GitHub user edespino added a comment to the discussion: [feature] cherry-pick 
ABI Check for GitHub Actions

Thank you @yihong0618  and @avamingli  for this thoughtful discussion about ABI 
compatibility checking.

I agree that adopting ABI testing is valuable, but I think we need to frame it 
within a broader versioning and compatibility strategy. I previously proposed a 
PostgreSQL-style branching strategy with SemVer 
([#591](https://github.com/apache/cloudberry/discussions/591)) that could help 
address some of the concerns @avamingli raised:

1. For ABI stability promises:
- Major versions (X.0.0): Can include ABI breaks with clear documentation
- Minor versions (0.X.0): Should maintain ABI compatibility for supported 
extensions
- Patch versions (0.0.X): Must maintain strict ABI compatibility

This would give us clearer guidelines for:
- When we can make breaking changes
- What level of compatibility testing is needed
- How to handle emergency fixes that might break ABI

The beauty of adopting SemVer is that it provides clear expectations to the 
community. While PostgreSQL doesn't follow SemVer, we have the opportunity to 
provide stronger compatibility guarantees that would benefit extension 
developers.

Regarding the specific concerns about private/unsupported extensions - we could 
maintain a list of "supported extensions" that are part of our ABI testing 
suite. This would give us a clear scope for compatibility promises while still 
allowing the community to build and maintain their own extensions with clear 
understanding of support levels.

Would it make sense to combine the ABI testing implementation with the 
versioning proposal? We could start with a small set of core extensions for 
testing and expand from there.

GitHub link: 
https://github.com/apache/cloudberry/discussions/728#discussioncomment-11369675

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: 
[email protected]


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

Reply via email to