podman-remote build: -f gets confused

[edsantiago]

The -f flag is looking for love Dockerfiles in all the wrong places:

$ bin/podman system service --timeout=0

In window 2:

$ mkdir ccc
$ echo "FROM scratch" >ccc/Dockerfile2
$ ./bin/podman-remote build -t foo -f Dockerfile2 ccc
ERRO[0000] 1 error occurred:
        * lstat /path/to/my/current/directory/Dockerfile2: no such file or directory

Error: Post: "http:...long stuff probably related to the previous failure so I'm not going to bother filing an issue for it"

Non-remote podman build works fine.

[rhatdan]

@vrothberg @nalind is there an easy way to add a single file to the copier of the context directory?

[rhatdan]

@edsantiago is finding all of the little warts in podman-remote build. :^).

I have been trying to kill them off as quickly as possible.

[vrothberg]

@vrothberg @nalind is there an easy way to add a single file to the copier of the context directory?

Yes, that's possible (see https://github.com/containers/podman/blob/master/cmd/podman/containers/cp.go#L362). Looking that the bindings for build, I think that nTar(...) should be replaced with using the copier package. It also supports Excludes among many other things.

[nalind]

The logic there is already checking for exclusions, and I think that slipping additional files into the tar stream (if we need to emulate something like the multiple-Dockerfiles feature) is going to be marginally easier with the current logic.

[edsantiago]

As an FYI: there are 52 tests in bud.bats that I'm currently skipping because they don't work with podman-remote due to this bug; all of them of the form buildah bud -f path. It seems like a common pattern.

[vrothberg]

I had a look at the code and it seems that the logic in cmd/podman/images/build.go needs some polishing. IMHO the data passed to the bindings should be fully evaluated since we must tar the files up. Using @edsantiago's reproducer above, we Dockerfile2 as the Dockerfile and /home/vrothberg/ccc as the context directory. I think the Dockerfile should have already been evaluated or at least be relative.

[vrothberg]

Ah, I have another idea.

[vrothberg]

Something for bug week. The longer I read the code, the more questions I have.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[edsantiago]

Still present. This is the biggest problem facing buildah-bud tests under podman-remote: there are 53 buildah-bud tests that need to be skipped because of this bug.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[edsantiago]

Still present: https://storage.googleapis.com/cirrus-ci-6707778565701632-fcae48/artifacts/containers/podman/6152338764201984/html/bud-remote-fedora-34-root-host.log.html#t-00205