[aarch64] assert(is_power_of_2(x)) failed: x must be a power of 2: 0xffffffff80000000

[Schmidt, Lutz]

On 24.02.20, 17:09, "Andrew Haley" <aph at redhat.com> wrote:

    On 2/24/20 3:51 PM, Schmidt, Lutz wrote:
    > I would locate the real bug in is_power_of_2(jint). Why?
    >
    > The function is passed in a jint (= signed int) of value 0x8000_0000. This clearly is a negative number. There does not exist a negative number which is a power of 2.
    
    I'm pretty sure you're wrong. jint multiplication is defined as being
    the low-order 32 bits of the true result; i.e. it is modulo 2**32. It
    is different from C++ arithmetic in this regard, in which integer
    arithmetic may overflow. In Java, if you start with 1 and multiply it
    by 2, 31 times, you get exactly 0x8000_0000. The inverse operation of
    those multiplication is exact_log2, and given that the result of the
    multiplications is well defined so is its inverse, exact_log2().

OK. That's a different way to look at it.
    
    > Separate from that, exact_log2(jint) sign-extends the passed-in
    > value to long, creating a bit pattern which obviously is not a power
    > of two.
    
    And in my opinion that sign extension should not be done in the case
    of a jint argument. We should try to do the most useful thing.

I agree. Just didn't say it explicitly. 
    
    > A possible fix would include:
    >  - make is_power_of_2() and exact_log_2() sign-aware.
    >  - call is_power_of_2() and exact_log_2() with arguments of proper
    > type (juint in this case).
    
    Or fix exact_log2(jint) by specializing it appropriately.

Yes, by treating the jint argument as unsigned. 
    
    -- 
    Andrew Haley  (he/him)
    Java Platform Lead Engineer
    Red Hat UK Ltd. <https://www.redhat.com>
    https://keybase.io/andrewhaley
    EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671