7.0: Docker Tag Parsing Issue with Registry Ports

[bbezak]

Description:

We're facing a parsing issue in docker.py version 7.0, where registry server addresses with port numbers are incorrectly processed as part of the Docker tag, leading to an invalid tag format error.

Error:

it' happening in openstack kolla upstream CI:
docker.errors.DockerException: invalid tag 'primary:4000/lokolla/base:change_903165': invalid reference format

Possible issue:

The current regex introduced in commit: a9b5494
in docker.py does not accommodate port numbers in registry addresses:

_TAG = re.compile(
    r"^[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*(\/[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*)*" \
    + "(:[a-zA-Z0-9_][a-zA-Z0-9._-]{0,127})?$"
)

[6ax]

We faced the same problem. Apparently the regular expression should be like this:
r"^[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*(:[0-9]{2,5})?" \ + "(\/[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*)*$"

[artificial-intelligence]

well, looking at the OCI-Spec I think the new code is more correct/matches what OCI mandates.

a name must match [a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*(\/[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*)*
and a tag must match [a-zA-Z0-9_][a-zA-Z0-9._-]{0,127}

[artificial-intelligence]

:edit: this was wrong information, I misinterpreted the regex. still very weird.

[6ax]

It looks like the resulting regex doesn't include this:

// optionalPort matches an optional port-number including the port separator
// (e.g. ":80").
optionalPort = `(?::[0-9]+)?`

https://github.com/distribution/distribution/blob/97b1d649c4938d0f608d96454d6a8326b1f96acd/reference/regexp.go#L65

and the code should look like this? :

_TAG = re.compile(
    r"^[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*(?::[0-9]+)?(\/[a-z0-9]+((\.|_|__|-+)[a-z0-9]+)*)*" \
    + "(:[a-zA-Z0-9_][a-zA-Z0-9._-]{0,127})?$"
)