The GHC version detection didn't actually work right, resulting in the __GLASGOW_HASKELL__ < 800 branch always being taken. This became apparent once I tried using GHC 8.8, as the fallback of fail inside Monad is no longer there, so it fails to compile. We need to fix this by playing the same sort of double-#define games that GenericTemplate.hs does. I plan to do that in a later PR but would like to unbreak this for now. Since there's been no release of Alex since this was added, this shouldn't qualify as a breaking change.
The GHC version detection didn't actually work right, resulting in the
__GLASGOW_HASKELL__ < 800branch always being taken. This became apparent once I tried using GHC 8.8, as the fallback offailinsideMonadis no longer there, so it fails to compile. We need to fix this by playing the same sort of double-#define games that GenericTemplate.hs does. I plan to do that in a later PR but would like to unbreak this for now. Since there's been no release of Alex since this was added, this shouldn't qualify as a breaking change.Reverts simonmar/alex#136