Changelog for libostree-1-1-2021.6-48.12.x86_64.rpm :

* Fri Dec 10 2021 Bjørn Lie - Update to version 2021.6: + Most of the fixes are related to warnings highlighted by gcc - fanalyzer static source analysis. + Performance of pruning logic has been improved, avoiding unnecessary trips through redundant serialization. + A regression has been fixed so that ostree is properly behaving again when used from the initramfs, at a point where /sysroot may not be mounted yet. + A race condition related to sysroot.readonly has been addressed by directly setting up sysroot readonly in initramfs.- Changes from version 2020.8 to 2021.5 please see upstreams list https://github.com/ostreedev/ostree/releases- Switch to obs_scm from tar_scm, and use obscpio instead of generated tarball. Also stop autogeneration of .changes, upstream now have proper release notes that should be used.- Use ldconfig_scriptlets macro for post(un) handling for shared library, modernize spec.
* Tue Dec 15 2020 Martin Liška - Enable LTO (boo#1133120) as it works now.
* Thu Nov 19 2020 dimstarAATTopensuse.org- Update to version 2020.8: + This release mostly contains scalability improvements and bugfixes. + Caching-related HTTP headers are now supported on summaries and signatures, so that they do not have to be re-downloaded if not changed in the meanwhile. + Summaries and delta have been reworked to allow more fine-grained fetching. + Finally, this fixes several bugs related to atomic variables, HTTP timeouts, and 32-bit architectures.- Changes from version 2020.7: + Static deltas can now be signed to more easily support offline verification. + There\'s now support for multiple initramfs images; the idea here is that one can have a \"main\" initramfs image and a secondary one which represents local configuration. + The documentation is now moved to https://ostreedev.github.io/ostree/ + Lot of preparatory cleanups to the pull code landed for upcoming work on indexing deltas outside of the summary. + On the bugfix side, the biggest one is a fix for an assertion failure when upgrading from systems before ostree supported devicetree. + Also notable is that ostree no longer hardlinks zero sized files to avoid hitting filesystem maximum link counts.- Changes from version 2020.6: + One notable feature: ostree now supports / and /boot being on the same filesystem. + Other than that it\'s mostly bugfixes; there is one quite important one for anyone using the readonly=true for /sysroot (which is still just Fedora CoreOS I suspect). + There\'s some improvements to the GObject Introspection metadata, some (cosmetic) static analyzer fixes, a fix for the immutable bit on s390x, dropping a deprecated bit in the systemd unit file, etc.- Changes from version 2020.5: + This release primarily fixes a regression in 2020.4 where the \"readonly sysroot\" changes incorrectly left the sysroot read-only on systems that started out with a read-only / (most of them, e.g. Fedora Silverblue/IoT at least). + There\'s some additions to the pull API to aid flatpak. + There were a few fixes to the man pages, and ostree show now displays the parent commit. + The default dracut config now enables reproducibility. + On the \"feature\" side, there is a new ostree admin unlock - -transient. We expect this to be a foundation for further support for \"live\" updates.- Changes from version 2020.4: + By far the biggest change in this release is new ed25519 signing support, powered by libsodium. + stree commit gained a new --base argument, which significantly simplifies constructing \"derived\" commits, particularly for systems using SELinux. + Handling of the read-only sysroot was reimplemented to run in the initramfs and be more reliable. Enabling the readonly=true flag in the repo config is recommended. + Several bugs were fixed in locking for the temporary \"staging\" directories OSTree creates, particularly on NFS. + lib: Coerce flags enums to GIR bitfields changed some values to be (correctly) flags - this may show up as incompatible for GObject Introspection consumers (but not C). + A new timestamp-check-from-rev option was added for pulls, which makes downgrade protection more reliable and will be used by Fedora CoreOS. + Several fixes and enhancements were made for \"collection\" pulls including a new --mirror option. + The ostree commit command learned a new --mode-ro-executables which enforces W^R semantics on all executables. + A new commit metadata key (OSTREE_COMMIT_META_KEY_ARCHITECTURE) was added to help standardize the architecture of the OSTree commit. This could be used on the client side for example to sanity-check that the commit matches the architecture of the machine before deploying.
* Thu Apr 30 2020 Dominique Leuenberger - Stop invalid usage of %_libexecdir: + Use %{_prefix}/lib where appropriate. + Use _systemdgeneratordir for the systemd-generators. + Define _dracutmodulesdir based on dracut.pc. Add BuildRequires(dracut) for this to work.
