-
Notifications
You must be signed in to change notification settings - Fork 16
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add functionality for verifying tarball signatures to automated ingestion scripts #208
Conversation
The staging PR for the test tarball was opened: https://github.com/EESSI/staging/pull/2544. Before merging that PR, I've manually changed the signature of the metadata file to verify that the ingestion would fail, and it did:
Right now it only prints this in the log file, and continues. It should also open an issue in the staging repo, I'll have to add that. |
It now opens an issue if the verification fails, see https://github.com/EESSI/staging/issues/2545. |
Another test: I've uploaded a tarball and metadata file without signatures, and set |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good.
- A few suggestions for naming attributes/variables. (if done would require renaming across the whole script.)
- A tiny typo.
- Question about logic when to return False (two places).
(self.object, self.local_path, self.object_sig, self.local_sig_path), | ||
(self.metadata_file, self.local_metadata_path, self.metadata_sig_file, self.local_metadata_sig_path), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe worth to rename elements to clearly notify what they refer to and have the naming consistent? For example,
(self.object_remote_path, self.object_local_path, self.object_sig_remote_path, self.object_sig_local_path),
(self.metadata_remote_path, self.metadata_local_path, self.metadata_sig_remote_path, self.metadata_sig_local_path),
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This originated from the object terminology used by the boto client, but I agree that it would be more clear to use the names like you proposed. Is it okay to do that in a follow-up PR, as it would require changes across the entire script?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, perfectly fine to do it in a follow-up PR.
(self.metadata_file, self.local_metadata_path, self.metadata_sig_file, self.local_metadata_sig_path), | ||
] | ||
skip = False | ||
for (object, local_file, sig_object, local_sig_file) in files: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Possibly synchronise naming with potential changes in files
array, e.g.,
for (remote_path, local_path, sig_remote_path, sig_local_path) in files:
Co-authored-by: Thomas Röblitz <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
Still testing it, but first tests look good. I'm running a separate instance/fork of the automated ingestion on the Stratum 0, configured to use a test bucket and non-existing target repo. This successfully picked up the tarball+metadata file and their signatures from the test bucket (uploaded by @trz42 in EESSI/software-layer#968):
The log shows that the signatures were successfully verified:
We should do the same for a non-valid signature (and make sure they fail), and for tarballs without a signature file (this should succeed if
signatures_required = no
).