Ah yes, in assembler it is easy, but it can be done in C.
I had been relying on failing human memory, instead of looking it up.
The principle is there, and it works, but my memory was faulty.
I had to look it up.
In main SP = x
Serial int
In SerialISR SP = x + 2
void Serialint (void){
....
//TX part
sendchar(char ch);
}
void sendchar (char ch){
freeint();
...
}
void freeint (void){
#pragma asm
reti
#endpragma asm
}
In sendchar SP = x+4
In Freeint SP = x+6
Reti
SP = x + 4 = sendchar
ret from sendchar SP = x+2
In serialISR
reti from serialISR
SP = x
Back at main
You are correct, the reti will be done twice.
The point is that reti is the same as ret software wise, but additionally it handles the hardware interrupt unlock.
So you unlock the interrupts twice. No problem.
References:
http://www.embedded.com/columns/t.../190302110
www.state-machine.com/8051/QDKn_8051-IAR.pdf
Regards
Johan Smit