* Mon Apr 06 2020 alarrosaAATTsuse.com- Update to version 2020.3:
* A quick followup to 2020.2, which introduced support for read-only sysroot ended up breaking some of the Fedora CoreOS tests in coreos-assembler which in turn holds back ostree going into FCOS
* Now that gap has been closed and more of those tests are being run on the new CI.- Update to version 2020.2:
* lib: Fix Since versions for 2020.1
* Post-release version bump
* \"Brown paper bag\" release that actually sets the is_release_build=yes flag and also fixes the Since: on a few new functions.- Update to version 2020.1:
* There is now support for making the /sysroot mount point read-only to start. This protects against a lot of accidental damage, and also generalizes and improves the previous special case handling of having /boot read-only. One known issue is that ostree pull is broken with this enabled, and this will be fixed.
* Error-handling around GPG verification has had an overhaul. Specifically, libostree now has more specific error codes to distinguish between different verification failures. This should allow apps to have more fine-grained control over how to respond to errors. Do note that the error messages themselves have changed, and we strongly suggest that anyone relying on a specific error message string to migrate to using the API directly.
* The original \"archive\" (split up objects) format didn\'t make it easy for a client system to know how much data it would be downloading. Later, static deltas were added which addressed this problem, but there are situations in which object fetches still occur. Later then support for optional sizes metadata in commit objects was added but was never really stabilized/publicized. There were also some bugs in it. That is now completed - the sizes data is now stable. and new API was added to read it.
* This release adds initial fs-verity support; it doesn\'t do too much today. Bigger picture it\'s important to understand that the vision of OSTree is to enable Linux systems that feel like they\'re \"image based\" (transactional, versioned updates, no dependency resolution client side), but also to enable things like doing commits on the client side. Today rpm-ostree supports replacing the kernel client side as a first class operation. This is crucially important to make it feel truly like a Linux system that you own.
* A small tweak was made to have OSTree create repo structure directories and files (such as objects/ or .lock) with group write permissions. This is useful for managing OSTree remote servers from multiple UIDs. For systems with the default umask of 0022, this should have no effect.
* We\'ve extensively reworked CI for the upstream repo. In addition to Travis, testing is now done on top of Fedora CoreOS. Not all tests have been carried over, but expect to see more coming. This rework will also allow us to have more comprehensive tests previously not possible.
* Several fixes were made to the test suite to handle the cases of systemd vs no-systemd, and systemd is now advertised in the list of features in ostree --version if present.
* Tue Jan 14 2020 Antonio Larrosa - Update to version 2019.6:
* Nothing major in this release, but there is some bigger stuff outstanding and ready to merge, so this version was released so that work will have time to stabilize.
* A few build/CI fixes. A new progress API which will be used by flatpak (and can be used by others). Finally, we also avoid reordering kernel arguments.- Update to version 2019.5:
* We discovered that CLang has a static analyzer scan-build; it found some small memory leaks so far, otherwise mostly noise, but we haven\'t dug through all the errors yet.
* Gained a new zipl (s390x bootloader) backend
* Install the .hmac files needed for FIPS mode in /boot too
* This is also the first release where we switched to using the OpenShift Prow as a merge bot, though a lot more CI work is pending.- Update to version 2019.4:
* This is mostly a bugfix release. Notably, the 2019.3 release caused some issues related to the gpg-agent code spewing messages on the terminal. Additionally, Fedora 31 users have hit upon issues with ostree-finalize-staged.service running too late to be able to write back its logs to the journal. This then confused rpm-ostree after reboot, because it looks at the previous boot\'s journal for this message.
* The biggest feature-ish change is support for a partial commit \"reason\" so that after ostree fsck --delete was used, subsequent ostree fsck will continue to report an error. This should be used by higher level tools that want to do \"fsck and repair\". It\'s likely at some point that \"fsck and repair\" logic will move down into the libostree core as well.
* There are ongoing efforts to port Fedora CoreOS to s390x: one fix landed here to add the deployment prefix to BLS entries since it\'s what the zipl bootloader expected.- Update to version 2019.3:
* A lot of changes since the last release. On the feature side, probably the biggest is we\'ve made public the internal API for kernel arguments, which rpm-ostree now uses.
* Other things include a new --modern switch for init-fs, better support in pull for downgrade protection, better use of mmap, support for committing archives (tarballs) from stdin, etc.
* Finally, libostree now supports being built without GPG, which is an important preparatory piece for introducing an alternative signature system.- Update to version 2019.2: + It\'s been some time since the last release, so this is a slightly larger one! There\'s lots of new features, and a few bug fixes. + New features:
* A new sysroot.bootloader key was added to be more explicit about which bootloader OSTree should use. Notably a none value is supported, in which OSTree is solely responsible for writing the BLS entries. This can then be used by bootloaders like GRUB2, which now supports BLS natively.
