update meson

[jepio]

update meson

Update meson to a current version, which is needed when introducing clevis ebuilds.

How to use

./build_packages && ./build_image

Testing done

CI running: http://jenkins.infra.kinvolk.io:8080/job/os/job/manifest/4810/cldsv/

• Changelog entries added in the respective changelog/ directory (user-facing change, bug fix, security fix, update)

[dongsupark]

Looks good if CI passed.

Can you please also delete rng-tools from https://github.com/flatcar-linux/coreos-overlay/blob/233d43aa144dfa508739d4b31644478919591cef/profiles/coreos/arm64/package.accept_keywords#L57 ?

[tormath1]

Mm, polkit build is failing because of this I think: https://mesonbuild.com/Release-notes-for-0-60-0.html#unknown-options-are-now-always-fatal.
I think we could replace the useless -Dwith-duktape=yes by this option: https://gitlab.freedesktop.org/polkit/polkit/-/blob/master/meson_options.txt#L19. It was correctly built previously because the option seems to default to duktape.

EDIT: build has been retriggered with this commit: flatcar-archive/coreos-overlay@7e303cc

[jepio]

Thanks @tormath1 , can you submit the duktape fix as a PR? The remaining issue is that now the rngd service exits with error when hwrng is not available, so i'll adjust the ExecStart line to tolerate failure.

[pothos]

Is rngd now starting by default or still only in the initramfs? We had a discussion in in flatcar/Flatcar#207 about that. I think it won't hurt to try to run it and handle it gracefully. Another idea was to only start it based on seeing what /dev/random is backed by /sys/class/misc/hw_random/rng_* and whether rngd adds value or not but that is maybe a bit complex.

[jepio]

Yes, we automatically enable rngd but only for the initramfs (https://github.com/dracutdevs/dracut/blob/master/modules.d/06rngd/module-setup.sh). /sys/class/misc/hw_random/rng_* is related to /dev/hwrng but these don't get fed into /dev/random automatically, that's what rngd does. rngd support multiple backends and on x86 it uses the rdrand instruction. The new version also can checks whether /dev/hwrng has an backend before using.

[pothos]

Ok, should we enable it by default then in general instead of only the initramfs?

[JAORMX]

uhm... isn't it the case that rngd is no longer needed? At least it doesn't seem to be shipping anymore in Fedora [1] [2], and RHEL [3]

[pothos]

uhm... isn't it the case that rngd is no longer needed? At least it doesn't seem to be shipping anymore in Fedora [1] [2], and RHEL [3]

With the new development I hope yes, it won't be needed anymore: https://lwn.net/Articles/884875/

[jepio]

@JAORMX might be right - I will test removing rngd altogether. There were many changes to kernel rng other the year,s most of which have the goal of removing blocking.

[jepio]

Dropped the rng-tools part because we removed it. Rerunning CI.

[tormath1]

let's wait for the CI but it's a "go" for me.

[dongsupark]

Looks good if it has changelog as well.

EDIT: ah, no changelog is needed, as it is just meson.