Default logging is very verbose

[michielproce]

If you're experienced with DCEVM, HotswapAgent is easy to get running.
My only problem with the default is that the logging is a very verbose. I'm having trouble following my own application's logging.

Most of the logging looks like this:

(-1 / -1) (org/hotswap/agent/plugin/jvm/AnonymousClassInfos$AnonymousClassInfoMatcher/org/hotswap/agent/plugin/jvm/AnonymousClassInfos$AnonymousClassInfoMatcher)
{instance class} 
 - klass: {other class}
{instance class} 
 - klass: {other class}

I've tried disabling the logging but it appears my LOGGER=warn in a hotswap-agent.properties in my classpath doesn't seem to be picked up.

(BTW: Thanks for a pretty great alternative to JRebel :))

[edudant]

Hi, can you please describe your setup? This logging is directly from DCEVM and is only related to Hotswap Plugin class redefinition.

Default HotswapAgent logging level is INFO.

Do you use latest DCEVM?

[michielproce]

I'm using JDK 1.7.0_51 64 bit (host: Win7), DCEVM 24.51-b03-dcvmfull-3 (latest on dcevm.github.io).
I'm running an webapplication on Tomcat 7.0.50.

I've noticed that the verbose logging does not happen at other webapps, so there might be something application specific going on. If I run the application with DCEVM but without HotswapAgent the logging stops.
How can I suppress these messages?

If you need additional information, let me know.

[skybber]

Logging Source of this stuff is definitively DCEVM or OpenJDK respectively. I did some private patch, because I hate this mess.

[victorpablosceruelo]

Do you have the patch somewhere?
Could you tell us how to patch?
Thanks in advance.

[skybber]

Comment following lines (this is for v7u85):

diff --git a/src/share/vm/classfile/systemDictionary.cpp b/src/share/vm/classfile/systemDictionary.cpp
index 39ec219..1321603 100644
--- a/src/share/vm/classfile/systemDictionary.cpp
+++ b/src/share/vm/classfile/systemDictionary.cpp
@@ -2032,7 +2032,7 @@ void SystemDictionary::check_constraints(int d_index, unsigned int d_hash,
       assert(check->klass_part()->oop_is_instance(), "noninstance in systemdictionary");
       if ((defining == true) && ((k() != check) && k->old_version() != check)) {
         ResourceMark rm(Thread::current());
-        tty->print_cr("(%d / %d) (%s/%s)", k->revision_number(), check->klass_part()->revision_number(), k->name()->as_C_string(), check->klass_part()->name()->as_C_string());
+        // tty->print_cr("(%d / %d) (%s/%s)", k->revision_number(), check->klass_part()->revision_number(), k->name()->as_C_string(), check->klass_part()->name()->as_C_string());
         k()->print();
         check->print();
         linkage_error = "loader (instance of  %s): attempted  duplicate class "
diff --git a/src/share/vm/oops/klass.cpp b/src/share/vm/oops/klass.cpp
index 19a9f5b..8361662 100644
--- a/src/share/vm/oops/klass.cpp
+++ b/src/share/vm/oops/klass.cpp
@@ -601,7 +601,7 @@ jint Klass::jvmti_class_status() const {
 void Klass::oop_print_on(oop obj, outputStream* st) {
   ResourceMark rm;
   // print title
-  st->print_cr("%s ", internal_name());
+  // st->print_cr("%s ", internal_name());
   obj->print_address_on(st);

   if (WizardMode) {
@@ -610,15 +610,15 @@ void Klass::oop_print_on(oop obj, outputStream* st) {
   }

   // print class
-  st->print(" - klass: ");
+  // st->print(" - klass: ");
   obj->klass()->print_value_on(st);
-  st->cr();
+  //st->cr();
 }

 void Klass::oop_print_value_on(oop obj, outputStream* st) {
   // print title
   ResourceMark rm;              // Cannot print in debug mode without this
-  st->print("%s", internal_name());
+  // st->print("%s", internal_name());
   obj->print_address_on(st);
 }

[victorpablosceruelo]

Thank you very much. It works perfectly for 7u79 too.

I've asked DCEVM developers to apply your patch or something similar when in production mode
dcevm/dcevm#86

Hope it helps other people.

[skybber]

My solution is hack. I didn't study the purpose, why the messages are printed directly to stdout. There should be another solution how to avoid this.

[PrabakaranSK]

@skybber - Would you be kind enough to patch the DCE VM source all by yourself, as it seems no one is assigned to fix this issue on DCEVM side?

[cchilds]

I have spent days trying to get dcevm to compile with no success. If someone here has or could get a build of java 7u79 full, with these logging fixes applied, I would love you forever.

[victorpablosceruelo]

cchilds, if you use debian/ubuntu you can use the following url as repository:
http://babel.ls.fi.upm.es/~vpablos/others/packages/debian/

It is 7u79 without patching, but it is the only version I have successfully patched.
The other ones gave me headaches ... :-(

If not, you can always download the package and extract the files from it.
By the way, eclipse Mars does not work with DCEVM.

[cchilds]

Thanks @vpablos, but unfortunately, I'm using Windows. Thanks anyway.

What issues are you having with Mars and DCEVM? I have used it with Mars quite a bit with no problems.