tests/integration: install libzstd-devel

[jeckersb]

Required to build zstd-sys - https://artifacts.osci.redhat.com/testing-farm/66cab0ec-1653-42ec-b482-a90001dbd65a/work-awskhuoovz0/plans/install-upgrade/aws/execute/data/guest/default-0/tests/integration/install-upgrade/rpm-build-1/output.txt

[jmarrero]

lgtm
however can you add to the description what this fixes or preps for?

[cgwalters]

This also relates to 0455f80 - in chatting with @henrywang I was hoping that we could rework the integration tests to always start from our hack/Containerfile to get closer to a single source of truth for the build starting point.

[jeckersb]

Hm well this fixes the zstd-sys thing, but now mock is failing and the logs are not particularly helpful - https://artifacts.osci.redhat.com/testing-farm/ab704601-09e0-4473-a945-c9dfbe5931b7/work-awsbf0m5vzu/plans/install-upgrade/aws/execute/data/guest/default-0/tests/integration/install-upgrade/rpm-build-1/output.txt

[cgwalters]

I think it's because the framework is passing --resultsdir to mock, but the test harness isn't picking up the files written there.

[cgwalters]

I'm not sure; maybe it's somehow a flake. I tried restarting the jobs.

[cgwalters]

It's also possible that this was actually broken by #581 - I think I ended up merging without those jobs being run because they were running for over 2 hours and we had a big TF backlog.

[jeckersb]

Ah I think we need to update the buildrequires from the spec in contrib/, testing that now locally to see if that's sufficient.

[jeckersb]

Ah I think we need to update the buildrequires from the spec in contrib/, testing that now locally to see if that's sufficient.

Seems to fix it for me locally, added a new commit with that, we'll see how it shakes out in CI

[cgwalters]

Ah right. Hmm...I might lean a bit towards always fetching the spec from fedora instead. This is a classic problem with specs/dpkg/arch etc in upstream, it just bitrots as it's not canonical.

(Actually though, in theory in Fedora we could make use of the "dynamic buildrequires" stuff to auto-determine from our source external dependencies we have, but not worth doing right now)

[cgwalters]

It's actually kind of tempting to change ostree-ext to also just support forking off /usr/bin/zstd, I don't have any problem with taking a hard dep on it...

[jeckersb]

Now it looks like quay issues:
Error: authenticating creds for "quay.io": received unexpected HTTP status: 504 Gateway Time-out

[cgwalters]

I guess let's give the other jobs a little bit more time to see if they get farther. But if not, we can just merge over red for those too.

[cgwalters]

Ooh no it looks like #581 did break more things:

[31mERROR[0m Installing to filesystem: Verifying fetch: Creating importer: Conflicting authentication options

[cgwalters]

ostreedev/ostree-rs-ext#639