RFR: 8253891: Debug x86_32 builds fail after JDK-8239090

[Thomas Stuefe]

On Thu, 1 Oct 2020 09:52:54 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> `CPU_MAX_FEATURE` is actually a `uint64_t`, with at least 46 bits set. `exact_log2` expects `intptr_t`. The implicit
> conversion works on 64-bit, but fails on 32-bit. Calling to `exact_log2_long` seems to cater for both bitnesses.
> Testing:
>   - [x] tier1 on Linux x86_64
>   - [x] tier1 on Linux x86_32 (some unrelated failures)

Looks good. I like the plural of bitness.
jlong is signed and intptr_t is unsigned, but I don't think that is a problem here. I see a gtest for the
exact_log2_long testing all bits. Would be clearer though to have an uint64_t variant.

-------------

Marked as reviewed by stuefe (Reviewer).

PR: https://git.openjdk.java.net/jdk/pull/455