| *** hashar has joined #openstack-jjb | 06:12 | |
| *** hashar is now known as hasharAway | 13:25 | |
| *** mnencia has joined #openstack-jjb | 15:22 | |
| zxiiro | mnencia: thanks for joining. Just an FYI our main focus right now is to try to get the much needed JJB 2.0 released soon. | 15:27 |
|---|---|---|
| zxiiro | electrofelix: speaking of that the YAML parser patch seems to have broken macro parsing inside of plugins. You might want to take a look at discussions from Tuesday above. | 15:30 |
| zxiiro | I think we cannot release until we have a solution to this since it breaks a core feature that we're using heavily | 15:30 |
| mnencia | Hi, I'm looking at https://blueprints.launchpad.net/openstack-ci/+spec/jenkins-job-builder-folders because I've a lot of jobs to generate and I'd like to organize them in folders.? I see there are two patch attached to that blueprint, and one is abandoned. I could work a bit to make this advancing, but I don't want to waste time trying to fix something that is obsolete and cannot be merged. | 15:37 |
| electrofelix | hmm, fun, I guess we were essentially doing a recursive lookup as we dispatched to each in turn | 15:37 |
| electrofelix | mnencia: I expect that just needs the conflict resolved and it should be ok, the other missing piece was something to create folders and I believe there is a patch floating around to handle that as well that needs updating | 15:39 |
| electrofelix | sorry forgot to reference the patch | 15:39 |
| electrofelix | mnencia: https://review.openstack.org/134307 just needs rebasing for the recent conflict due to changes to the parser.py | 15:40 |
| electrofelix | I've just not been able to devote enough time to jjb for a while to get it over the line | 15:40 |
| mnencia | there is there are two plugin that can create folders jenkins-job-builder-addons (but it breaks generation of other thinks, not sure why) and https://github.com/jagardaniel/jenkins-job-builder-folder (tat is very embrional) | 15:41 |
| electrofelix | https://review.openstack.org/189697 comprises of part of the work for creation, but it needs something to ensure that the folder is created/updated before jobs are uploaded to it and ensure that is true for when using the parallel workers (probably by processing all folders first) | 15:42 |
| electrofelix | but even if you process all folders first, you need to make sure that any nested folders are done in the correct order as well so with parallel workers there is a little bit of work to solve this | 15:43 |
| electrofelix | zxiiro: you know that one is an unusual case, it's not that nested macros were supported, but that we were using registry.dispatch from the postbuildscript plugin code | 15:45 |
| electrofelix | So in removing the macro resolving from the dispatch it means we need to change the call in the postbuildscript plugin call to either expand first or find a way to spot at the top level | 15:47 |
| electrofelix | and perform the expansion first | 15:47 |
| electrofelix | btw, you know that postbuildscript plugin is no longer available online to install? | 15:48 |
| zxiiro | electrofelix: ah really ok. But it does seem like other plugins need updating too. like the Maven Project plugin or seems like any plugins that need to call another macro of the same module. | 15:48 |
| zxiiro | electrofelix: yes we are aware. in fact we're trying to find a replacement but haven't gotten one we like yet. So we're side loading that plugin for now. | 15:49 |
| zxiiro | there's some concern that we might have to develop our own plugin that does what we need. | 15:49 |
| electrofelix | zxiiro: posttask does something similar, we just have it search for the first line when we want a task to always run at the end in case of failure/success ;-) | 15:51 |
| mnencia | electrofelix there is also an issue deleting old jobs: if they reside in a folder they get ignored. Also delete-all doesn't work for me. Not sure why | 15:52 |
| electrofelix | I'm pretty certain the security issue was to do with the groovy code, so if it was refactored slightly to reuse the groovy plugin instead the problem should go away | 15:52 |
| electrofelix | mnencia: that should be solved by the first patch, as to delete jobs you need to delete based on the full name rather than the short name of the job | 15:53 |
| mnencia | ok, I will take a look in the weekend | 15:53 |
| electrofelix | zxiiro: "grep -Ir dispatch jenkins_jobs/modules/" shows all the problem ones, once we have one solution, it's just a case of applying wherever else we call dispatch to hand off to the correct module | 15:53 |
| electrofelix | It does mean that there is a limitation that needs to be discussed with waynr about expansion works | 15:54 |
| zxiiro | electrofelix: ah that makes sense | 15:55 |
| electrofelix | which means we might need to reconsider the approach and pass down some objects to do expansion at the last possible moment instead of upfront in order to get consistent behaviour at all points | 15:55 |
| zxiiro | electrofelix: yeah we don't need groovy at all. we just need a thing that let's us run builders | 15:56 |
| electrofelix | mnencia: have a look at the changes in https://review.openstack.org/#/c/134307/12/jenkins_jobs/builder.py, the changes to the delete_old_managed should handle deleting from inside folders | 15:57 |
| electrofelix | mnencia: would be good to have that confirmed though | 15:57 |
| *** electrofelix has quit IRC | 16:47 | |
| *** hasharAway has quit IRC | 18:51 | |
| *** hashar has joined #openstack-jjb | 18:52 | |
| *** hashar has quit IRC | 21:54 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!