Re: [Valgrind-developers] Fixes for regression failures, part 1 arch independent

[John R. <jr...@bi...>]

On 2/14/20 13:25 UTC, Mark Wielaard wrote:
> On ppc64[be] I do see an issue with the vbit-test:

The test itself is incorrect.  First of all, the expected value is known but not shown.
Second, for a bit-wise OR operation, if either input has a valid '1' bit in some bit position,
then the result also has a valid '1' bit in that bit position.  An input bit that is a valid '1',
makes the other input bit in that bit position into a "don't care" bit.
[A similar rule applies to AND operation with a valid '0' bit.]

Corrected lines are shown below the incorrect ones:

> 
> *** Incorrect result for operator Iop_Or1
>      opnd 0:  vbits = 00000001   value = 00
>      opnd 1:  vbits = 00000000   value = 01
>      result:  vbits = 00000001   value = 00
>      expect:  vbits = 00000000
        expect:  vbits = 00000000   value = 01
> *** Incorrect result for operator Iop_Or1
>      opnd 0:  vbits = 00000000   value = 01
>      opnd 1:  vbits = 00000001   value = 01
>      result:  vbits = 00000001   value = 00
>      expect:  vbits = 00000000
        expect:  vbits = 00000001   value = 01
> *** Incorrect result for operator Iop_Or1
>      opnd 0:  vbits = 00000001   value = 01
>      opnd 1:  vbits = 00000000   value = 00
>      result:  vbits = 00000001   value = 00
>      expect:  vbits = 00000000        expect:  vbits = 00000001   value = 01
> *** Incorrect result for operator Iop_Or1
>      opnd 0:  vbits = 00000000   value = 00
>      opnd 1:  vbits = 00000001   value = 00
>      result:  vbits = 00000001   value = 00
>      expect:  vbits = 00000000
        expect:  vbits = 00000000   value = 00