You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by johnnylj98 <jo...@gmail.com> on 2009/11/25 17:27:53 UTC
[DRLVM][JVMTI]a bug when running SpecJVM98 _201_compress
Hi all,
I'm doing some tuning work for Harmony running SpecJVM98.
I write a JVMTI agent to catch VMDeath event and do something at the end of JVM's life, but that doesn't work(the same agent work well for hotspot).
my agent likes bellow:
...
static void JNICALL
callbackVMDeath(jvmtiEnv *jvmti, JNIEnv *jni)
{
assert(false);
printf("vm death\n");
}
JNIEXPORT void JNICALL Agent_OnUnload(JavaVM *jvm)
{
assert(false);
printf("unload!");
}
JNIEXPORT jint JNICALL Agent_OnLoad(JavaVM *jvm, char *options, void *reserved)
{
...
error = jvmti->SetEventNotificationMode(JVMTI_ENABLE,JVMTI_EVENT_VM_DEATH, (jthread)NULL);
callbacks.VMDeath = &callbackVMDeath;
error = jvmti->SetEventCallbacks(&callbacks,(jint)sizeof(callbacks));
...
}
...
The program ends correctly. This is really strange. Can anyone try this case and explain for me?
Thanks a lot.
2009-11-26
johnnylj98
Re: Re: [DRLVM][JVMTI]a bug when running SpecJVM98 _201_compress
Posted by johnnylj98 <jo...@gmail.com>.
Thanks a lot! I modified the source code of SpecJVM and the agent works well now~!
2009-11-27
johnnylj98
发件人: Gregory Shimansky
发送时间: 2009-11-26 02:46:12
收件人: dev
抄送:
主题: Re: [DRLVM][JVMTI]a bug when running SpecJVM98 _201_compress
IIRC VMDeath is not sent if System.exit is used by application. This is a
limitation of JMVTI implementation in DRLVM. Most likely Spec uses exit to
finish.
> Hi all,
> I'm doing some tuning work for Harmony running SpecJVM98.
> I write a JVMTI agent to catch VMDeath event and do something at the end of
> JVM's life, but that doesn't work(the same agent work well for hotspot).
>
> my agent likes bellow:
> ...
> static void JNICALL
> callbackVMDeath(jvmtiEnv *jvmti, JNIEnv *jni)
> {
> assert(false);
> printf("vm death\n");
> }
> JNIEXPORT void JNICALL Agent_OnUnload(JavaVM *jvm)
> {
> assert(false);
> printf("unload!");
> }
> JNIEXPORT jint JNICALL Agent_OnLoad(JavaVM *jvm, char *options, void
> *reserved) {
> ...
> error = jvmti->SetEventNotificationMode(JVMTI_ENABLE,JVMTI_EVENT_VM_DEATH,
> (jthread)NULL); callbacks.VMDeath = &callbackVMDeath;
> error = jvmti->SetEventCallbacks(&callbacks,(jint)sizeof(callbacks));
> ...
> }
> ...
> The program ends correctly. This is really strange. Can anyone try this
> case and explain for me?
>
> Thanks a lot.
>
> 2009-11-26
>
>
>
> johnnylj98
>
Re: [DRLVM][JVMTI]a bug when running SpecJVM98 _201_compress
Posted by Gregory Shimansky <gs...@gmail.com>.
IIRC VMDeath is not sent if System.exit is used by application. This is a
limitation of JMVTI implementation in DRLVM. Most likely Spec uses exit to
finish.
> Hi all,
> I'm doing some tuning work for Harmony running SpecJVM98.
> I write a JVMTI agent to catch VMDeath event and do something at the end of
> JVM's life, but that doesn't work(the same agent work well for hotspot).
>
> my agent likes bellow:
> ...
> static void JNICALL
> callbackVMDeath(jvmtiEnv *jvmti, JNIEnv *jni)
> {
> assert(false);
> printf("vm death\n");
> }
> JNIEXPORT void JNICALL Agent_OnUnload(JavaVM *jvm)
> {
> assert(false);
> printf("unload!");
> }
> JNIEXPORT jint JNICALL Agent_OnLoad(JavaVM *jvm, char *options, void
> *reserved) {
> ...
> error = jvmti->SetEventNotificationMode(JVMTI_ENABLE,JVMTI_EVENT_VM_DEATH,
> (jthread)NULL); callbacks.VMDeath = &callbackVMDeath;
> error = jvmti->SetEventCallbacks(&callbacks,(jint)sizeof(callbacks));
> ...
> }
> ...
> The program ends correctly. This is really strange. Can anyone try this
> case and explain for me?
>
> Thanks a lot.
>
> 2009-11-26
>
>
>
> johnnylj98
>