#9564: Floating point subnormals overrounded on output -------------------------------------+------------------------------------- Reporter: jrp | Owner: simonmar Type: bug | Status: new Priority: normal | Milestone: Component: Runtime | Version: 7.8.3 System | Keywords: Resolution: | Architecture: x86_64 (amd64) Operating System: MacOS X | Difficulty: Unknown Type of failure: Incorrect | Blocked By: result at runtime | Related Tickets: Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Comment (by carter): the standard defines the read/show for Float/Double as {{{ instance Show Float where showsPrec p = showFloat instance Read Float where readsPrec p = readSigned readFloat instance Show Double where showsPrec p = showFloat instance Read Double where readsPrec p = readSigned readFloat }}} at the bottom of https://www.haskell.org/onlinereport/haskell2010/haskellch9.html#x16-1710009 then per https://www.haskell.org/onlinereport/haskell2010/haskellch38.html#x46-314000... {{{ showFloat :: RealFloat a => a -> ShowS Show a signed RealFloat value to full precision using standard decimal notation for arguments whose absolute value lies between 0.1 and 9,999,999, and scientific notation otherwise. }}} then looking at the stuff in 4.7.0.1 base http://hackage.haskell.org/package/base-4.7.0.1/docs/src/GHC- Float.html#showFloat {{{ -- | Show a signed 'RealFloat' value to full precision -- using standard decimal notation for arguments whose absolute value lies -- between @0.1@ and @9,999,999@, and scientific notation otherwise. showFloat :: (RealFloat a) => a -> ShowS showFloat x = showString (formatRealFloat FFGeneric Nothing x) }}} I'll try to dig into this a teeny bit more, but I think thank as long as `(read . show) :: Float -> Float` acts as the identity function on all floating point values when we roundtrip them. (and as long as they get correctly parsed to that same internal value when read/showed between haskell and another language) phrased differently, if we dont have roughly that `read_hs . show_clang == read_clang . show_hs == read_clang . show_clang == read_hs . show_hs == id ` then yes we have a problem, but we dont quite show that problem with theses tests as above, right? We just demonstrate that the particular choice in default representation for show differs from C, right? Its important to remember that parsing a floating point number itself will do rounding to the nearest floating point value too. i'm trying to focus on some work work this week, but if someone could test that these roundtripining identities work out ok, that'd be awesome (I may try to do it soon myself, but who knows how long that will take to happen) -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/9564#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler