The Stability Working Group has continued to meet and this is to provide a monthly view into the discussions, and projects.
The group met twice during May 2023.
Main topics of discussion included:
The CLC proposal around clarifying GHC internals exposed in base.
Strategies for migrating those GHC internals out of base.
Recruitment for the group
On going and completed projects:
(Completed) Add a glossary to the GHC documentation to give a clarifying space on terms such as “technology preview” when they are used elsewhere in documentation and release notes.
Examine possibilities for a more stable TemplateHaskell interface.
Additional clarifications around some GHC features.
As noted above we are actively seeking more membership, particularly as the current members remain bandwidth limited for seeing actions completed. If you would like to contribute in some way please reach out.
Further, the group exists to serve the community looking for paths to reduce the cost of inevitable change throughout the ecosystem. If there is anything you want to be sure the group sees please reach out!
You can also reach out to me directly if for any reason you do wish to reach the group directly. A direct message here, via email mailto:trevis@flipstone.com, on Github with username telser, are just a few of the ways to contact me. I’m happy to discuss any ideas for the group as well as proxy any request that the group talk about a particular issue.
Finally, all feedback on this update is greatly appreciated.
As someone who implemented breaking changes in template-haskell for the GHC 9.8 release, I’d be interested to hear your thoughts as to how I could have avoided that.
We have compatibility libraries th-compat and th-desugar, but not all packages are using those. Is there anything that I, as a GHC developer implementing a new feature, could have done differently, in your opinion?
I’ll definitely bring this to the group to discuss, thank you!
Am I reading the head.hackage change correctly that with th-abstraction and th-desugar patched, there were only 10 other packages in need of patching? I have seemingly forgotten just how much of hackage gets tested, but unless it is a very small subset that already feels like a pretty big win for general use of those two libraries as providing a more stable interface.
Again, I’ll share with the group to see if anyone else has ideas. I’m still ingesting this change and what it means, but someone who uses Template Haskell more might have a quicker take.
