These now regularly run upstream through packit, and thus are reasonably
well gated.
We unfortunately can't switch from STI to FMF completely, as in RHEL the
tests are part of the container-tools module, which uses STI. But the
remaining two ansible files change very rarely, getting rid of the
browser.sh and runt-est.sh duplication gives us most of the benefit
already.
Revert the downgrade from commit
d313d525e8, and install the full chromium
package instead. This has a lot more dependencies, but is not affected
by the Ctrl+A keyDown crash.
It used to be a dependency of cockpit-podman, but it is still needed on the
first host where cockpit is being accessed - which is the case how
gating is accessing cockpit.
Only set `$SOURCE` to ./source/ if that actually exists, which is when
it is being called through Standard Test Interface.
Provide fallbacks for using FMF [1] or when calling the script manually.
[1] https://github.com/cockpit-project/cockpit-podman/pull/649
In the presence of a package-lock.json, `npm install` apparently
downloads all packages even when giving explicit ones on the CLI. Avoid
that by removing package-lock.json before.
So that it can be called from the container-tools module without
duplication. For technical reasons this requires calling the test entry
script from a different directory, so parameterize it.
Rename the test from "browser" to "cockpit-podman" so that it is
more descriptive within a module test.
Drop the TEST_OS hack, tests of current version recognizes RHEL 8.3 just
fine (and in fact depends on it).
Related: rhbz#1821193
Re-use the test skeleton from cockpit. Just run a single integration
test for now, until we make them properly @nondestructive upstream [1].
Run chromium and the test directly on the test bed. For cockpit-podman
tests it's not practical to run them in a cockpit/tasks podman
container, as that container interferes too much with the tests.
[1] https://github.com/cockpit-project/cockpit-podman/pull/356
