zhangyue19921010 opened a new issue, #11309: URL: https://github.com/apache/paimon/issues/11309
### Search before asking - [x] I searched in the issues and found nothing similar. ### Paimon version master at 71082ea86a57930f3fa00bb74f60d0a5dd676e35 (2.2-SNAPSHOT). ### Compute Engine Java API; the same filesystem schema merge is used by Spark path-based writes. ### Minimal reproduce step 1. Create an ordinary append table with an `id INT` column and open it with `FileStoreTableFactory.create(fileIO, path)`. 2. Call `table.store().mergeSchema` with an incoming row type containing `id INT, requested STRING`. 3. After the merge reads schema 0, let another schema writer add `other INT`, publishing schema 1 before the first writer's atomic schema-file publication. 4. The first writer's atomic publication returns false because schema 1 already exists. `mergeSchema` also returns false, although `requested` was never added. The regression in `SchemaManagerTest` injects the competing real schema commit immediately before the atomic write, without timing-dependent threads. ### What doesn't meet your expectations? `mergeSchema` documents false as a no-op, but a lost commit is not a no-op. Callers can continue without the requested schema; Spark's schema helper only reloads its table when the result is true. A failed publication should raise an exception so the whole operation can be retried with fresh schema state. ### Anything else? The proposed fix checks the atomic commit result and reports the conflict. Successful merges and true no-op merges retain their current return values. ### Are you willing to submit a PR? - [x] I'm willing to submit a PR! -- 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]
