RFR (S) 8001538: hs_err file does not list anymore compiled methods in compilation events

[Christian Thalinger]

On Nov 13, 2012, at 7:53 AM, Vladimir Ivanov <vladimir.x.ivanov at oracle.com> wrote:

> http://cr.openjdk.java.net/~vlivanov/8001538/webrev.00/
> 
> Due to incorrect message buffer size calculation, a message may be truncated to 4/8 characters.
> 
> Also, changed how deoptimization events are logged:
>  - moved [1] from "Deoptimization events" to "Events" section.
>  - added [2] to "Deoptimization events" section
> 
> Sample output from a hs_err file:
> Deoptimization events (2 events):
> Event: 3.438 Thread 0x0000000103804800 Uncommon trap: Test::f @ 23, reason: unloaded, action: reinterpret
> Event: 3.482 Thread 0x0000000103804800 Uncommon trap: Test::main @ 38, reason: unloaded, action: reinterpret
> 
> Internal exceptions (10 events):
> ...
> 
> Events (10 events):
> Event: 3.441 Thread 0x0000000103804800 DEOPT UNPACKING pc=0x00000001046cf214 sp=0x0000000100700548 mode 2
> Event: 3.482 Thread 0x0000000103804800 Uncommon trap 3 fr.pc 0x00000001046ffe1c
> ...

I have similar change in one of my workspaces:

diff --git a/src/share/vm/runtime/deoptimization.cpp b/src/share/vm/runtime/deoptimization.cpp
--- a/src/share/vm/runtime/deoptimization.cpp
+++ b/src/share/vm/runtime/deoptimization.cpp
@@ -1241,10 +1241,6 @@
   // before we are done with it.
   nmethodLocker nl(fr.pc());
 
-  // Log a message
-  Events::log_deopt_message(thread, "Uncommon trap %d fr.pc " INTPTR_FORMAT,
-                            trap_request, fr.pc());
-
   {
     ResourceMark rm;
 
@@ -1265,6 +1261,11 @@
     int             trap_bci    = trap_scope->bci();
     Bytecodes::Code trap_bc     = trap_method->java_code_at(trap_bci);
 
+    // Log a message
+    Events::log_deopt_message(thread, "Uncommon trap: reason=%s action=%s pc=" INTPTR_FORMAT " method=%s @ %d",
+                              trap_reason_name(reason), trap_action_name(action), fr.pc(),
+                              trap_method->name_and_sig_as_C_string(), trap_bci);
+
     // Record this event in the histogram.
     gather_statistics(reason, action, trap_bc);
 
Maybe you can merge them.  Btw. I find the PC also very helpful in the log_deopt_message.

-- Chris

> 
> Testing: JPRT, simple test w/ deopt events in the log.
> 
> Best regards,
> Vladimir Ivanov
> 
> [1] Uncommon trap 3 fr.pc 0x00000001046ffe1c
> [2] Uncommon trap: Test::f @ 23, reason=unloaded action=reinterpret