On 2020-04-16 04:37, coleen.phillim...@oracle.com wrote:
On 4/15/20 9:37 PM, David Holmes wrote:
Hi Coleen,
On 16/04/2020 10:59 am, coleen.phillim...@oracle.com wrote:
open webrev at
http://cr.openjdk.java.net/~coleenp/2020/8242896.01/webrev
bug link https://bugs.openjdk.java.net/browse/JDK-8242896
Looks good but ...
Built and ran vmTestbase RedefineTests which use the protected code.
... you need to ensure that builds that don't INCLUDE_JVMTI still
work okay as they will now actually be excluding this code for the
first time.
Last time I tried to build minimal, it failed for some odd problem on
my system. I'll wait until someone from the openjdk can build it then.
thanks,
Building minimal should work. Try something like this: "jib configure
-- --with-jvm-variants=minimal". It should work.
I just tried with your patch, and it does not work.
/localhome/hg/jdk-BAR/open/src/hotspot/share/classfile/metadataOnStackMark.cpp:70:
error: undefined reference to 'CodeCache::old_nmethods_do(MetadataClosure*)'
/localhome/hg/jdk-BAR/open/src/hotspot/share/code/nmethod.cpp:1495:
error: undefined reference to
'CodeCache::unregister_old_nmethod(CompiledMethod*)'
I made a quick check but it was not clear to me if the call sites in
metadataOnStackMark.cpp and nmethod.cpp should be excluded if missing
jvmti, or if the code in codeCache.cpp should really be present even
with jvmti.
/Magnus
Coleen
Thanks,
David
thanks,
Coleen