gaoyunhaii commented on PR #21736:
URL: https://github.com/apache/flink/pull/21736#issuecomment-1457659571

   Hi @XComp 
   
   For the first point, sorry it is indeed my misunderstanding, and we are in 
fact consistent. 
   
   For the second point, after some more thoughts it now looks to me we indeed 
need to do the data generate on publishing instead of cutting branch, since 
there are still new codes merged during the period before formally published. 
   
   For migration test data generation, we in fact have a parameter to specify 
the target version so that the snapshots data could be located in the right 
location, thus I think this step is ok. 
   
   The main issue is that currently the migration tests relies on the 
mostRecentlyPublishedVersion() to list the versions to test against. Before 
1.17 is published, there is not test data for release 1.17, and the versions 
should be [some start version, 1.16], after 1.17 is published, the versions 
would become [some start version, 1.17]. 
   
   It looks its not easy to distinguish the two cases without more information. 
The git tags might not always exist (for example, users download sources from 
the website) and is not easy to acquire from the java code. Do you think it is 
ok for us to have a constant in `FlinkVersion` like 
   
   ```
   public static FlinkVersion MOSTLY_RECENT_PUBLISHED_VERSION = v1.17;
   ```
   
   and the RM will update the variable on publishing after generating the test 
data ?
   


-- 
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]

Reply via email to