[Haskell-cafe] GHC 8.8 and MonadFail proposal

[Kazu Yamamoto (=?iso-2022-jp?B?GyRCOzNLXE9CSScbKEI=?=)]

Hello,

I built the GHC 8.8 branch and tried to use it. I was very surprised
because many packages cannot be built due to the removal of the "fail"
method from the Monad typeclass.

GHC 8.6 (with -Wcompat) does not warn even if "fail" is defined.  I
had no chance to notice this change. Isn't this change too radical?

The same thing happened in GHC 8.4 due to Semigroup. I could not
notice this change beforehand because I did not know the "-Wcompat"
option. In my opinion, we should have included "-Wcampat" in "-Wall".

Anyway, if I understand correctly, "-Wcompat" does not help in 
the MonadFail case.

I'm a fan of both proposal. But I feel that no period of warnings
is not a good approach.

--Kazu