[Haskell-cafe] '#' in literate haskell

[John MacFarlane]

Can anyone explain why ghc does not treat the following
as a valid literate haskell program?

--------- test.lhs ----
# This is a test

> foo = reverse . words

------------------------

When I try to load this in ghci (or compile it using ghc),
I get:

test.lhs:1:2: lexical error at character 'T'

It seems that the problem is the '#' character in the first
column.  Replacing it with something else, or moving it to
the right one space, solves the problem.

The following literate haskell program, from
http://notvincenz.blogspot.com/2008/01/literate-haskell-and-c.html,
also fails to load for me, for the same reason (the leading
'#' in line 8).

---- literate-haskell-and-c.lhs ---
/* c and lhs file

>
> module Foo where
> main = print "Haskell"
> 

*/
#include 

int main() {
  printf("C\n");
  return 0;
}
------------------------------------

I've reproduced this with ghc 6.10.1 and ghc 6.8.3 (linux binaries
from haskell.org) and with ghc 6.8.2 (Ubuntu intrepid).
Interestingly, hugs (September 2006 version) has no trouble with
test.lhs. I haven't tried ghc 6.6.

I care about this because I'd like to use markdown conventions
to format the comment parts of literate haskell programs.  Markdown
supports atx-style headers, which begin with strings of '#'
characters starting in the first column. I know that some people use
markdown with literate haskell, so there must be something basic here
that I'm missing!

John