Doesn't allow building against parsec >= 3

[GoogleCodeExporter]

My libraries have updated to parsec 3.0.0 (some dependency pulled it in), 
which stopped pandoc from building.  But if I just change the 
pandoc.cabal file to allow parsec 3.0.0, then it actually builds fine.  
Obviously parsec 3 is quite backwards compatible.

Original issue reported on code.google.com by L.Plant...@cantab.net on 30 Dec 2008 at 11:41

[GoogleCodeExporter]

My main reason for requiring parsec < 3 is performance: pandoc is quite a bit 
slower
when compiled with parsec 3.  I understand that the parsec 3 developers are 
working
on this issue, but until it is resolved, I want to make sure pandoc will build 
with
parsec 2 by default.

Original comment by fiddloso...@gmail.com on 31 Dec 2008 at 2:44

[GoogleCodeExporter]

Original comment by fiddloso...@gmail.com on 13 Jul 2009 at 3:03

• Changed state: WontFix

[GoogleCodeExporter]

Does parsec 3.1 change things?
http://www.haskell.org/pipermail/haskell-cafe/2010-March/074076.html

Original comment by eric....@gmail.com on 4 Apr 2010 at 10:04

[GoogleCodeExporter]

I think that parsec 3.1 does fix the performance problems.  But I'm still 
reluctant to 
change the default, for the following reason.  You'll get problems if you try 
to link 
libraries compiled against parsec 3 with libraries compiled against parsec 2.  
There 
are still lots of libraries that require parsec 2, so I think that the whole 
community 
has to decide to make the change to parsec 3, or there will be problems linking 
pandoc with those libraries.

Two possible ways forward:
- whole community can decide to lift the parsec < 3 restrictions.
- we can rename one of the packages, as was suggested on one of the haskell 
lists, 
either naming parsec 2 "parsec98", or renaming parsec 3.  Then the link 
conflict 
problem goes away.

I think the second solution is best, myself.

Original comment by fiddloso...@gmail.com on 4 Apr 2010 at 4:29

[GoogleCodeExporter]

PS. I could, in the mean time, add a cabal flag that allows pandoc to be built 
against 
parsec 3 (defaulting to False), if that would be useful.

Original comment by fiddloso...@gmail.com on 4 Apr 2010 at 4:31