I replaced the eprintln!s with println!, but it didn’t make any difference. Maybe we should remove those statements completely.

I can’t really find the place that kills the kernel. The last command to run is xvc file list:

print(xvc_test_data.file().list("test-data/dir-0002"))

and the output it produces is:

[src/output.rs:144:13] &output_str = "SS         131 2024-06-05 08:57:11 41e16be7          test-data/dir-0002/file-0003.bin\nSS         131 2024-06-05 08:57:11 27f0
efd0          test-data/dir-0002/file-0002.bin\nSS         131 2024-06-05 08:57:11 66de5084          test-data/dir-0002/file-0001.bin\nTotal #: 3 Workspace Size:
      393 Cached Size:        6006\n"

print may be causing the crash, but the more likely cause is the command that comes after this:

!ls -l test-data/dir-0001/

I replaced this with lsd, which also failed. Maybe it’s actually a Python crash or bug.

The way to understand is to create a notebook file with only that cell and try to run it.

The ls line runs fine with a new notebook. It even runs on the README file when run at the beginning. The line that makes the kernel crash is:

xvc_test_data.storage().new_s3(name="backup", bucket_name="xvc-test", region="eu-central-1", storage_prefix="xvc-storage")

We can start by removing the new_s3 part.

The storage() method runs fine. It returns an XvcStorage() object, as it should.

When I run storage().list(), it takes a very long time. The bug is likely related to storage().

It looks like the storage object was adding file instead of storage as a subcommand. I’ve fixed it now.

That was the bug. The README notebook now creates the S3 storage.

What was the reason behind this?

Parsing the CLI to the XvcCLI object was perhaps the culprit. Let’s look at it more clearly.

Let’s try xvc file new s3 as a command to see how it behaves.

It says unrecognized subcommand for new.

This is how it should be, but I wonder why it doesn’t work for the XvcCLI parser.

Anyway, it’s already 13:00, so let’s stop here for today.

export AWS_ACCESS_KEY_ID="XXXX"
export AWS_SECRET_ACCESS_KEY="YYYY"

you can install the AWS CLI with:

pip3 install --user awscli

and create an S3 bucket from the command line without using a web browser:

aws s3api create-bucket --acl public-read --bucket my-unique-bucket-name --region eu-central-1 --create-bucket-configuration LocationConstraint=eu-central-1

This will create a publicly readable bucket at http://my-unique-bucket-name.s3.amazonaws.com.

Then you can copy your files with:

aws s3 cp my-file.zip s3://my-unique-bucket-name/subdir/my-file.zip

and share the link http://my-unique-bucket-name.s3.amazonaws.com/subdir/my-file.zip with anyone you want.