hashar has quit IRC | 22:00 | |
| corvus | https://etherpad.openstack.org/p/M58BwQJ7l1 | 22:24 |
| corvus | clarkb: how's that look? | 22:24 |
| clarkb | corvus: ya I think that is helpful for explaining how to use branches | 22:28 |
| clarkb | in the config I hsould say | 22:28 |
| corvus | i did not set out to answer the question 3 different ways, i just couldn't stop typing :) | 22:28 |
| clarkb | might even point mriedem to it once its sent as it is realted to the nova stable job experience they had recently | 22:29 |
| corvus | sent! | 22:30 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Late bind projects https://review.openstack.org/553618 | 22:37 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP: Make config objects freezable https://review.openstack.org/562816 | 22:37 |
| tobiash | clarkb: with yocto you can create multiple partitions (using parted) | 23:36 |
| tobiash | e.g. see https://github.com/Freescale/meta-fsl-arm/blob/master/classes/image_types_fsl.bbclass | 23:36 |
| clarkb | tobiash: doesn't that require root? | 23:37 |
| clarkb | tobiash: to partition a device in particular | 23:37 |
| tobiash | Nope that works against a local (sparse) file | 23:37 |
| clarkb | huh ok so ${SDCARD} is a sparse file that gets an mbr/gpt partition table written to it and multiple filesystems? /me looks to see how mkfs targets specific parittions | 23:38 |
| tobiash | clarkb: mkfs probably works on its own files and get dd'ed finally into the partition | 23:40 |
| clarkb | ya that would line up with my reading of the ext* mkfs stuff | 23:41 |
| clarkb | it oeprates on its own file for sure | 23:42 |
| tobiash | corvus: I like the approach of freezing the reusable config options | 23:42 |
| clarkb | and then that file can be converted to things like qcow2 in the same .bb | 23:42 |
| clarkb | (so I guess you'd layer in this sdacrd stuff as needed) | 23:42 |
| tobiash | yes | 23:42 |
| clarkb | https://github.com/Freescale/meta-fsl-arm/blob/master/classes/image_types_fsl.bbclass#L305 there it is | 23:43 |
| tobiash | by default yocto just generates filesystem images | 23:43 |
| clarkb | that seems incredibly hacky but I guess it works | 23:43 |
| clarkb | so ya dib could construct a raw disk in this manner too. Then do conversions as a regular user | 23:44 |
| clarkb | I wonder what the performance comparison looks like | 23:46 |
| tobiash | I think both do a final cop, | 23:47 |
| tobiash | copy | 23:47 |
| clarkb | ya they do | 23:47 |
| tobiash | Maybe yocto does an extra one but not sure | 23:48 |
| clarkb | but one goes through a loopback device (dib) and the other uses dd on file directly | 23:48 |
| clarkb | (in this case I wouldn't be surprised if the dd'ing was faster) | 23:48 |
| clarkb | oh except right it writes it to one file (the fs file) that then gets written to the image file | 23:48 |
| clarkb | so that would be an extra copy | 23:49 |
| tobiash | The dd could benefit more from caching | 23:49 |
| clarkb | its unfortunate all the mkfs's have different flags to do the fs copy but otherwise this is reasonably straightforward | 23:59 |
| clarkb | doesn't looks like they support xfs? | 23:59 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!