duncan.coutts:
On Thu, 2008-08-14 at 10:21 -0700, Don Stewart wrote:
I think you're right. The Binary instances cannot and must not read more than they need to, so that gives us the behaviour that we read exactly the length of the file, but no more, and thus we never hit EOF, so we don't close the file. So yes, decode should force the tail so that it can indeed hit EOF.
Duncan,
You're suggesting that decode and decodeFile should whnf the next cell?
At least decodeFile should, since it doesn't give you any other access to the file handle otherwise.
Does decode return the tail? I don't remember. If not it should also whnf it. If it does then the user can choose (they might want to do something else with the trailing data).
I've pushed a decodeFile that does a whnf on the tail after decoding. If you're at the end of the file, that's sufficient to close the Handle. You'll also need bytestring >= 0.9.1.0 (note, not the one that comes by default with ghc 6.8.x) -- Don