Closed StefanScherer closed 1 year ago
Should this fix the CI?
Yes, I hope so. CI is only triggered in the upstream repo, but not in forks, that's why we don't see a status here. This PR should fix the errors we can see in the "Release and Weekly Build" workflow in https://github.com/docker/scan-cli-plugin/actions. Even if I have created this PR directly in upstream repo, we couldn't run it there, only the build-pr workflow is triggered for PR's and that runs only on Linux nodes, but not on Windows. I suggest we merge it and #204 and then see if that can build the missing linux/arm64 binary.
Yes, CI is green again -> https://github.com/docker/scan-cli-plugin/actions/runs/2789915771
So we don’t have CI on the PRs? 😞
Good catch, yes. Looking at the comments in #180 I think we should just skip the e2e test in PR's to avoid providing the env variables in PR builds from random forks.
Signed-off-by: Stefan Scherer stefan.scherer@docker.com
- What I did
Make the
PWD
variable on Windows a path with forward slashes (again) instead of backslashes. I saw in older workflow runs that the path used forward slashes, then started failing running the e2e tests on Windows because of it having backslashes.I was able to reproduce locally with MinGW64 11.2.0 installed:
- How I did it
Add a
subst
to replace all\
with/
. Surprisingly in the Makefile I didn't had to use things like double\\
backslash, just a single one is fine here.- How to verify it
mingw32-make -f builder.Makefile e2e
that failed for me without the substitution. With this fix it only complains about missing E2E vars, but it came across the path- Description for the changelog
- A picture of a cute animal (not mandatory)