C implementations of the C++ ABI

[Christophe de Dinechin]

Mark Mitchell wrote:
> 
> 
> All in all, I'm not really suggesting we change anything.  I'm just
> trying to make sure we're aware that there's an issue, in case we're
> not already aware.
> 
We discussed it, and we have at least one company (EDG) in the team who may have
the need to generate C code in some cases. I think the agreement is that we do
care, but that a C-generating tool will either not be ABI compliant in some
cases, or require specific extensions (such as inline assembly) to work.

The case of EH tables is particularly interesting: any C-compatible EH model I
know of has a definite runtime cost (something like setjmp/longjmp, for
instance). We don't want to pay that price.


Best regards,
Christophe