Re: [cxx-abi-dev] non-public non-static members vs POD layout

[David Vandevoorde <daveed@xxxxxxx>]

On Apr 28, 2008, at 6:59 PM, Mark Mitchell wrote:
> Scott --
>
> My personal opinion is that the ABI does not guarantee that "almost"  
> PODs are laid out as PODs. In fact I think it says that such things  
> must be laid out using the rules in the C++ ABI -- surprising though  
> that might be.
>
> Also since the ABI predates TC1, the inclusion of pointers to  
> members on PODs should not apply. I suppose the spec would  
> explicitly say "PODs as defined prior to TC1"
>
> Does anyone object to making that change?


I think so.  Hereunder is an e-mail to this list (from a few years  
ago) that seems to indicate we already agreed to use the TC1  
definition of PODs.

	Daveed



Date: Mon, 25 Oct 2004 18:13:01 -0700
From: Mark Mitchell <mark@xxxxxxxxxxxxxxxx>
Organization: CodeSourcery, LLC
To: Mark Mitchell <mark@xxxxxxxxxxxxxxxx>
CC: Nathan Sidwell <nathan@xxxxxxxxxxxxxxxx>,
  Dennis Handly <dhandly@xxxxxxxxxx>, cxx-abi-dev@xxxxxxxxxxxxxxxx
Subject: Re: [cxx-abi-dev] What is a POD?  TC1 or first C++ Standard


Mark Mitchell wrote:

> Yes, you're correct, the standard does guarantee that you can do
> bitwise copies for PODs  using their size.
> In other words, your point is that unless we interpret the ABI as
> meaning "TC1 POD", it is not the ABI for a correct C++ compiler,
> because it will fall afoul of this rule.  This is [basic.types]/3,  
for
> reference.
>
> I think that's a pretty compelling agument that we should clarify the
> ABI to indicate that POD means "POD as defined by TC1".  Any
> objections to that proposal?  Are there any compilers other than aCC
> implementing the pre-TC1 definition?

Nobody objected, so I committed the following change, requiring the TC1
definition.  The change will be reflected in the online copy shortly.  I
also clarified the fact that an array that is a POD type, but whose
elements are not a POD for the purpose of layout is not itself a POD for
the purpose of layout.

FYI,

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@xxxxxxxxxxxxxxxx


Index: abi.html
===================================================================
RCS file: /home/cvs/Repository/cxx-abi/abi.html,v
retrieving revision 1.78
diff -c -5 -p -r1.78 abi.html
*** abi.html	24 Sep 2004 03:47:55 -0000	1.78
--- abi.html	26 Oct 2004 01:09:26 -0000
*************** sharing a virtual pointer with the deriv
*** 228,242 ****

 <p>
 <a name="POD" />
 <dt> <i>POD for the purpose of layout</i><dt>
 <dd>
! A type is considered a POD for the purposes of layout if it is a POD
! type (in the sense of [basic.types]), and is not a POD-struct or
! POD-union (in the sense of [class]) with a bitfield member whose  
declared
! width is wider than the declared type of the bitfield
! </dd>
 </p>

 <p>
 <dt> <i>primary base class</i> </dt> <dd> For a dynamic class, the
 unique base class (if any) with which it shares the virtual pointer at
--- 228,258 ----

 <p>
 <a name="POD" />
 <dt> <i>POD for the purpose of layout</i><dt>
 <dd>
! <p>
! In general, a type is considered a POD for the purposes of layout if
! it is a POD type (in the sense of ISO C++ [basic.types]).  However, a
! POD-struct or POD-union (in the sense of ISO C++ [class]) with a
! bitfield member whose declared width is wider than the declared type
! of the bitfield is not a POD for the purpose of layout.  Similarly, an
! array type is not a POD for the purpose of layout if the element type
! of the array is not a POD for the purpose of layout.  Where references
! to the ISO C++ are made in this paragraph, the Technical Corrigendum 1
! version of the standard is intended.
!
! <p>
! <img src=warning.gif alt="<b>NOTE</b>:">
! The ISO C++ standard published in 1998 had a different definition of
! POD types.  In particular, a class with a non-static data member of
! pointer-to-member type was not considered a POD in C++98, but is
! considered a POD in TC1.  Because the C++ standard requires that
! compilers not overlay the tail padding in a POD, using the C++98
! definition in this ABI would prevent a conforming compiler from
! correctly implementing the TC1 version of the C++ standard.
! Therefore, this ABI uses the TC1 definition of POD.</dd>
 </p>

 <p>
 <dt> <i>primary base class</i> </dt> <dd> For a dynamic class, the
 unique base class (if any) with which it shares the virtual pointer at
*************** unwind table location.
*** 4939,4948 ****
--- 4955,4972 ----
 <p> <hr> <p>

 <p>This version of this document is $Revision: 1.78 $.  No special
 significance should be attached to the form of the revision number; it
 is simply a identifying number.</p>
+
+ <p>
+ <font color=blue>[041025]</font>
+ Indicate that the TC1 definition of POD is intended in the section
+ defining a &quot;POD for the purpose of layout&quot;.  Clearly
+ indicate that an array whose elements are not PODs for the purpose of
+ layout is itself not a POD for the purpose of layout.
+ </p>

 <p>
 <font color=blue>[040923]</font>
 Clarify behavior of <code>__cxa_vec_delete</code>.
 </p>