RFR(XS): 8232616: JVMCI_lock fails to get initialized when cds is disabled

[David Holmes]

Hi Jie,

That fix looks good to me too. Looks like the nesting inside the 
INCLUDE_CDS ifdef was unintentional.

Thanks,
David

On 18/10/2019 10:35 pm, Jie Fu wrote:
> Hi all,
> 
> JBS:    https://bugs.openjdk.java.net/browse/JDK-8232616
> Webrev: http://cr.openjdk.java.net/~jiefu/8232616/webrev.00/
> 
> When cds was disabled, many tests in compiler/aot and compiler/jvmci 
> crashed due to SIGSEGV in Mutex::unlock().
> For the list of failing tests, please refer to the JBS.
> 
> The reason for these crashes is that JVMCI_lock fails to get initialized 
> when cds is disabled [1].
> 
> I suggest to fix this bug since aot and jvmci seem not to be dependent 
> on cds.
> Could you please review it and give me some advice?
> 
> Testing:
>   - The list of failing tests due to this bug  on Linux/x64
>   - make test TEST="tier1" CONF=server-release on Linux/x64
> 
> Thanks a lot.
> Best regards,
> Jie
> 
> [1] 
> http://hg.openjdk.java.net/jdk/jdk/file/2e0462f4cf4e/src/hotspot/share/runtime/mutexLocker.cpp#l339 
> 
> 
>