postgresql: 9.5.20 -> 9.5.21, 9.6.16 -> 9.6.17, 10.11 -> 10.12, 11.6 -> 11.7, 12.1 -> 12.2 #80328
Conversation
|
I think we should not treat those as mass rebuilds, so no need to base on I recall someone told me it is fine to merge pg to master. But let's check that. I took as an example one of previous CVE pg updates: #63601 Looking at Hydra, I found the evaluation which picked up my commit: https://hydra.nixos.org/eval/1526884?filter=&compare=1526814&full=#tabs-still-succeed Hydra eval had revision 8634c3b. We can then lookup when was 19.03 channel switched to that revision using an awesome @grahamc service https://channels.nix.gsc.io/nixos-19.03/history-v2 Indeed, there is a line Format is "commit-hash date-of-commit date-of-advancement". We don't care about date-of-commit, because it is not date of git push, but date-of-advancement is one when all builds were finished and channel advanced. So, comparing it to my backport push date, I conclude it took ~12 hours to rebuild all the Qt stuff which depends on postgresql (modulo non-adjusted timezone errors). Is 12 hours a mass-rebuild? I have vague memories that threshold is 3 days, but cc @vcunat as knowledge bearer and potentially interested (#19045) |
Forward-picked from merge 036c524 in staging to master, as upstream claims it fixes CVE-2020-1720. https://www.postgresql.org/about/news/2011/
|
My understanding is that there's a security issue fixed and Hydra's currently not much loaded, so in that light the rebuild certainly doesn't seem large, and I also picked this to master in 329102c. |
|
@vcunat how many hours is "large"? Do we have ETA on channel advancement? Also, are there historical graphs of "hydra much loaded"? |
|
IIRC @grahamc has shown some graphs, but I can't find any URL now (except for OfBorg). There's certainly no hard limit what's large – for me it's mainly intuitive estimate; the counts of rebuilds also aren't perfect in the sense that they don't easily translate into e.g. those "hours" of Hydra work. It also depends on urgency of the change and its likelihood to break things – mass rebuilds sometimes get slowed down not just due to rebuilds but also due to mass breakages. |
|
@vcunat thanks. Hadn't thought about mass breakages. |
|
I found the Hydra graphs: https://status.nixos.org/grafana/?orgId=1 |
Motivation for this change
https://www.postgresql.org/about/news/2011/
Things done
sandboxinnix.confon non-NixOS linux)nix-shell -p nixpkgs-review --run "nixpkgs-review wip"./result/bin/)nix path-info -Sbefore and after)