- A flagfile requires manual management with touch and rm, replace it
with an environment variable that can be set more conveniently.
- Removing tools/tests/opgpcard.rs is not necessary, the file it links
to is always generated in build.rs, either with subplot tests or
without. But it is always there, so cargo test does not fail for
that reason.
- Some subplot tests require a virtual card. Those are not always
available on a system, so subplot uses the
tests/virtual-card-available file to note whether a virtual card is
available, and only generates those tests if it is.
- That file is included in the cache, so the cache should not be
shared.
- Subplot requires the presence of virtual cards, so it needs an image
where they are present. However this interferes with the intentions
of the cargo-test and cargo-test-debian-bookworm jobs.
- Revert the cargo-test and cargo-test-debian-bookworm jobs to the
environments they used to test, rust:latest and
debain:bookworm-slim, respectively.
- Add a test "subplot" that runs the subplot tests as part of the
virtual-test stage.
This doesn't change anything for existing tests but allows testing
opgpcard against a virtual smartcard later on.
Sponsored-by: NLnet Foundation; NGI Assure
- Forks of this project do not have access to cookiejar's hardware
keys. They do not have a gitlab CI runner that can pick up those
jobs, so the pipeline eventually fails with a timeout.
- Prevent hardware related jobs from being created for forked
projects.
The virtual card images used to do this kind of redirect. Now the consumer of the images needs to decide if they want to see virtual card debug output. We don't want to see it, by default.
