RFR(xs) 8232030: HelloDynamic.java fails with latest Graal
Ioi Lam
IOI.LAM at ORACLE.COM
Thu Oct 17 20:08:11 UTC 2019
Hi Calvin, this looks good. Thanks!
Ioi
> On Oct 17, 2019, at 11:50 AM, Calvin Cheung <calvin.cheung at oracle.com> wrote:
>
> bug: https://bugs.openjdk.java.net/browse/JDK-8232030
>
> During CDS dynamic dumping and during the loading of a class whose super class has default methods, the method_ordering is not set correctly after merging in the default methods. A simple fix is in the merge_in_new_methods() of defaultMethods.cpp to include the dynamic dumping condition when setting up the merged_ordering array.
>
> bash-4.2$ hg diff src/hotspot/share/classfile/defaultMethods.cpp
> diff --git a/src/hotspot/share/classfile/defaultMethods.cpp b/src/hotspot/share/classfile/defaultMethods.cpp
> --- a/src/hotspot/share/classfile/defaultMethods.cpp
> +++ b/src/hotspot/share/classfile/defaultMethods.cpp
> @@ -1012,7 +1012,7 @@
> klass->class_loader_data(), new_size, NULL, CHECK);
>
> // original_ordering might be empty if this class has no methods of its own
> - if (JvmtiExport::can_maintain_original_method_order() || DumpSharedSpaces) {
> + if (JvmtiExport::can_maintain_original_method_order() || Arguments::is_dumping_archive()) {
> merged_ordering = MetadataFactory::new_array<int>(
> klass->class_loader_data(), new_size, CHECK);
> }
>
> Tested manually with the vmoptions listed in the bug report.
>
> Also ran tier1 - 2 tests.
>
> thanks,
>
> Calvin
>
More information about the hotspot-runtime-dev
mailing list