Error accessing GCP artifacts when using special characters in task name

[materight]

Describe the bug

When uploading to GCP and using special characters in the task name, e.g. :, the artifact can't be retrived correctly. The returned artifact url is quoted, but GCP requires the url to be unquoted. Adding a call to urllib.parse.unquote(artifact._url) fixes the issue.

To reproduce

import urllib
from clearml import Task

task = Task.init(project_name='TEST', task_name='test:1')
task.upload_artifact('test_artifact', {'a': 1}, wait_on_upload=True)
artifact = task.artifacts['test_artifact']
#artifact._url = urllib.parse.unquote(artifact._url) # Fix
val = artifact.get(force_download=True)
print(val)

Expected behaviour

The artifact should be downloaded correctly. Without the fix, it tries to access: gs://cerrion-clearml/TEST/test%253A1.7c3dc2f40455456e88cf4c154edb965c/artifacts/test_artifact/test_artifact.json while the correct url would be without the %25: gs://cerrion-clearml/TEST/test%3A1.7c3dc2f40455456e88cf4c154edb965c/artifacts/test_artifact/test_artifact.json

Environment

• Server type: self hosted
• ClearML SDK Version: 1.11.0
• ClearML Server Version: WebApp: 1.11.0-373 • Server: 1.11.0-373 • API: 2.25
• Python Version: 3.10.10
• OS: Linux

[ainoam]

Thanks for reporting @materight.

Hope to have a fix out soon.

[alex-burlacu-clear-ml]

Hey @materight, we'll release a fix for this issue in the next few days, but it will require some adjustments on for the objects stored on the server. We'll follow up with more details

[materight]

Great! Thanks for the update

[materight]

Fixed in v1.12.0