draft release notes for GCC 3.2

draft release notes for GCC 3.2

[Joe Buck]

(As usual, this info is obtained from reading ChangeLog files; if anything
 is missing or wrong, please let me know).
--------------------------------------------------------------------------
Draft release notes for GCC 3.2

3.2 is a small bug fix release, but there is a change to the application
binary interface (ABI), hence the change to the second part of the version
number.

The main purpose of the 3.2 release is to correct a couple of problems in
the C++ ABI, with the intention of providing a stable interface going
forward.  Accordingly, 3.2 is only a small change to 3.1.1.

In addition, a number of bug fixes for the C++ standard library are
included in this release.

There is one platform-specific bug fix (for x86_64 - AMD's 64-bit
architecture) Other than that, for users of C, Java, Objective-C, or
Fortran, there are no changes with respect to 3.1.1.

(Jakub's change to avoid emitting compound literals that are unused: worth
mentioning?)

Below you will find a list of the problem reports (PRs) from GCC's bug
tracking system (GNATS) that are known to be fixed in the 3.2
release. This list might not be complete (that is, it is likely that some
bug fixes have been missed).

To learn more about a particular PR, visit the link

	http://gcc.gnu.org/cgi-bin/gnatsweb.pl

and enter the PR number in the "View Problem Report" dialog box.  Then
click the "View" button.  It's not necessary to log in to view reports.

C++:
7320: g++ 3.2 relocation problem

libstdc++:
6410: Trouble with non-Ascii monetary symbols and wchar_t
6503, 6642, 7186: Problems with comparing or subtracting various types
		  of const and non-const iterators
7216: ambiguity with basic_iostream::traits_type
7220: problem with basic_istream::ignore(0,delimiter)
7222: locale::operator==() doesn't work on std::locale("")
7286: placement operator delete issue
7442: cxxabi.h does not match the C++ ABI
7445: poor performance of std::locale::classic() in multi-threaded applications

x86_64 specific:
7291: off-by-one in generated inline bzero code for x86_64

Re: draft release notes for GCC 3.2

[Nathan Sidwell]

Joe Buck wrote:
> --------------------------------------------------------------------------
> Draft release notes for GCC 3.2

> C++:

PR 7432 reports an (unintentional) ABI change between 3.1 and 3.1.1
The alignment of int bitfields which were extended to long long
changed on x86 systems. A 'long long' field is aligned to a 4
byte boundary, but a 'long long' bitfield is aligned to an
8 byte boundary. The new behaviour matches the C compiler's
behaviour (but, I believe disagrees with intel's compiler) --
this may be a long standing x86 ABI incompatibility.

PR 7470 reports another ABI discrepancy, which I am working on.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: draft release notes for GCC 3.2

[Joe Buck]

> Joe Buck wrote:
> > --------------------------------------------------------------------------
> > Draft release notes for GCC 3.2
> 
> > C++:
> 
> PR 7432 reports an (unintentional) ABI change between 3.1 and 3.1.1
> The alignment of int bitfields which were extended to long long
> changed on x86 systems. A 'long long' field is aligned to a 4
> byte boundary, but a 'long long' bitfield is aligned to an
> 8 byte boundary. The new behaviour matches the C compiler's
> behaviour (but, I believe disagrees with intel's compiler) --
> this may be a long standing x86 ABI incompatibility.

So, is it agreed that the 3.1.1 behavior is correct?  Matching the
behavior of C is a good thing.  To me, matching the behavior of Intel's
compiler is low priority, but preserving the C ABI with respect to older
GCC versions is vital.

Re: draft release notes for GCC 3.2

[Nathan Sidwell]

Joe Buck wrote:
> 
> > Joe Buck wrote:
> > > --------------------------------------------------------------------------
> > > Draft release notes for GCC 3.2
> >
> > > C++:
> >
> > PR 7432 reports an (unintentional) ABI change between 3.1 and 3.1.1
> > The alignment of int bitfields which were extended to long long
> > changed on x86 systems. A 'long long' field is aligned to a 4
> > byte boundary, but a 'long long' bitfield is aligned to an
> > 8 byte boundary. The new behaviour matches the C compiler's
> > behaviour (but, I believe disagrees with intel's compiler) --
> > this may be a long standing x86 ABI incompatibility.
> 
> So, is it agreed that the 3.1.1 behavior is correct?  Matching the
> behavior of C is a good thing.  To me, matching the behavior of Intel's
> compiler is low priority, but preserving the C ABI with respect to older
> GCC versions is vital.
This is the conclusion Mark and I came to.

nathan
-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: draft release notes for GCC 3.2

[Andreas Jaeger]

Joe Buck <Joe.Buck@synopsys.com> writes:

> There is one platform-specific bug fix (for x86_64 - AMD's 64-bit

The platform is called x86-64.  The underscore is there just for
config.guess.  Can you change the documentation, please?

> [...]
> x86_64 specific:
> 7291: off-by-one in generated inline bzero code for x86_64

Thanks for making these notes,
Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj