On 27/08/2024 1:50 pm, Jan Beulich wrote:
> On 27.08.2024 14:01, Andrew Cooper wrote:
>> On 26/08/2024 11:30 am, Jan Beulich wrote:
>>> On 23.08.2024 01:06, Andrew Cooper wrote:
>>>> --- a/xen/common/bitops.c
>>>> +++ b/xen/common/bitops.c
>>>> @@ -84,8 +84,32 @@ static void __init test_fls(void)
>>>> CHECK(fls64, 0x8000000000000001ULL, 64);
>>>> }
>>>>
>>>> +static void __init test_multiple_bits_set(void)
>>>> +{
>>>> + /*
>>>> + * multiple_bits_set() is generic on the type of it's parameter, as
>>>> the
>>>> + * internal expression is so simple.
>>>> + */
>>>> +
>>>> + CHECK(multiple_bits_set, 0, false);
>>>> + CHECK(multiple_bits_set, 1, false);
>>>> + CHECK(multiple_bits_set, 2, false);
>>>> + CHECK(multiple_bits_set, 3, true);
>>>> +
>>>> + CHECK(multiple_bits_set, 1 | (1UL << (BITS_PER_LONG - 1)), true);
>>> This is really the same as ...
>>>
>>>> +#if BITS_PER_LONG > 32
>>>> + CHECK(multiple_bits_set, 1 | (1UL << 32), true);
>>>> + CHECK(multiple_bits_set, 1 | (1UL << 63), true);
>>> ... this, at least as long as BITS_PER_LONG > 32 in practice means
>>> BITS_PER_LONG == 64. Perhaps not really worth keeping that line?
>> I suppose not. I'll drop it.
>>
>> However, It occurs to me that I do need a test of 0x8000000000000001ULL
>> mostly for 32bit builds to check that even when the argument is split,
>> the answer is still accurate.
> IOW you'll insert an #else in the middle. Fine with me; keep the R-b.
Oh, I already had that case, out of context below what you quoted. I'll
just drop the one intermediate test.
~Andrew
