[avr-gcc-list] maybe a bug in gcc-avr?

[lbuchman]

Hello together,

I've been thinking about something for the last days. Atm I guess the
bug of my problem is in the avr-gcc.

Here my code:

uint16_t total=0, temp=0;
uint8_t bufs[64];
uint8_t bufr[64];
...
if (usb_rawhid_recv(bufr, 50) == 64) {
#if 1 //depending on 'if that calculation is done', the result of bufs[0] will 
be 0x01 or 0x02...
        total = (((uint16_t)bufr[0]) << 8) + ((uint16_t) (bufr[1] & 0xFF));
#endif
temp = (((uint16_t)bufr[0]) << 8); //temp is 0x0200
bufs[0] = temp >> 8;
bufs[1] = (((uint16_t)bufr[0]) << 8) >> 8;

Well, if bufr[0] is 0x02, then bufs[1] is also 0x02. That works fine.
But bufs[0] is calculated wrong IN CASE that i also calculate the variable 
"total" inside the "#if 1".
(bufs[0] is then 0x01 instead of 0x02)

The two assemblies are vimdiff'ed and shown in this image:
http://www10.pic-upload.de/28.11.12/1afla5oyatlv.jpg

On the left side the working case (where total isn't calculated due to the "#if 
0")
On the right side is my problem shown. I calculate the variable "total" but gcc
seems to ignore my typecasts.

What do you think?
I hope the mailing list is the right place, where I should post my
problem :)

Have a nice evening :)