Hi all, Yesterday afternoon I actioned the rollout of VM based CI, which is now in production for all platforms.
As part of this, Snap builds are now generally available and may be used freely. Publishing of Snaps still requires a notary to be built, so that component remains unavailable for now. The previous dedicated VM providing support for Snaps has also been retired. FreeBSD has also updated to Qt 6.9 as part of this changeover, and all other images (SUSE Qt 5.15, Qt 6.9 and Qt 6.10, Alpine Qt 6.8) have also been rebuilt and are fully updated as well. Builder wise, 4 of our previous 6 builders have been converted to be VM based. The remaining two are scheduled for retirement to allow for an ARM builder to be provided, and will be temporarily available for a further week or so to allow for old pipelines to be completed. The ARM builder has also been provisioned and connected as a VM based runner to Gitlab. Should your project be Qt 5 based still, you will find that all support for everything except Linux has been removed. This is in line with what was previously announced and you will need to remove those jobs from your CI configuration. All build artifacts relating to Qt 5 on FreeBSD and Windows have already been purged from the system, as support for those platforms has now ended. If your project has custom jobs, it would be advisable to check that those jobs are making use of VM based CI where possible, as build power on the Docker side following this conversion will be more limited (being primarily intended to support building websites and running linter checks). Should you be running custom workflows it is important to note that the location for caching artifacts has changed as part of this migration. It is no longer at /mnt/artifacts/$PLATFORM/ or /mnt/caches/$PLATFORM/ - instead things are now found at /mnt/$PLATFORM/artifacts/ and /mnt/$PLATFORM/caches/. It is imperative that this change is reflected in your jobs, otherwise you may encounter permission related failures due to different distributions having different user/group IDs. For those curious, the rollout procedure across all four machines took less than an hour to process (including the Hetzner re-imaging of the machines), and is largely fully automated. There is now also automation in place to clean up old images as well, which will hopefully reduce the risk that the builders run out of disk space. Please let me know if you have any questions. Many thanks, Ben
