Re: AW: [avr-gcc-list] jump_to_null question

[Dr. Jan Menzel]

Haase Bjoern (PT/EMM1) wrote:
> Just one question: Did you extract the line
>
> >         RCALL   PC-0x0009
>
> > From disassembling the completeley linked executable or from disassembly of the 
> > object file?
> Possibly the linker might have replaced (PC-0x0009) by the true distance 
> between the PC position of the PC at the RCALL.?
I used the disassembly windows while single stepping in AVR Studio. I 
guess its the linked binaries disassembly.
	The main problem still remains: why is the complier jumping to a 
different position if the position is defined as const...

> -----Ursprüngliche Nachricht-----
> Von: address@hidden [mailto:address@hidden Im Auftrag von Eric Weddington
> Gesendet: Donnerstag, 24. Mai 2007 15:17
> An: 'Dr. Jan Menzel'; address@hidden
> Betreff: RE: [avr-gcc-list] jump_to_null question
>
>  
>
> > -----Original Message-----
> > From: 
> > address@hidden 
> > [mailto:address@hidden
> > org] On Behalf Of Dr. Jan Menzel
> > Sent: Thursday, May 24, 2007 6:40 AM
> > To: address@hidden
> > Subject: [avr-gcc-list] jump_to_null question
> >
> > Hi all!
> > 	I'm currently working on a bootloader for Mega8 and 88 
> > and found a 
> > problem I do not yet understand. At the end of the bootloader the code
> > has to jump to the main reset vector at address 0x0000. If I do the
> > usual way
> >         void (*jump_to_null)(void) = 0x0000;
> > gcc 3.4.6 complies:
> >         LDI     R30, 0x0102
> >         LDI     R31, 0x0103
> >         ICALL
> > and everything works fine, as expected.
> >         But if I try to be more specific and define
> >         void (* const jump_to_main)(void) = 0x0000;
> > gcc 3.4.6 complies
> >         RCALL   PC-0x0009
> > which unfortunately does not what I intended.
> > 	Could anyone please point me to my error in reasoning 
> > or lack of 
> > understanding? Many thanks!
>
> 1. Why do you need to declare it const, when the first way works fine?
> 2. The best way to reset the AVR, is not by jumping to the reset vector.
> When you do that, some of the registers may be left in a state that is not
> the default. The best way to reset is to enable the watchdog timer on the
> shortest time-out, go into an infinite loop, and let the watchdog reset the
> AVR. Depending on which AVR you are using, you will have to turn off the
> watchdog timer after reset, with code in the .init3 section, for example.
>
> Eric 
>
>
>
> _______________________________________________
> AVR-GCC-list mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/avr-gcc-list