From: Shyouhei Urabe <shyouhei@...>
Date: 2011-05-21T08:26:15+09:00
Subject: [ruby-core:36361] [Backport87 - Backport #3297][Feedback] [patch] Array#pack("i_") is broken in big endian 64bit systems


Issue #3297 has been updated by Shyouhei Urabe.

Status changed from Open to Feedback

Sorry for this late reply, does this still happen?  I guess it has already been fixed for a while.
----------------------------------------
Backport #3297: [patch] Array#pack("i_") is broken in big endian 64bit systems
http://redmine.ruby-lang.org/issues/3297

Author: Klaus K��mpf
Status: Feedback
Priority: Normal
Assignee: 
Category: core
Target version: 
ruby -v: ruby 1.8.7 (2008-08-11 patchlevel 72) [powerpc64-linux] 


=begin
 See also http://redmine.ruby-lang.org/issues/show/1131 for a similar issue (String#unpack("V"))
 
 Array#pack("i_") is broken in big endian 64 bit systems
 
 irb> [1].pack("i_")
 => "\000\000\000\000"
 
 
 The root cause is similar to the one of issue 1131, use of a wrong offset in a 64bit long.
 But the fix is a bit simpler and is probably correct for all other architectures too.
 
 According to the documentation, the "i" format of pack/unpack uses a fixed-size (32bit) integer, while adding "_" uses the system-size.
 However, the implementation of Array#pack uses a 'long' type (64bit !) to store an intermediate value, thus breaking OFF() in accessing the wrong 4 bytes of the value.
 
 Fix: Use 'int' to store the intermediate value (use of int is correct for the fixed-size as well as the system-size case) and don't use OFF()
 
 With this patch, the pack_spec and unpack_spec tests of rubyspec.org run without issues.
=end



-- 
http://redmine.ruby-lang.org