[GHC] #647: Socket bug on Mac OS X, with patch

[GHC]

#647: Socket bug on Mac OS X, with patch
----------------------------------+-----------------------------------------
  Reporter:  gwright at antiope.com  |          Owner:         
      Type:  bug                  |         Status:  new    
  Priority:  normal               |      Milestone:  6.4.2  
 Component:  libraries/network    |        Version:  6.4.1  
  Severity:  normal               |     Resolution:         
  Keywords:  "OS X" Sockets       |             Os:  MacOS X
Difficulty:  Easy (1 hr)          |   Architecture:  powerpc
----------------------------------+-----------------------------------------
Comment (by gwright at antiope.com):

 As far as I know, this bug is really only for darwin and the older BSDs.
 Traditional BSDs
 required the sockaddr_in structure to be completely zeroed before use even
 if all of the
 fields were properly initialized.  This is necessary because the structure
 is padded out to
 16 bytes, and functions like bind check the pad space to make sure it is
 all zero.  This is a sanity
 check from the earliest days of BSD, when there were other 16 byte
 sockaddr_* structures,
 like sockaddr_ns for XNS networks.  The sockaddr_ns structure had a
 different field layout.
 Checking the pad space was a way to ensure that a socket address for the
 internet family
 (AF_INET) was not filled in with an XNS address.

 Yes, this is all the result of some Berkeley graduate student being over-
 cautious
 a quarter of a century ago.

 Linux and recent FreeBSDs do not require zeroing the sockaddr_* structure
 before use.
 (I've tested on FreeBSD 6.0, and zeroing first is not needed.  I think
 Solaris does not
 require it either.)  It is a bit annoying that OS X could not fix this.

 If you look at the first edition of Stevens' _UNIX Network Programming_,
 you see that
 every time before a sockaddr_* struct is used, it is bzero'ed first.
 Although as Wolfgang
 notes, it was never well documented.

 Probably more than you wanted to know about sockets, no?

 -Greg

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/647>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler