Thanks for picking this up! I'm not surprised that the TH error is gone; this patch didn't change anything TH-related, I think it was just tripping on some unrelated bug which has been fixed in the meantime.
Nov 23 2015
Sep 27 2015
Have you considered adding one single datatype to the universe, like Rec0 but of a different kind (perhaps URec0 :: # -> *)? That way you can still address all unboxed cases with a single URec0 a instance. Alternatively, make Rec0 poly-kinded.
Sep 12 2015
Looks good to me, thanks. Perhaps just worth adding somewhere that the reason why the new datatypes are introduced is that the kinds of the unlifted types prevent Generic instances being given to them.
Aug 27 2015
I'm afraid not, at least not in the foreseeable future. I'd love it if someone were to pick it up, though!
Aug 22 2015
Looks good to me too.
Aug 5 2015
I was waiting for @ezyang to reply back on the changes after his comment.
Jul 20 2015
I'm happy to review this next week, once I'm back from holidays.
Jan 27 2015
Looks good to me.
Jan 19 2015
Nov 26 2014
Currently this diff panics when haddock runs on Language.Haskell.TH.Syntax. Removing all derive Generic clauses from that file makes it go through. Surprisingly, removing the NoInline constructor on line 1268 also makes the panic go away (though things fail later on due to a missing constructor).
Nov 23 2014
Patch is rebased, but I don't really know how I can see exactly what failed...
Update the patch to the latest HEAD.
Nov 18 2014
Superseded by D493.
- Fix two mistakes
Whitespace changes, plus address some of Simon's remarks.