* ostree config now supports the unset command to unset a key from the OSTree repo config
* ostree remote add now supports the --force flag to replace a remote of the same name if it exists
* ostree-prepare-root now logs a structured journal message after finding the deployment to which to pivot. This can be used by higher-level apps like RPM-OSTree to build a history of the deployments the machine was booted into.
* The staging API now supports a lockfile which prevents finalization at shutdown. This is intended to be used in systems that need more fine-tuned control between staging a deployment, and setting it as the default deployment on reboot.
* ostree static-delta show now prints the From and To commits for which the delta was generated + Bug fixes:
* Make looking up collection-refs similar to how regular refs are looked up, i.e. first in the transaction, then in the current repo, and then in the parent repo
* Don\'t include the OSTree commit version number twice in the boot menu title. This affected at least Fedora-based systems composed with RPM-OSTree\'s mutate-os-release.
* Activate ostree-finalize-staged earlier; this should hopefully make deployment finalization more reliable by running it later in the shutdown sequence, when less services are running
* Run grub2-mkconfig on the filesystem tree of the pending deployment, rather than the previously deployed tree. This was a corner case where the deployment failed if a previous deployment did not exist, on systems where grub.cfg is used.
* Wed Apr 24 2019 Martin Liška - Disable LTO (boo#1133120).
* Mon Jan 28 2019 bjorn.lieAATTgmail.com- Update to version 2019.1: + This is the first libostree release of 2019; no big changes, just a collection of smaller features and bugfixes. + On the features side, a good example is: lib/repo: Search a list of paths in gpgkeypath for gpg keys. + Another feature is grub2: add support for devicetree. + lib/kargs: allow empty-list arguments i.e it ensures libostree supports \"empty list\" kernel arguments. + There\'s also some ongoing work to have libostree be a \"backend\" for OCI/Docker container storage; checkout: honor opaque checkouts. + If built with --disable-http2, allow enabling via http2=1 will allow people to more easily play with HTTP2 if it\'s disabled by default.
* Wed Nov 14 2018 Antonio Larrosa - Update to version 2018.9: + New features:
* Allow disabling pulling from LAN/USB/Internet
* lib/repo: Add an API to get min-free-space-
* reserved bytes
* OstreeMutableTree: add _remove method
* repo: Add a checkout option to not hardlink zero-sized files + Bugfixes (apart from regular memory leaks fixes):
* finalize-staged: Bump timeout to 5 minutes
* deploy: Fix removing /var/.updated with separate /var mount
* src/ostree: Don\'t delete refs having aliases + One notable change in this release is the initrd service ostree-prepare-root.service now runs earlier in the boot process. This shouldn\'t actually affect OSes, unless there\'s extended logic in the initrd that integrates tightly with OSTree. + Another systemd related change is the introduction of a path unit: ostree-finalize-staged.path. This allows the service of the same name to be path activated instead of explicitly started at deployment staging time. This release however does not yet rely on this mechanism to give time for packagers and integrators to adapt to the new unit (e.g. by enabling it in the systemd presets). A future release will require this. Note that deployment staging is still not the default for libostree, although at least rpm-ostree now unconditionally uses staging, and while it generally worked well, we hit issues with people using slower hard drives, hence the increase in timeout in PR #1755 . + Another change to call out is: lib/commit: Don\'t chown objects to repo target owner. We previously had incomplete support for a process running as uid 0 writing to a repository owned by a non-zero uid, but it was never finished. This will likely be revisited at a later time.
* Mon Aug 27 2018 bjorn.lieAATTgmail.com- Update to version 2018.8: + This release is pretty much all minor bugfixes: memory leaks, fixing error messages and docs, handling a race condition on pull with summary updates. There\'s one new feature (noted below), and we also gained a new contributing tutorial: https://mail.gnome.org/archives/ostree-list/2018-August/msg00005.html + The one bugfix I want to call out explicitly is: ostree-remount.service: RemainAfterExit=yes (gh#ostreedev/ostree#1697). It\'s surprising it took us so long to find and fix this; I\'ve seen occasional boot failures that I believe trace down to this problem. The behavior of systemd units of Type=simple without RemainAfterExit=yes set is rather nonsensical; I may try to push to have a warning emitted upstream if such a unit is a dependency of another. + And the one new feature is the auto-update-summary config option for repositories. For more information, see the docs and gh#ostreedev/ostree#1681.- Rebase ostree-grub2-location.patch with quilt.- Drop libostree-fix-wformat-warnings-on-i586.patch: Fixed upstream.