Ah, welcome to another in my file of insane bugs. It's not pressing, becuase when running a debug session with a JtagIceII, the problem goes away, so I can develop from a debug session. Non-optimum, of course.
Anyone ever have a mega644 that resets constantly? Details follow:
1) With WinAvr200601xx, it would happen with an 'empty' program, that just turned on an led ( i.e. main() { DDRC = 0x01; PORTC = 0x01; while(1) {} } )
2) Installed WinAvr200604xx(latest), suddenly that started working fine, so I put my serial port code back in which worked fine for the first 2 puts(), then started reseting again in the !middle! of the second puts()
3) MCUCSR shows 0, so it must be a bug, not Hw, right? I sure thought it was a runaway watchdog for a long time the way the cpu reset in the middle of a puts (and yes, it's a 20 char string in a 128 byte buffer, so it's not that). Also "WDT always on" fuse shows blank in avrstudio
4) call cli() from .init3, and have a default interrupt handler that just returns (verified in .lst file). Uart code is polled and is used in several projects without error.
5) Optimization off. -O0 (-Os same behavior, just gets further through the line before resetting)
6) I also do some spi between the two puts(), but I can see that communication works perfectly in the debugger, and I get the expected result from the second puts() (doing an spi tx to a device, of AA, then readback, then 55, then readback, and if match, print one thing, no match print something else).
7) prog size is under 4000 in a mega644
8) Single stepping in avrstudio, entire program will execute, including the part where it's supposed to remain indefinately in the while(1) loop at the end. The second you F5 (run), it starts resetting again, until you do the "reset" command from the debug menu.
Maybe someone's been here before...;) At least I don't have any deadlines for this (hobby project this time around).