Released on 2020-02-07.

Mock 2.0 highlights:

• The mock versioning policy (or rather style) has changed from three to two-number pattern. Don’t panic, this isn’t really special major release - the change was only done to move from the <UNUSED>.<MAJOR>.<MINOR> pattern to <MAJOR>.<MINOR>, so practically we went with v2.0 instead of previously planned v1.5.0.

• The --bootstrap-chroot option is newly enabled by default, this can be disabled by --no-bootstrap-chroot, or by config_opts['use_bootstrap'] = False. The content of bootstrap chroot is cached by default and never automatically updated, but one can use the new --scrub=bootstrap to remove related caches. The --scrub=all was updated to clean bootstrap as well (but --clean doesn’t touch bootstrap chroot at all).

• The use_bootstrap_container configuration option was renamed to use_bootstrap to better describe it’s purpose (it never implied usage of container technology) and to align with use_bootstrap_image option. Please migrate your custom configuration.

• The output from --debug-config option now only shows the differences from mock’s defaults, and the output doesn’t have the Jinja templates expanded.

• The config_opts['dnf.conf'] replaced config_opts['yum.conf']. Both still work, but only one of them can exist one config file.

• The --old-chroot and --new-chroot options were obsoleted by --isolation=chroot|nspawn, and still default to --isolation=nspawn. Please migrate your tooling.

• Mock now can now pre-configure DNF variables (#346), e.g. config_opts['dnf_vars'] = { 'stream': '8-stream' }

• The regression in --use-bootstrap-image implementation was fixed (did not work at all in v1.4.21), and should work reliably now (podman still needs to be installed manually to make it work).

• In mock config files we now prefer Jinja templates, instead of previously used python expansion "%(variable)" % ... It is not likely, but if you use this in your custom config files, please migrate.

Mock 2.0 other fixes and enhancements:

• Loop device files are pre-populated even in --isolation=nspawn chroots, similarly to what is done with --isolation=chroot (#298).

• The include() statement in mock config now also accepts relative path names (relative against config_opts['config_path'] for now).

• The host local repositories from mock config files (like baseurl=file://) are now correctly bind-mounted to bootstrap chroot. So installing RPM from such repositories with --bootstrap-chroot now works (related #381).

• Non-interactive commands in chroot are executed through systemd-nspawn --console=pipe (when --isolation=nspawn, default) (#432).

• Better detection of host’s package manager (DNF vs. YUM), for both bootstrap and normal chroot. This should demotivate people from using --dnf and --yum options (#233). More, on Fedora 31+ there’s no real YUM package manager anymore (there only is yum.rpm which actually provides /bin/yum symlink to /bin/dnf). This situation is now properly detected in mock, and the symlink is ignored (we fallback to DNF).

• Better re-using of DNF/YUM caches, in both normal and bootstrap chroot. This is mostly given by previous bullet (YUM vs. DNF detection). To be 100% sure, we also newly rather bind-mount both DNF and DNF cache directories into the chroot.

• Mock expands the config templates (aka include()) completely before executing it by eval(), and the implementation is now much simpler and clear.

• Mock doesn’t ignore cleanup_on_success configuration option after --postinstall action.

• mock --chain file descriptor leak was fixed, so the descriptor usage is constant with multiple builds.

• The Jinja templating is now iteratively re-rendered (when Jinja template expands to another Jinja template), till there is something to expand. Also we start the Jinja rendering mechanism a bit earlier in the codebase so the mock configuration isn’t really order-dependant (no matter which configuration option is set first).

• Fix lvm plugin volume removal feature on modern systems (rhbz#1762728).

• We don’t install shadow-utils (we don’t need this one) and distribution-gpg-keys (we copy the keys from host instead), so this makes the initial dnf_install_command transaction shorter, and more reliable across all the variety distributions we support.

• The --sources parameter is not mandatory in --buildsrpm mode.

• Mock now copies /etc/pkg/ca-trust/extracted into chroot (#397).

• The success and fail files are created under mockbuild user, not root.

• The compress_logs, when turned on, have predefined default gzip method.

• We turned --forcearch on long time ago, but mock exited with cryptic error when qemu-user-static wasn’t installed. Mock now detects that qemu-user-static is missing and throws instructions instead.

Mock-core-configs 32.0

• Added configs for Fedora 32, Fedora Rawhide configs moved to F33. The new package depends on updated distribution-gpg-keys 1.36 package (avaiable in Fedora updates at the time of release).

• Fedora 29 configs EOLed (moved below eol subdirectory).

• All the configuration files were modified to use templates, to de-duplicate a lot of stuff and many inconsistencies were fixed.

• On el7, mock/mock-core-configs automatically enable use_bootstrap_image option for Fedora 31+ chroots (ZSTD compression enabled for RPMs) because without this option it wouldn’t make sense to do anything (neither bootstrap chroot is installable).

Both mock and mock-core-configs packages need to be updated together as pair.

Following contributors contributed to this release:

• Dominik Tureček
• Jakub Čajka
• Jakub Kadlčík
• Merlin Mathesius
• Scott K Logan
• Sérgio M. Basto
• Silvie Chlupová
• Tomas Hrnciar

Thank you.