cp support

[ibuildthecloud]

FEATURE REQUEST

mount/umount is more powerful than cp, but just as buildah is more powerful than a Dockerfile, you still support a Dockerfile. cp is such a common task it would be nice if it was supported.

[mheon]

I can see making a cp command that mounts (if unmounted), copies, and then unmounts (if you initially did a mount) under the hood. We have the code to do all three, combining them shouldn't be that difficult.

[mheon]

If you were thinking of a more Docker-style copy via stdin, I'm not specifically opposed to that, but the simpler approach would be to reuse the existing mount-based code.

[rhatdan]

@ibuildthecloud so you want podman cp to work.
I have no problem implementing this my just execing the CP command, but putting the effort/support into actually doing cp seems a lot of work. And docker cp is one of the things I tend to show as a reason why podman/buildah mount is better.

[rhatdan]

@ibuildthecloud Any comment on our suggestions? Or should I close this?

[mheon]

@rhatdan I think this is still a good enhancement, we should leave it open

[vrothberg]

I can pick that up if you want to, and would follow @mheon's inital proposal to map cp to mount, cp, unmount. Are you okay with that?

[rhatdan]

Im fine with it. My only concern is handling multiple mounts/unmounts and making sure we handle this correctly.

[mheon]

I think mount/unmount session-based concurrency would need to happen first

[vrothberg]

My only concern is handling multiple mounts/unmounts and making sure we handle this correctly.
[...]
I think mount/unmount session-based concurrency would need to happen first

@mheon, @rhatdan, I think we handle this now correctly, do we? I think the internal counter solves most of it. But I am not sure if multiple unmounts should break the cp or not?

[mheon]

I think the internal counter does solve this, as long as each mount only unmounts once.

…

On Mon, Nov 19, 2018, 06:55 Valentin Rothberg ***@***.*** wrote: My only concern is handling multiple mounts/unmounts and making sure we handle this correctly. [...] I think mount/unmount session-based concurrency would need to happen first @mheon <https://github.com/mheon>, @rhatdan <https://github.com/rhatdan>, I think we handle this now correctly, do we? I think the internal counter solves most of it. But I am not sure if multiple unmounts should break the cp or not? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#613 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHYHCF7rNj-WceIZWwOE7wTznOVke21Yks5uwpwygaJpZM4TSnED> .

[rhatdan]

I agree.

[vrothberg]

Great, this seems ready to be implemented 👍

[ktdreyer]

I was using podman today, and I was sad about the lack of "cp" parity with docker. Now I'm happy to find this issue.

[rhatdan]

Would be a simple bash script to generate the equivalent. (I am not a big fan of docker cp). But I guess we will have to implement it.

mnt=$(podman mount $ctr)
cp $mnt/var/lib/foobar /tmp
cp /etc/mydata.conf $mnt/etc
podman umount $ctr

[rhatdan]

@QiWang19 Could you look into adding this command.

[mheon]

I think this is definitely more urgent now that podman mount is broken for rootless