From: James T. <ja...@fl...> - 2020-11-11 10:17:46
|
(If you’ll pardon the old ATC reference) I was making a coffee and thinking about my sanity, and have a proposal for landing all the things currently on approach: - Compositor as default renderer - CompositeViewer - Nathan’s ortho-scenery patches What I’m keen to avoid, is a situation where we have regressions but don't know which change caused it. So I’d like to propose the follow timeline: - starting Monday next week, Fernando merges the compositor stuff this will be on by default (or rather, there will no longer be a switch to do anything else) so we will see some regressions around it, but hopefully very few (and we will know they are caused by the compositor) - once that is green, Jules rebases on top and tests, so we can do a straight merge of CompositeViewer CompositeViewer is opt-in, so anyone encountering issues with it, can trivially test *without it* as well, to decide if the issue is Compositor or CompositeViewer - one that merge is green, Nathan can rebase on top (which I suspect may be trivial, since his changes are fairly isolated) Again Nathan’s stuff is-opt in, so anyone seeing regressions can turn it off easily to check if it’s Compositor or Ortho-scenery as the culprit The timeline for this is ‘as fast as we can get all the platforms green’, but ideally with a clean nightly build in between each landing; this means we can save off those nightly installers and have a way to check ‘compositor without compositeViewer merged’ if the need arises (which hopefully it won’t) Does this seem safe / practical / reasonable? Kind regards, James |