BuildKit can't COPY paths with non-ASCII characters

[felixkrull-neuland]

When trying to build a Dockerfile using BuildKit that contains paths with non-ASCII characters in COPY commands, I get the following error message

ERROR: failed to solve: Internal: rpc error: code = Internal desc = rpc error: code = Internal desc = header key "followpaths" contains value with non-printable ASCII characters

I was able to reproduce this on Ubuntu 22.04 using Docker Engine 24.0.2 installed from the docker.com APT repository:

vagrant@ubuntu-22-04:~/repro-non-ascii$ ls -lh
total 8.0K
-rw-rw-r-- 1 vagrant vagrant 43 Jun  2 10:42 Dockerfile
-rw-rw-r-- 1 vagrant vagrant  5 Jun  2 10:42 test-äöü.txt
vagrant@ubuntu-22-04:~/repro-non-ascii$ cat test-äöü.txt 
test
vagrant@ubuntu-22-04:~/repro-non-ascii$ cat Dockerfile 
FROM alpine:latest
COPY test-äöü.txt /

Building with BuildKit fails:

vagrant@ubuntu-22-04:~/repro-non-ascii$ sudo docker build .
[+] Building 1.1s (5/6)                                                         
 => [internal] load build definition from Dockerfile                       0.0s
 => => transferring dockerfile: 80B                                        0.0s
 => [internal] load .dockerignore                                          0.0s
 => => transferring context: 2B                                            0.0s
 => [internal] load metadata for docker.io/library/alpine:latest           1.1s
 => ERROR [internal] load build context                                    0.0s
 => => transferring context:                                               0.0s
 => CACHED [1/2] FROM docker.io/library/alpine:latest@sha256:02bb6f428431  0.0s
------
 > [internal] load build context:
------
ERROR: failed to solve: Internal: rpc error: code = Internal desc = rpc error: code = Internal desc = header key "followpaths" contains value with non-printable ASCII characters

Building without Buildkit works:

vagrant@ubuntu-22-04:~/repro-non-ascii$ sudo DOCKER_BUILDKIT=0 docker build .
DEPRECATED: The legacy builder is deprecated and will be removed in a future release.
            BuildKit is currently disabled; enable it by removing the DOCKER_BUILDKIT=0
            environment-variable.

Sending build context to Docker daemon  4.096kB
Step 1/2 : FROM alpine:latest
latest: Pulling from library/alpine
8a49fdb3b6a5: Already exists 
Digest: sha256:02bb6f428431fbc2809c5d1b41eab5a68350194fb508869a33cb1af4444c9b11
Status: Downloaded newer image for alpine:latest
 ---> 5e2b554c1c45
Step 2/2 : COPY test-äöü.txt /
 ---> 25b30012519d
Successfully built 25b30012519d

---

I sincerely hope this is a regression and not intended behaviour. This would be a severe limitation.

[jedevc]

Hm, this definitely needs further investigation. I think we should definitely be supporting this, but I couldn't find a test for this kind of behavior 👀

I worked on a quick test for this: f9f2a00, I'll try and take a bit of a deeper dive into it when I have some time, though if anyone else is interested, feel free to pick it up 🎉