AVR-Chat

Is there something unusual I should know with this device?

I'm watching it step through a simple routine to light some LEDs, and
after stepping through a single CBI instruction it's telling me that
half of SRAM has been changed..

There are four alarm timers, and if any of them is > 0, I want to
light the appropriate LED, if the timer is Zero, then turn the LED
off.

Sane at this point:

1638:     	if (A_Alarm_Timer > 0) {
+000004C1:   9180022E    LDS       R24,0x022E     Load direct from data space
+000004C3:   2388        TST       R24            Test for Zero or Minus
+000004C4:   F019        BREQ      PC+0x04        Branch if equal
1639:     		LED_DDR		|=  (1 << LED_1);		// Output
+000004C5:   9A24        SBI       0x04,4         Set bit in I/O register
1640:     		LED_Port	|=  (1 << LED_1);		// High	
+000004C6:   9A2C        SBI       0x05,4         Set bit in I/O register
+000004C7:   C001        RJMP      PC+0x0002      Relative jump
1642:     		LED_Port  &= ~(1 << LED_1);			// Low
+000004C8:   982C        CBI       0x05,4         Clear bit in I/O register
1646:     	if (A_Warn_Timer > 0) {
+000004C9:   91800230    LDS       R24,0x0230     Load direct from data space
+000004CB:   2388        TST       R24            Test for Zero or Minus
+000004CC:   F019        BREQ      PC+0x04        Branch if equal
1647:     		LED_DDR		|=  (1 << LED_2);		// Output
+000004CD:   9A25        SBI       0x04,5         Set bit in I/O register
1648:     		LED_Port	|=  (1 << LED_2);		// High	
+000004CE:   9A2D        SBI       0x05,5         Set bit in I/O register
+000004CF:   C001        RJMP      PC+0x0002      Relative jump

Sane to this point

1650:     		LED_Port  &= ~(1 << LED_2);			// Low
+000004D0:   982D        CBI       0x05,5         Clear bit in I/O register

BOOM. Total corruption of SRAM, stack pointer blown, all hope abandon
ye who enter here..


The device itself, without the ICE, is crashing when I run this
routine, so I'm thinking it muse be some wild config problem.
I'm running 20 MHz external crystal, and that's confirmed by the 1mS
interrupt and serial baud rate being correct.

Fuses:
BODLEVEL is set at 4.3V, (VCC is solid at 5.05V)
JTAGEN is checked
SPIEN is checked
EESAVE is checked
BOOTSZ is 512 words (I'm not using it)
SUT_CKSEL is Ext Xtal osc, 8.0 -  16k clock + 65mS



-- 
There is no computer problem which cannot be solved by proper
application of a sufficiently large hammer.