Re: [avr-libc-dev] Better Interrupt Macro

[Dean Camera]

> Will this allow for the following syntax:
>
>
>   ISR(VectorName, ISR_NOBLOCK)
>   {
>   // Code goes here
>   }
>
>
> Wouldn't this set both the 'interrupt' and 'signal' attributes at the
> same time?

Yes indeed, very astute. the reason for declaring the ISR macro with the 
ISR_NOBLOCK (signal) is for backwards compatibility. If no second argument, 
the SIGNAL attribute is applied.

I have discovered in testing that having the signal as well as another 
ISR-related attribute specified at the same time works correctly and gives 
the expected output. If ISR_NOBLOCK is specified, then the resultant ISR 
contains a SEI as expected, if ISR_NAKED is specified, then the resulting 
code lacks a prologue or epilogue. Since it works fine, I don't think it is 
a problem. As a bonus, the SIGNAL attribute forces GCC to spell-check the 
vector name.

The only real issue with my macros as I can see it is a lack of a way to 
create an ISR_EMPTY equivalent. The only way I can think of doing this would 
be to create a "ISR_return()" macro ( #define ISR_return() __asm__ 
__volatile__ ("reti" ::); ) and then in the docs tell the user the correct 
way of doing so would be:

ISR(VectorName, ISR_NAKED)
{
   ISR_return();
}

Subject to individual's coding style, of course. I actually see this as a 
plus; the "ISR_return()" macro is in a similar style to the sei() and cli() 
macros, and so would be useful to those creating their own (non-empty) 
ISR_NAKED routines also. Although If the current convention was to be 
adhered to, the macro name would be "reti();", I suppose.

Thoughts?

- Dean