Re: [avr-gcc-list] compiler bug?

[Graham Davies]

Richard F wrote:
[in] rtc.time_element.tm_wday = (rtc.time_element.tm_wday < 6) ? 
rtc.time_element.tm_wday++ : 0;
... the expression rtc.tm_wday++ doesn't appear to be evaluated.

If the conditional expression evaluates TRUE, the statement reduces to this:
rtc.time_element.tm_wday = rtc.time_element.tm_wday++;

The C language is defined such that the RHS will be evaluated completely 
before assignment to the variable on the left.  So what happens is that the 
value of rtc.time_element.tm_wday is retrieved and stored, 
rtc.time_element.tm_wday is then incremented and, finally, the stored value 
(the original value of rtc.time_element.tm_wday) is assigned to 
rtc.time_element.tm_wday (wiping out the increment).  The compiler is at 
liberty to optimize away the redundant increment (unless 
rtc.time_element.tm_wday is declared volatile, in which case it would be 
incremented and then put back where it was).

Graham.