Hi Markus,
The fix looks good.
Thanks,
Serguei
On 3/2/16 16:15, Markus Gronlund wrote:
Greetings,
Could a please ask for reviews for the following simple fix to resolve
a test issue associated with
test/java/lang/instrument/NativeMethodPrefixAgent.java
Bug: https://bugs.openjdk.java.net/browse/JDK-8151100
Webrev/diff:
diff --git a/test/java/lang/instrument/NativeMethodPrefixAgent.java
b/test/java/lang/instrument/NativeMethodPrefixAgent.java
--- a/test/java/lang/instrument/NativeMethodPrefixAgent.java
+++ b/test/java/lang/instrument/NativeMethodPrefixAgent.java
@@ -31,7 +31,7 @@
* java.management
* java.instrument
* @run shell/timeout=240 MakeJAR2.sh NativeMethodPrefixAgent
NativeMethodPrefixApp 'Can-Retransform-Classes: true'
'Can-Set-Native-Method-Prefix: true'
- * @run main/othervm -javaagent:NativeMethodPrefixAgent.jar
NativeMethodPrefixApp
+ * @run main/othervm -XX:-CheckIntrinsics
-javaagent:NativeMethodPrefixAgent.jar NativeMethodPrefixApp
*/
import java.lang.instrument.*;
Description:
The java/lang/instrument/NativeMethodPrefixAgent.java modifies (wraps)
the names of native methods.
If a class is loaded that contain the @HotSpotIntrinsicCandidate
annotation on a native method, one step of the class file parser is an
attempt to validate the method name against a registered intrinsic in
the VM.
By default, the flag CheckIntrinsics is true, which will yield:
“Method
[<class>.wrapped_tr0_wrapped_tr1_wrapped_tr2_<original_method_name>()<return>]
is annotated with @HotSpotIntrinsicCandidate, but no compiler
intrinsic is defined for the method. Exiting.”
This "wrapped" prepending to the method name invalidates the intrinsic
check.
In order for the test to pass even when loading a class with a native
method containing the @HotSpotIntrinsicCandidate annotation, the test
should explicitly disable the CheckIntrinsics flag:
-XX:-CheckIntrinsics
Thanks in advance
Markus
