Variants of ticket #1200 still cropping up

[rrnewton]

I see that ticket #1200 from several years back is currently closed.

However, I was in the midst of writing some simple tests to help myself understand ReaderT ContT interaction and again ran into these frustrating undefined errors coming from printf:

import Control.Monad.Cont as C
import qualified Control.Monad.Reader as R
import Data.IORef
import Text.Printf

test ref = do 
  x <- R.ask 
  liftIO$ printf "Observed value %d before callCC\n" x 
  callCC$ \cont -> do
     y <- R.ask 
     liftIO$ writeIORef ref cont
     liftIO$ printf "Observed value %d inside callCC\n" y
  z <- R.ask 
  liftIO$ printf "Observed value %d in invoked continuation\n" z

main = do ref <- newIORef (error "unused")
          let test' = do test ref
-- Uncommenting the following will fix it:
--                         return ()
	      m1 = R.runReaderT test' (100::Int)
              m2 = C.runContT m1 (\ () -> return ())
          m2
          putStrLn "Done with main."

I see only two occurrences of undefined in Printf.hs. Is there any disadvantage to changing them to appropriate errors? In fact, is there any reason that undefined shouldn't be banned from the standard libraries in favor of error?

[simonpj]

Absolutely. Please send patch!

S

[rrnewton]

Trivial patch uploaded. It looks like I botched the naming convention. By the way, the wiki section on the patch naming convention seems to be missing. It's referred to from here:

 http://hackage.haskell.org/trac/ghc/wiki/WorkingConventions/FixingBugs

And should live here:

 http://hackage.haskell.org/trac/ghc/wiki/WorkingConventions/Git#PatchNaming

[igloo]

Replying to rrnewton:

Trivial patch uploaded.

Applied, thanks!

By the way, the wiki section on the patch naming convention seems to be missing. It's referred to from here:  http://hackage.haskell.org/trac/ghc/wiki/WorkingConventions/FixingBugs And should live here:  http://hackage.haskell.org/trac/ghc/wiki/WorkingConventions/Git#PatchNaming

Fixed, ta.