sundapeng commented on PR #10178: URL: https://github.com/apache/paimon/pull/10178#issuecomment-6128881059
While fixing the presign finding I found the root cause was structural: `RESTTokenFileIO` still refreshed the token with its own logic next to `RESTTokenRefresher`, so the wrapper's token and the signing token could diverge, and the two had already drifted in how they assemble a token and when they reload. 4507284 makes `RESTTokenFileIO` load its token through `RESTTokenRefresher`, so a table's token follows one policy. This also changes three existing behaviors of the wrapper: a token living an hour or less is no longer reloaded, with a new delegate, on every call; a reload no longer blocks other callers; and a failed reload no longer fails the call while the token is still valid. `testRefreshFileIOWhenExpired` relied on the first one and now moves the clock instead. I left out sharing one refresher per table and catalog user across the JVM, which Python does since #7571 but which would change when a new table load sees new permissions in Java. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
