Re: [MiNT] TimerA interrupt handler

["markus@mubf.de" <markus@mubf.de>]

On 04.01.2013 18:31, Vincent Rivière wrote:
> Anyway, I agree that such compiler dependent features are dangerous 
> when we don't know exactly what happens behind the scenes. Looking at 
> the generated assembler code is always a good practice.
There's another trap you'll need to take care of. 
__attribute__((interrupt)) is perfect to write trap handlers and usually 
doesn't cause any headaches there.

If you want to write "real" interrupt handlers, however, you must be 
aware that there is no way (at least none that I knew of) to atomically 
adjust IPL on handler entry as the very first instruction (which you 
must do to be sure your interrupt handler doesn't get called while it's 
already handled) since there is no way to intercept the gcc function 
prologue generated (to mask interrupts writing to SR).

In my opinion "real" interrupt handlers are better written in pure 
asssembler.