From: Justin Bailey <jgbailey@...>
Date: 2006-07-15T00:36:37+09:00
Subject: Unexpected pointer behavior with unpack

------=_Part_37845_31708227.1152891394993
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

I have had the opportunity to work [1] a lot with Ruby's ability to create
and manipulate pointers via pack/unpack recently, and have discovered
surprising behavior when dealing with nested pointers. I'm interested in
other's thoughts on the analysis below. I wasn't able to find a lot of
straightforward documentation about this Ruby ability.

If I take a string, pack it to a pointer, and then pack that pointer into
another string, I lose the ability to retrieve the inner pointer. However,
if I make a reference to the "inner" pointer in a local variable, I can
still retrieve it. An example is best:

  puts (inner = ["foo"].pack("P")).inspect # "xxxx" - some 4 byte string
  puts inner.unpack("P3").first.inspect # "foo"
  puts (outer = [inner].pack("P")).inspect # "yyyy" - another 4 byte string
  puts outer.unpack("P4").first.inspect # "xxxx" - first 4 byte string from
above
  puts inner.unpack("P3").first.inspect # "foo" - still can get to inner
structure

As is shown, I can create a two packed pointers and get access to them
through individual variables. However, if I try to access the inner
structure from the outer, like so:

  puts outer.unpack("P4").first.unpack("P3") # ArgumentError - no associated
pointer

I get an error. Now, I think I understand why this happens - pointers have
to be associated with variables. It would be dangerous to unpack to treat
any string as a pointer. However, I do think the above is surprising
behavior. It took me awhile to understand why my nested structures didn't
seem to work.

Does anyone else agree with that this is suprising behavior or am I missing
something?

Justin

[1] I needed to call some complex Win32 API functions which took nested
structures, that were allocated in Ruby but filled out by the API, and
that's what led me down this path.

------=_Part_37845_31708227.1152891394993--