Re: [avr-gcc-list] Optimizer bug, again?

[David Brown]

Erik Zane wrote:
> I'm working on Ubuntu, and I've downloaded GCC 4.3.0 and applied all the 
> patches delivered with WinAVR, and it appears that there is what could 
> be a reappearance of an optimization bug.  Is the distribution missing a 
> patch, or do I have to start trying to figure out where my code is broken.
>
> My problem seems to occur when I try to send a message to my application 
> requesting a USART bitrate change.  If I had to hazard a guess, it seems 
> that the target has a hard time dealing with the 32-bit data and 
> generates a bad value for UBRR register.  Once that happens, it's time 
> to reset.  Unfortunately, one of my applications is essentially a 4-way 
> soft UART, which gets the problem doubled.
>
> For the record, it seems fine if I compile with -O0 on 4.3.0, and I 
> still have one of my PCs with a 4.2.2 avr-gcc in which I compiled doing 
> the same as I did above for 4.3.0, and anything of mine compiled with 
> that continues to work. as expected.

Can you isolate the problem and post minimal sample code illustrating 
the effect?  There is almost no way anyone can help without such details.