httputil: work around NPE in signer.Add() #1999
Conversation
|
Can one of the admins verify this patch? |
glebsts
changed the title
|
@hdonnay Hello, could you please clarify about "conditionals being other way"? ? |
There was a problem hiding this comment.
Sorry for letting this languish.
I think the test's error message is backwards.
With that changed, this just needs a rebase and sign-off (git commit -s).
rebase steps
Just adding a note here in case you're unfamiliar with a rebase workflow.
Conceptually, you're just taking your commits and applying the diff to the current main instead of where they were originally committed.
Assuming origin is the quay/claircore repo, and fork is your fork:
git checkout BRANCH_NAMEgit fetch origingit rebase origin/maingit push --force-with-lease fork
If you get a merge conflict in the rebase step, just replace the conflict block with the code you think should be there. If something's wrong, we'll catch it. If it's for an autogenerated file like go.sum, just recreate it.
If we asked you to change a commit message, add the -i flag to the rebase command and use the reword command.
| t.Error("signer initialization with empty config should succeed") | ||
| } | ||
| if signer.use != nil { | ||
| t.Error("signed request authority map should be non-initialized") |
There was a problem hiding this comment.
| t.Error("signed request authority map should be non-initialized") | |
| t.Error("signed request authority map should be initialized") |
If
authsection is empty in config file, instance ofSigneris initialized with default field values:https://github.com/quay/clair/blob/v4.7.3/internal/httputil/signer.go#L29
Later e.g. in
reportcommand processing if configuration is used, methodAddis used as "marks the authority in uri as one that expects signed requests":https://github.com/quay/clair/blob/v4.7.3/cmd/clairctl/report.go#L145
In this method an entry is being added to
usefield ofsignerreceiver no matter what. But ifauthsection was missing in config,usehas default value of nil and producesassignment to entry in nil map.This patch fixes internal logic, but I decided to keep URI parsing before actually checking for nil to give faster feedback to user if uri is e.g. malformed. Feel free to suggest change in such a flow.
Method
Addis called from multiple commands, so I changed it inside the method, not pre-checking the calls.Other
Signerfields seem to be safe (ranging overnilmap is ok).There is no test file for signer,
so I did not create a new test caseI added one