Rebase on 2022.1

[dbnicholson]

This is a rebase on top of 2022.1. Here's a list of our commits on master relative to our previous 2020.8:

• 7610f0c repo: Delete compat files with commit
• e727966 tests: Add tests for Endless compat files

I think it's time to cut the cord on these two. The last time the ostree client would pull these legacy files was in EOS 2.3(!). Even if someone followed the manual EOS 2 to 3 upgrade, the chances that one of the commits that was pulled from that time hasn't been pruned away is very low. I would say the systems this affects is 0.

• 6607b65 lib/repo-pull: Retry pulls without static deltas if they fail

See ostreedev/ostree#1612. I think the last comment from jlebon makes sense but I haven't looked into the feasability. Will file a task to upstream this.

• 1319537 grub2: Disable use of grub-mkconfig on Endless

This is really unfortunate. The upstream way to handle this is to specify the bootloader in the repo config like so:

[sysroot]
booloader=none

Maybe we should start having the image builder setup repos that way and add another ugly eos-boot-helper script to set it on existing repos so we can someday rely on it. Will file a task to handle that. The other part of reliably saying bootloader=none is that we could stop trimming the ostree-boot package.

• 052f2fd deploy: Create fake symlinks when on FAT filesystems
• 09b24be deploy: Handle efi blobs

Ugly hacks to make ostree work when /boot is a FAT filesystem like when it's the UEFI ESP on PAYG. GNOME OS developer has a nicer solution in ostreedev/ostree#1967, but I don't believe it's ready yet. Will file a downstream task to clean this up.

• 090932d Skip test-pull-repeated during CI

From debian/patches. We could probably drop this as I've never seen this failure locally or in OBS. From other discussions it sounds like debian's s390x builder has other problems. Leaving it for now.

• ef2d75c pull: Allow disabling commit binding verification
• 1c04dd0 pull: Use GNU coding style
• 5d130e7 tests: Ensure no dangling commit partials on remote depth pull
• bec8ceb pull: Error on depth pull with missing head commit
• 514dbad pull: Fix local pull with depth and truncated source history
• 9533a18 Fix translation of file:// URIs into paths
• 528e32c test-pull-summary-sigs: Set timestamps to serve expected files

Upstream backports.

• 7ce86b9 deploy: Don't try to create symlinks on FAT /boot

Another FAT hack. See above.

• a8b52f6 repo: Make locking APIs public
• 19f8684 repo: Require lock type in ostree_repo_lock_pop
• 62b592e build-sys: Bump required GLib to 2.44
• 4204a7b repo: Make locking per-OstreeRepo
• 46e74d3 repo: Make locking precondition failures fatal
• 7271863 test-concurrency: Lower lock timeout
• 2c228c8 tests: Add single process repo locking tests

Upstream backports.

https://phabricator.endlessm.com/T32964

[dbnicholson]

The post-rebase queue can more clearly be viewed in Version_2022.1...T32964-rebase-2022.1.

[dbnicholson]

retest this please

[pwithnall]

LGTM.

From debian/patches. We could probably drop this as I've never seen this failure locally or in OBS. From other discussions it sounds like debian's s390x builder has other problems. Leaving it for now.

I think it makes sense to carry all the debian/patches even if, like this one, we’re fairly sure they’re not relevant, as diverging from Debian potentially makes future debian/patches harder to apply.

[dbnicholson]

retest this please

[dbnicholson]

retest this please