robertmu commented on issue #65:
URL: 
https://github.com/apache/cloudberry-backup/issues/65#issuecomment-3876673242

   > > > > Thanks for reporting this issue.
   > > > > Currently, cloudberry-backup primarily supports Cloudberry Database 
2.0. Could you please let us know which version of Cloudberry Database you are 
using for your tests?
   > > > 
   > > > 
   > > > Cloudberry 2.1 - I use the `REL_2_STABLE` as the build source.
   > > 
   > > 
   > > Currently, cloudberry-backup only supports up to 
apache-cloudberry-2.0.0-incubating, so version 2.1 isn't supported yet. The 
integration test failures are likely caused by catalog differences between 
2.0.0 and 2.1. Interestingly, according to PostgreSQL rules, minor version 
changes shouldn't really have catalog changes, but that seems to be the case 
here
   > 
   > Aha. Now the Cloudberry main branch has the catalog changes compared to 
`REL_2_STABLE`, but for 2.1 and 2.0, they keep campatible: 
[apache/cloudberry#1547](https://github.com/apache/cloudberry/pull/1547) (CI: 
https://github.com/apache/cloudberry/actions/runs/21424095867)
   
   Thanks for the clarification. Although 2.0 and 2.1 are intended to be 
compatible, these specific integration test failures confirm that there are 
indeed catalog or behavioral changes in version 2.1:
   
   1. RejectLimitType Mismatch: The mismatch where the tool receives \xff 
instead of the expected empty string "" for pg_exttable.rejectlimittype 
indicates a change in how this "char" field is initialized or stored. 0xFF is 
often an internal marker in C, and its appearance suggests 2.1 handles "empty" 
states differently in the catalog.
   
   2. Statistics Restore Failure: The error invalid input syntax for type 
boolean: "2" during the pg_statistic restore is very telling. It means the 
stavalues column in pg_statistic (populated by ANALYZE) now contains values 
(like "2") that the database's own boolean input parser no longer accepts 
during an INSERT. This confirms that the data format or the ANALYZE logic in 
2.1 has diverged from 2.0.These subtle data-level differences are precisely why 
the integration tests are failing, despite the branches being theoretically 
compatible.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to