Re: [Qemu-devel] [Bug 1655708] [NEW] target/ppc/int_helper.c:2806: strange expression ?

[joserz@linux.vnet.ibm.com]

On Wed, Jan 11, 2017 at 05:12:38PM -0000, Eric Blake wrote:
> On 01/11/2017 10:41 AM, dcb wrote:
> > Public bug reported:
> > 
> > target/ppc/int_helper.c:2806:25: warning: ‘*’ in boolean context,
> > suggest ‘&&’ instead [-Wint-in-bool-context]
> > 
> > Source code is
> > 
> >        zone_digit = (i * 2) ? b->u8[BCD_DIG_BYTE(i * 2)] >> 4 :
> > zone_lead;
> 
> Also, looking at BCD_DIG_BYTE():
> 
> #if defined(HOST_WORDS_BIGENDIAN)
> #define BCD_DIG_BYTE(n) (15 - (n/2))
> #else
> #define BCD_DIG_BYTE(n) (n/2)
> #endif
> 
> Oops. n is under-parenthesized, and will cause invalid expansions for
> some expressions.  Let's fix that as well.
> 
Hello guys!!

Do you mean something like:

BCD_DIG_BYTE(n) (15 - ((n)/2)) and BCD_DIG_BYTE(n) ((n)/2) ?

to avoid issues like BCD_DIG_BYTE(i + 3) that will expand to (i + 3/2)
instead of ((i + 3)/2), right?

> 
> > so I think the compiler warning is for the i * 2 lhs of the ?.
> 
> Yes - the compiler is complaining that 'i * 2' can only be non-zero if
> 'i' was non-zero (given that the code occurs in a loop for i between 0
> and 16), so it is just as easy to write 'i ? ...' instead of the weirder
> '(i * 2) ? ...'.
> 

Yes, Eric is correct it's meant to be "(i) ? .. : .." because the sign
is located at position 0. I'm improving it. I'll send a patch for them
today.

Thanks!