14 Jul
2014
14 Jul
'14
2:24 a.m.
Hi all, I've been noticing some strange Float behaviour (not rounding errors but totally erroneous values for non-pathological values). It seems to be a manifestation of GHC ticket #9125 which was reported from a Raspberry Pi (arm) version. A quick example:
print (29.0 :: Float)
gives
2109.0
There is more detail at https://ghc.haskell.org/trac/ghc/ticket/9125#comment:1 Has anyone else seen this? I'm running a compiled-from-source-tarball 7.8.3 as I couldn't get past the dreaded 'illegal instruction:4' binary installation bug. Simulator and device behave the same. Thanks.