[avr-libc-dev] RE: [bug #24231] AVR Mistakes Value in Register

[Weddington, Eric]

> -----Original Message-----
> From: Chris Troutner [mailto:address@hidden 
> Sent: Tuesday, September 09, 2008 3:27 PM
> To: Eric Weddington
> Cc: Weddington, Eric; address@hidden
> Subject: Re: [bug #24231] AVR Mistakes Value in Register
> 
> I'm using the latest, 1.6.2 of AVR Libc. As to the rest, I'm 
> using whatever comes with the latest 20080610 release of WinAVR. 
> 
> I've noticed some discrepancy in simulation behavior in AVR 
> Studio between the Simulator, Simulator 2, and use of JTAG 
> mkII for emulation. 
> 
> If you set a break point at the "service = 0" line, and then 
> go to dissassebly when the simulator reaches that point, you 
> should see it call the "STS 0x0100,R1" instruction. Are you 
> not seeing it on your machine?
> 
> I was able to get around it by using the instruction "service 
> &= ~(1<<0)".
> 

One thing at a time.
At this point I'm only concerned about the compiler, not the AVR Studio 
Simulator 1/2, or JTAG debugging.

I see the instruction in e.g., the .lss file:
                                //Clear service variable
                                service = 0;
 230:   10 92 00 01     sts     0x0100, r1

What I do *not* see in the assembly is r1 getting set to some other value other 
than 0. So I'm still not sure about your claim that this a compiler error.

Are you compiling with -Os and then debugging? If so, then turn of 
optimizations by specifying -O0, and then debug. 

Can you find in the assembly where r1 gets set to 1?