Hi 👋

When an error occurs during a multipart upload, BlobWriter.__exit__ is called with exception context. When that is the case, the multipart upload should be aborted. Currently, BlobWriter.__exit__ blindly calls BlobWriter.close, which in turn finishes (commits) the multipart upload. The user then ends up with a partial blob on gcs, whereas the multipart upload should have been aborted.

Steps to reproduce

• init a BlobWriter
• write some bytes that are to be uploaded
• raise an error that should result in aborting the upload

Code example

with blob.open('wb') as fp:
    fp.write(b'first chunk')  # not yet uploaded
    fp.write(b'big chunk' * 1024 ** 8)  # uploaded
    raise ValueError('SIGTERM received')

# the new blob should not have been created on gcs

For reference: smart_open.s3.MultiPartWriter.__exit__