[jdk11u-dev] RFR: 8234930: Use MAP_JIT when allocating pages for code cache on macOS [v7]

[Vladimir Kempik]

On Fri, 10 Dec 2021 19:28:48 GMT, Vladimir Kempik <vkempik at openjdk.org> wrote:

>> Vladimir Kempik has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   put multiline code difference under ifdefs instead of MACOS_ONLY
>
> I think the review from Thomas is enough, yours is optional, still nice to have,

> @VladimirKempik, SAP nightlies ran through with your patch on Windows x64, all Linux platforms and Mac intel; no problems attributable that I could see. Note that the tests did not include MacOS aarch64, since we don't have the machines yet. But I expect you tested that. We also don't test 32-bit, Solaris, or AIX. AIX is supposedly tested by IBM now since SAP stepped away from it.

Not exactly, none can test mac-arm on openjdk11 as it's not present there yet. I have made a sanity check on zulu11, applying this patch to it. C2/C1 still works which means this patch doesn't break memory management for codecache on mac-arm.
Solaris is another issue, as we can't build solaris from openjdk11 repo here, only from zulu11 repo. ( due to harfbuzz issues on jdk11u-dev).
Since this patch equals to "nothing" on platforms other than macos, it should be safe for them.
> 
> > I think the review from Thomas is enough, yours is optional, still nice to have,
> 
> Personally, I'm fine with it, but I'm not aware of the rules here.
> 
> IMHO in maintenance releases we should at least have the same number of reviewers we require on Head. After all, we are much closer to production in update releases; once out this hits many customer systems immediately.
> 
> Maybe not for simple patches that get downported and had been tested already in newer releases. But at least for complex stuff or for patches we develop explicitly for older releases.
> 
> Ping to @GoeLin ...

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

PR: https://git.openjdk.java.net/jdk11u-dev/pull/710