[cxx-abi-dev] C++0x: mangling of char16_t and char32_t

[Doug Gregor]

On Tue, Mar 25, 2008 at 2:13 PM, Michael Wong <michaelw at ca.ibm.com> wrote:
> I would prefer if we hold off using q for now, as it is in a contentious
>  mode for one of the DFP manglings which I still have to discuss with other
>  C++ ABI members for an alternative:
>  Currently IBM and I think GNU (but in 4.3) uses these for name mangling of
>  DFP builtin types:
>  >>     Decimal32:          "p"
>  >>     Decimal64:          "q"
>  >>     Decimal128:         "r"

Ah, I remember this discussion. I don't think GCC has a mangling for
these yet, but if IBM is already using 'q' then we need to be careful
here.

>  One of the issue is the same which is a wish for a more descriptive
>  mangling. Others include whether we really care about interoperability
>  given the different underlying representations, emulation modes vs hardware
>  modes issues.

It makes life easier for other tool vendors if name mangling is the
same across different compilers.

>  So in summary, I would appreciate it if you would avoid this area until we
>  can settle this issue.

GCC recently received a patch that adds support for char16_t and
char32_t, so that forces the issue a bit.

  - Doug