[ https://issues.apache.org/jira/browse/CASSANDRA-18001?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17633533#comment-17633533 ]
Ekaterina Dimitrova commented on CASSANDRA-18001: ------------------------------------------------- A few updates on my end... trying to be a good citizen it seems my try to do good to the HIGHRES and reduce the parallelism so that people do not spin empty containers for the new jobs didn't work. Manual verification of the file shows things shifted wrong. Also, I hit the issue with CircleCI where commands are squashed again and this makes it super hard for debugging. Last response from CircleCI team on this issue: {color:#de350b}After syncing with our team, I want to share that:{color} {color:#de350b}[YAML is essentially a specification|https://urldefense.com/v3/__https://yaml.org/__;!!PbtH5S7Ebw!b1SNWAq5oUmhmKoW44AWaB-eBX5wSa2T0dD6mpaTuIO6gYotfZnnuccv4Zc0vbn4bQ7ulVg9G_6Sz_k0lEhSMIY7oL_azCRsbw$], and its implementation will depend on the libraries any tooling uses.{color} {color:#de350b}[As stated in the specification|https://urldefense.com/v3/__https://yaml.org/spec/1.2.2/*chapter-6-structural-productions__;Iw!!PbtH5S7Ebw!b1SNWAq5oUmhmKoW44AWaB-eBX5wSa2T0dD6mpaTuIO6gYotfZnnuccv4Zc0vbn4bQ7ulVg9G_6Sz_k0lEhSMIY7oL_w7U4BbA$], "The amount of indentation is a presentation detail and must not be used to convey content information."{color} {color:#de350b}This means that as long as a YAML expression is still data-wise equivalent, the formatted output of a YAML expression can depend on the library itself.{color} {color:#de350b}As such, certain libraries can choose to "compress" the expressions, so long as the data is still the same data-wise.{color} {color:#de350b}Such compressions can be beneficial, since the library may see reducing the overall filesize being an important thing.{color} {color:#de350b}In other words, there is really no hard-guarantees to preserving a YAML expression's format.{color} {color:#de350b}We think, *if* you really want to ensure the {{commands}} format are always preserved / expressed in blocks, you would thus want to post-process.{color} {color:#de350b}I have set up an example here using the js-yaml Node package:{color} {color:#de350b}[https://replit.com/@kelvintaywl/YAMLprettifier?v=1|https://urldefense.com/v3/__https://replit.com/@kelvintaywl/YAMLprettifier?v=1__;!!PbtH5S7Ebw!b1SNWAq5oUmhmKoW44AWaB-eBX5wSa2T0dD6mpaTuIO6gYotfZnnuccv4Zc0vbn4bQ7ulVg9G_6Sz_k0lEhSMIY7oL-fJ6Nr3Q$]{color} {color:#de350b}(If you run the repl code, you should see the post-processing expanding the "compressed" command into a block){color} *I do not think we can/should fix the whole world in one ticket and also that trying that hides risks.* {color:#172b4d}On the bright side MIDRES tests and files seem good to me.{color} {color:#172b4d}*STEPS TAKEN TO UNBLOCK 4.1 OVER THE WEEKEND:* {color} - 4.1 and trunk finished as there are various things that differ in branches and few different bugs we caught... After that, when we are not pressed by the 4.1 release, I will fix 3.11, 4.0 and 3.0 patches over the next week. * I fixed now the 4.1+ patches for HIGHRES * I finished the cqlshlib * My testing and an old ticket for JDK 11 improvements proved that the burn tests in 4.1 and trunk (some of them) crash even with X-Large containers. I verified in Jenkins and Butler that all the burn tests are run and they finish successfully. SO I think those are not a blocker, we can use Jenkins ||Patch||LOWRES||MIDRES||HIGHRES|| |[4.1|https://github.com/ekaterinadimitrova2/cassandra/tree/18001-4.1-3]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra?branch=18001-4.1-3]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra?branch=18001-4.1-3-mid]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra?branch=18001-4.1-3-high]| |[trunk|https://github.com/ekaterinadimitrova2/cassandra/tree/18001-trunk-3]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra?branch=18001-trunk-3]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra/2130/workflows/9df2d073-c26a-4750-82e4-8b6645f37f9c]|[CI|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra?branch=18001-trunk-3-high2]| _*trunk still running - I will check it tomorrow morning_ Reminder about dtests large - _There are two tests that fail because there are not enough resources to bring up 9 nodes. Spinning a 2X Large container for all large tests because of 2 is not justified._ I can push the HIGHRES and pre-commit runs tomorrow if people are ok with the MIDRES and the current flow (trying to save some resources) > Add missing tests suites to CircleCI > ------------------------------------ > > Key: CASSANDRA-18001 > URL: https://issues.apache.org/jira/browse/CASSANDRA-18001 > Project: Cassandra > Issue Type: Task > Components: CI > Reporter: Ekaterina Dimitrova > Assignee: Ekaterina Dimitrova > Priority: Urgent > Fix For: 3.0.x, 3.11.x, 4.0.x, 4.1-rc, 4.1.x, 4.x > > > Burn tests to all branches, large Python DTests (with/without vnodes), > cqlshlib not tested in all branches and with all jdks; Java distributed tests > not running with J8/J11 4.0+ -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org