Hi Richard,

I believe that this is mostly due to plugin development happening to satisfy a plugin need. I doubt there is a grand unified vision for plugins. And I don't have one either. I've dabbled with codegen plugins a long time ago, these days I'm primarily concerned with plugins having a chance to work in cross compilation settings, and even that is still a very uncharted area, but Luite has come up with a hack and Sylvain is making progress :-) We still don't have the cabal side fixes, where we'd need some `plugin-depends` stanza, but all that only makes sense, once we have the fundamentals for plugins disentangled in ghc.

I agree that a discussion on discourse might help. But we won't know without trying.

On Tue, Aug 31, 2021 at 9:34 PM Richard Eisenberg <lists@richarde.dev> wrote:

Hi all,

I have seen a few posts from Sam Derbyshire here asking for feedback about plugin API design, and the responses have been minimal. This poses a design challenge, because the GHC folk who design the interface are sometimes distinct from the people who use the interface. We're trying to be good, seeking feedback from real, live clients. Is there a better way to do so than this mailing list? Example: we could create a Category on discourse.haskell.org, if that would reach the audience better. Or we could make a repo with issue trackers somewhere simply to track plugin design. What would work?

(I recognize that I'm asking in a perhaps-ineffective channel for advice, but I really don't have a better idea right now. Maybe some of you plugin authors are here and will point us in a better direction.)

Thanks,
Richard
_______________________________________________
ghc-devs mailing list
ghc-devs@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs