[C/C++ PATCH] Really disable fixed-point support on targets that don't support it and in C++ (PR inline-asm/39059)

[Sean D'Epagnier]

Hi,

I have been working on fixed-point for avr.  I have a complete
functional patch that gives native support.  As I know many people use
C++ on the avr backend, and I want them to be included in the people
who can benefit from fixed point.

>> The C++ FE doesn't support fixed point at all (_Fract/_Accum/_Sat keywords
>> aren't recognized at all), but again as fixed point constants made it
>> through with no diagnostics, we've already had a lot of various ICEs with
>> use of 0r and 0k in throw, template parameters, mangling etc.

Isn't this two problems?
First we need to support fixed point in the FE.  Next, why does 0r
cause problems?  Do you have an example?  Is this problem possible
overcome?

Thanks,
Sean

On 2/19/09, Jason Merrill <jason@redhat.com> wrote:
> Jakub Jelinek wrote:
>> I think that wasn't wasted effort.  When somebody adds fixed-point support
>> for C++, it can be used.  But really until somebody adds full
>> support for fixed-point to C++, supporting just fixed point constants and
>> not the their types is just a source of ICEs, nothing else.
>
> My impression was that the various ICEs would be pretty trivial to fix.
>   But I guess I don't actually care much about fixed point.  Is MIPS
> interested in fixing these bugs?
>
> Jason
>
>