Meeting URL: https://meet.jit.si/StableHaskellMeetBiWeekly
Previous meeting notes
- Left from last time
- CLC proposal to warn on
head
/tail
would break users of-Werror
.
- CLC proposal to warn on
- New items
- Organizing impact as discussed in this thread
- Updates on in-progress work
- GHC.X.Hackage second draft
- GHC tick-tock releases
- Last time we said this needs revision or closure, did this happen?
- Update on Language/compiler features to help stability
- Exposing package internals
- GHC warning policy document as discussed previously
- GHC API: following the Haskell Implementors Workshop and Haskell Symposium, David C offered to convene a working group to work out what kind of more stable GHC API would be most useful -- GHC as a library.
- At the moment it is not clear what the GHC API is (except for "every module in GHC" which is unhelpful).
- Having a clear API would make it much easier to use, much easier for GHC devs to know what APIs are public (and hence with a high stablity threshold) and which are entirely internal to GHC (and hence can be changed at will).
- We hope that this group's product will be a useful resource to GHC development going forward, and that the group can serve as a source of feedback
-
CLC proposal to warn on
head
/tail
would break users of-Werror
.- How concerned should be we about breaking libraries that use
-Werrror
? - GHC proposal about user-defined-warnings
- GHC proposal about Warning pragmas with categories This would allow users to disable specific warnings (eg. those for head and tail).
- Our role: we could (say) identify a GHC Proposal that would really help to smooth the path for otherwise-sensible CLC proposals.
- How concerned should be we about breaking libraries that use
-
Organizing impact as discussed in this thread
- Can we provide a framework for considering impact to the community?
- Outcome: a markdown file in our repo.
- Include: a list of implied priorities for each "slice" of users.
- David will create the file; Trevis will make the "list of implied priorities"
-
- No new developments; awaiting Ben.
- But a putative consensus that Something Should Be Done, and roughly what.
- David will talk to Ben about what to do next
-
- No consensus of "user voice" has emerged here. So let's just park this (unless Ben wants to pursue it).
-
GHC API
- Have a separate
ghc-api
package? Or a subset of modules in theghc
package that are advertised as stable.- An issue driving this is that ghc the library and ghc the application share versions.
- Could we collect from the community which parts of ghc are used?
- How does this group interact with the new working group?
- Have a separate
-
- Neither Simon nor Richard has enough cycles to actively push this.
- There have been surprisingly mixed reactions
- Several people have said "just use a convention" e.g. "it's unstable if it has
Internal
in the module name" - SPJ's suggestion: park this pending more user pressure. Encourage the GHC folk articulate a policy about what is public API for
ghc-prim
andbase
libraries. - Trevis has ideas, but is unsure how to amend the proposal to fit or if we should open new proposal(s).
-
GHC warning policy document as discussed previously