Re: EH patch for libstdc++

[Richard Earnshaw <Richard dot Earnshaw at buzzard dot freeserve dot co dot uk>]

On Sun, 2007-06-10 at 12:33 +0100, Richard Earnshaw wrote:
> On Fri, 2007-06-08 at 10:27 +0200, Benjamin Kosnik wrote:
> > > Daniel Jacobowitz pointed out that your EH patch for the ARM EH code:
> > > 
> > >   http://gcc.gnu.org/ml/gcc-patches/2007-05/msg01572.html
> > > 
> > > had never been officially reviewed.  (I had mistakenly seen Paul's
> > > review, and thought that was it -- but since this is libstdc++, Paul
> > > can't approve the patch.)  I've looked at it, and it looks sane to me
> > > as well, so please commit it.
> > 
> > I would prefer more minimal arm eh patching of eh_personality. See
> > attached.
> 
> Except that your more minimal version doesn't even compile, for the same
> reasons that Jason's original patch doesn't compile.
> 
> /home/rearnsha/gnusrc/gcc-cross/trunk/libstdc++-v3/libsupc
> ++/eh_personality.cc: In function '_Unwind_Reason_Code
> __cxxabiv1::__gxx_personality_v0(_Unwind_State, _Unwind_Control_Block*,
> _Unwind_Context*)':
> /home/rearnsha/gnusrc/gcc-cross/trunk/libstdc++-v3/libsupc
> ++/eh_personality.cc:548: warning: dereferencing type-punned pointer
> will break strict-aliasing rules
> /home/rearnsha/gnusrc/gcc-cross/trunk/libstdc++-v3/libsupc
> ++/eh_personality.cc:548: error: cannot convert 'const std::type_info*'
> to '_throw_typet*' in assignment
[...]

So how do we proceed on this?  I'm taking your counter proposal (even
though it doesn't work) as a rejection of my proposed patch.

I'd like to reach closure on this ASAP, building C++ for arm-eabi has
been broken for over a month now.

R.